Earth Surf Dynam 6 1101ndash1114 2018httpsdoiorg105194esurf-6-1101-2018copy Author(s) 2018 This work is distributed underthe Creative Commons Attribution 40 License

Geomorphic regulation of floodplain soil organic carbonconcentration in watersheds of the Rocky

and Cascade Mountains USA

Daniel N Scott and Ellen E WohlDepartment of Geosciences Colorado State University Fort Collins CO 80521 USA

Correspondence Daniel N Scott (danscottcolostateedu)

Received 16 March 2018 ndash Discussion started 10 April 2018Revised 5 October 2018 ndash Accepted 12 November 2018 ndash Published 23 November 2018

Abstract Mountain rivers have the potential to retain OC-rich soil and store large quantities of organic car-bon (OC) in floodplain soils We characterize valley bottom morphology floodplain soil and vegetation in twodisparate mountain river basins the Middle Fork Snoqualmie in the Cascade Mountains and the Big Sandy inthe Wind River Range of the Rocky Mountains We use this dataset to examine variability in OC concentrationbetween these basins as well as within them at multiple spatial scales We find that although there are somedifferences between basins much of the variability in OC concentration is due to local factors such as soilmoisture and valley bottom geometry From this we conclude that local factors likely play a dominant role inregulating OC concentration in valley bottoms and that interbasin differences in climate or vegetation character-istics may not translate directly into differences in OC storage We also use an analysis of OC concentration andsoil texture by depth to infer that OC is input to floodplain soils mainly by decaying vegetation not overbankdeposition of fine OC-bearing sediment Geomorphology and hydrology play strong roles in determining thespatial distribution of soil OC in mountain river corridors

1 Introduction

Terrestrial carbon storage plays an important role in regu-lating the global carbon cycle and the distribution of carbonbetween oceans the atmosphere long-term (105ndash109 years)storage in rock and short- to moderate-term storage in thebiosphere (101ndash104 years including vegetation and soil)(Aufdenkampe et al 2011 Battin et al 2009) Soils in par-ticular are a large organic carbon (OC) reservoir with signifi-cant spatial variability (Jobbaacutegy and Jackson 2000 Schmidtet al 2011) making them difficult to characterize in thecontext of global carbon cycling It is essential to quantifythe spatial variability of OC stored in the biosphere to con-strain the effects of climate change on feedbacks betweenbiospheric and atmospheric carbon storage (Ballantyne et al2012) To provide a more complete understanding of how thebiospheric carbon pool may change in the future and guidemanagement of soil OC we seek to provide a better con-

straint on where carbon is stored in the biosphere and theprocesses that regulate that storage

We focus here on river corridors defined as channels flu-vial deposits riparian zones and floodplains (Harvey andGooseff 2015) which process concentrate transport andstore carbon (Wohl et al 2017b) In the context of the car-bon cycle floodplains can act as a major component of thebiospheric carbon pool (Aufdenkampe et al 2011 Battin etal 2009) Floodplain soils can act as a substantial pool ofOC despite their relatively small aerial extent indicating thatfloodplains may be disproportionately important comparedto uplands in terms of carbon storage (DrsquoElia et al 2017Hanberry et al 2015 Sutfin et al 2016 Sutfin and Wohl2017 Wohl et al 2012 2017a) Mountainous regions dueto their high primary productivity (Schimel and Braswell2005 Sun et al 2004) may play a substantial role in thefreshwater processing and storage of OC whereby they re-tain sediment and water along the river network (Wohl et al2017b) Even laterally constrained floodplains in mountain-

Published by Copernicus Publications on behalf of the European Geosciences Union

1102 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

ous drainages can store significant quantities of OC that canbe mobilized during floods (Rathburn et al 2017) Howeverwe lack a comprehensive characterization of how floodplainsoil OC varies throughout watersheds It is important to un-derstand the spatial distribution of OC to predict its fate dur-ing floods and inform management to increase floodplain OCstorage (Bullinger-Weber et al 2014)

Floodplain OC enters river corridor soils via litterfall fromvegetation and the erosion of OC-bearing bedrock (Hilton etal 2011 Leithold et al 2016 Sutfin et al 2016) OC inputsare either allochthonous from upstream deposition of soilparticulate and dissolved OC or autochthonous from riparianvegetation (Omengo et al 2016 Ricker et al 2013 Sutfinet al 2016) As such OC input can be regulated by vegeta-tion dynamics and resulting litter input hydrologic and sed-iment transport regimes and water chemistry However it isunclear which OC inputs dominate under various conditionshampering the prediction of changes in OC delivery to soilsunder a changing climate that may have significant effects onvegetation dynamics and hydrology

OC concentration in soil is controlled by processes actingat multiple spatial scales At broad intra-basin scales OCconcentration in soil is regulated by the ability of carbon tosorb to soil particles and the ability of microbes and other or-ganisms to respire soil OC which can be controlled by rhizo-sphere dynamics moisture and temperature Sorption of OCto soil particles reduces OC lability and is controlled by grainsize and resulting available surface area as well as the avail-ability of calcium iron and aluminum (Kaiser and Guggen-berger 2000 Rasmussen et al 2018) Microbial respirationrepresents the primary pathway by which soil OC returns tothe atmosphere In general low temperatures and frequentsaturation inhibit microbial activity and promote OC stor-age (Falloon et al 2011 Jobbaacutegy and Jackson 2000 Sutfinet al 2016) At smaller interbasin scales the hydroclimaticregime controls vegetation dynamics moisture and temper-ature such that soil OC concentration in disparate regionscan be approximately characterized by these predictors (Auf-denkampe et al 2011 Schimel and Braswell 2005) How-ever at the scale of a single watershed hydrology ecologyand geomorphology play strong roles in determining soil tex-ture moisture and microbial dynamics in turn controllingOC storage in valley bottoms (Scott and Wohl 2017 Sutfinand Wohl 2017 Wohl and Pfeiffer 2018) We currently lacka comprehensive field-based examination of how processesacting at interbasin and intra-basin scales interact to regulatefloodplain soil OC concentrations

To address this knowledge gap we quantify spatial varia-tions in the OC concentration of floodplain soil across the en-tirety of two disparate mountain river networks This allowsus to examine interbasin hydroclimatic variation and intra-basin geomorphic and vegetation variation to understand themulti-scale controls on OC concentration We use this multi-scale approach to draw inferences regarding the spatial distri-

bution of floodplain OC controls on that spatial distributionand the dominant source of OC to mountain river floodplains

Objectives and hypotheses

Across a basin it is uncertain whether OC concentration infloodplain soils follows predictable longitudinal variation oris controlled by local factors Similarly in a vertical flood-plain soil profile it is uncertain whether OC concentrationfollows a trend similar to uplands with declining OC con-centration with depth or exhibits vertical heterogeneity as aresult of OC-rich layers deposited by floods It is also unclearwhether OC in floodplain soils is dominantly autochthonousor allochthonous Floodplain soil OC source may be evidentfrom the vertical heterogeneity of OC concentration domi-nantly autochthonous OC profiles should decline with depthwhereas dominantly allochthonous OC profiles should ex-hibit vertical heterogeneity reflecting episodic depositionOur primary objective here is to understand spatial variationsin OC concentration both with depth in a soil profile andacross a basin By quantifying these variations we hope toinfer the processes that regulate OC deposition in floodplainsoil

By examining two mountain river basins that differ interms of hydroclimatic regime and vegetation characteristicswe can quantify both interbasin variation in OC storage andvariation within each basin We hypothesize that at an in-terbasin scale the hydroclimatic regime and resulting rateof litterfall inputs in the riparian zone (Benfield 1997) willdominantly regulate OC concentration (H1) We define hy-droclimatic regime as the combination of precipitation andtemperature dynamics that result in the vegetation character-istics of a basin At an intra-basin scale we expect that val-ley bottom geometry and river lateral mobility will regulatefloodplain sediment characteristics and vegetation dynamicsThus we hypothesize that soil OC concentration does notvary along predictable longitudinal trends within mountainriver basins instead being more dominantly controlled by lo-cal fluvial processes and valley bottom form (H2a) We hy-pothesize that geomorphic process and form determine soiltexture and moisture which in turn set the boundary condi-tions that regulate the sorption of OC to mineral grains (pro-moting stabilization) and the potential of OC to be respiredby microbes (H2b) In terms of OC inputs to floodplain soilswe hypothesize that the source of OC is dominated by au-tochthonous vegetation and litter inputs in these basins (H3)As such we expect OC to dominantly decline with depthonly rarely exhibiting vertical heterogeneity that would rep-resent allochthonous deposition from flooding

2 Methods

This work was done alongside work presented in Scott andWohl (2018b) and hence shares field sites study design GISand sampling techniques

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1103

21 Field sites

We quantified soil organic carbon concentrations to a depthof approximately 1 m in the Big Sandy basin in the WindRiver Range of Wyoming and the Middle Fork Snoqualmiebasin in the central Cascade Mountains of Washington(Fig 1) These basins represent distinct bioclimatic and ge-omorphologic regions ranging from the wet high-relief Cas-cades to the semiarid moderate-relief Middle Rockies

The MF Snoqualmie has a mean annual precipitation of304 m (Oregon State University 2004) 2079 m of total re-lief over a 407 km2 drainage area and a mean basin slope of60 Topography in the MF Snoqualmie is largely glacio-genic with wide unconfined valleys at both high and lowelevations Streams range from steep debris-flow-dominatedheadwater channels to lower-gradient wide laterally uncon-fined channels in its lower reaches The lower reaches of theMF Snoqualmie have been clear-cut extensively since theearly 1900s although there is little logging activity todayVegetation follows an elevation gradient The talus activeglaciers and alpine tundra at the highest elevations transitionto subalpine forests dominated by mountain hemlock (Tsugamertensiana) (above approximately 1500 m) but also includ-ing Pacific silver fir (Abies amabilis) and noble fir (Abies pro-cera) in the lower subalpine and montane zones (above ap-proximately 900 m) Below the montane zone uplands andterraces are covered by Douglas fir (Pseudotsuga menziesii)and western hemlock (Tsuga heterophylla) whereas activeriparian zones are dominated by red alder (Alnus rubra) andbigleaf maple (Acer macrophyllum)

The semiarid Big Sandy is considerably drier than the MFSnoqualmie but also exhibits broad glacially carved valleysespecially in headwater reaches It has a mean annual precip-itation of 072 m (Oregon State University 2004) 1630 m oftotal relief over a 114 km2 drainage area and a mean basinslope of 25 The lower reaches of the Big Sandy are an-thropogenically impacted by moderate grazing use and an ac-cess road that crosses through part of the basin Herbaceousalpine tundra dominates higher elevations (above approxi-mately 3100 m) while the subalpine zone (approximately2900 to 3100 m) is characterized by forests of whitebarkpine (Pinus albicaulis) Engelmann spruce (Picea engelman-nii) and subalpine fir (Abies lasiocarpa) The montane zone(approximately 2600 to 2900 m) is comprised dominantlyof lodgepole pine (Pinus contorta) Only a small portion ofthis basin (approximately 1 ) resides below 2500 m whereshrub steppe begins to dominate (Fall 1994) Parklands andmeadows are abundant in this basin creating a patchy for-est structure Comparing this basin to the MF Snoqualmieprovides bioclimatic contrast that allows us to examine howfloodplain soil OC concentrations vary across a range ofstream morphologies and floodplain morphologic types in re-gions with differing precipitation forest characteristics andbasin morphology

To simplify comparison of these two basins we hence-forth refer to them by their dominant climate The MiddleFork Snoqualmie is the wet basin and the Big Sandy is thesemiarid basin (Fig 1)

22 Study design and sampling

We sampled the semiarid basin in summer 2016 and the MFSnoqualmie in summer 2017 During each sampling cam-paign no large floods occurred and we observed no flood-plain erosion or deposition Across both basins we cored atotal of 128 floodplain sites to determine soil OC concentra-tion Cores were collected as a series of individual soil sam-ples at both regular and irregular depth increments

221 Semiarid basin (Big Sandy)

The sparse vegetation in the semiarid basin enabled us touse a combination of a 10 m DEM and satellite imagery tomanually map the extent of the valley bottom along the en-tire stream network and delineate valley bottoms based onconfinement We defined unconfined valley bottoms as thosein which channel width occupied no more than half the val-ley bottom and confined valley bottoms as those in whichchannel width occupied greater than half the valley bottomWithin each confinement stratum we stratified the streamnetwork by five drainage area classes to produce a total of 10strata ensuring even sampling across the basin Within eachof the resulting 10 strata we randomly selected 5 reachesproducing a total of 50 sample sites throughout the basinDue to access issues we sampled 48 out of the 50 randomlylocated sites We supplemented these with 4 subjectively lo-cated sites that we felt enhanced our ability to capture vari-ation throughout the drainage based on observations in thefield resulting in a total of 52 sampled sites

222 Wet basin (Middle Fork Snoqualmie)

The wet basin is larger than the semiarid basin and has ex-tensive low-gradient floodplains in its downstream reachesThese extensive floodplains display high spatial variability invegetation surface water grain size and estimated surfaceage based on aerial imagery and ground reconnaissance Toensure an unbiased characterization of these heterogeneousfloodplains we used aerial imagery a 10 m DEM and pic-tures from field reconnaissance to delineate the floodplaininto patch categories fill channels (abandoned channels thathave had enough sediment deposited to prevent an oxbowlake from forming) point bars (actively accreting surfaceson the inside of bends) wetlands (areas with standing wa-ter in imagery that are not obviously oxbows) oxbow lakes(abandoned channels dammed at the upstream and down-stream ends to form a lake) and general floodplain surfaces(surfaces that cannot be classified into any of the above cat-

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1104 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

Figure 1 Hillshade showing the location sampling sites and stream network of the sampled basins Big Sandy Wyoming (4269minus10925) is on the left and MF Snoqualmie Washington (4753 minus12152) is on the right Circles represent sampling locations atwhich floodplain soil OC was measured Sample sites are colored by OC concentration Mean annual precipitation (MAP) drainage area(DA) and relief are given for each basin

egories) Within each of these five categories we randomlyselected six points at which to take soil cores

We also stratified the entire wet basin stream network bychannel slope into four strata Within each channel slopestrata (hereafter referred to as slope strata) we randomly se-lected 10 reaches to collect a single floodplain soil core re-sulting in 40 randomly located sample sites

To supplement randomly sampled sites and accommodatethe infeasibility of accessing two of the randomly sampledsites along the stream network we also subjectively selectedsample sites in places that we felt enhanced the degree towhich our sampling captured the variability present amongstreams in the basin This resulted in a total of 46 sites strat-ified by slope 38 of which were randomly sampled in addi-tion to 30 sites stratified by floodplain type

23 Reach-scale field measurements

At each sampled reach (100 m or 10 channel widthswhichever was shorter) we measured channel geometry andother characteristics although our measurements were notconsistent across all basins because field protocol evolvedduring the course of the study In both basins we measuredconfinement valley bottom width and channel bed slope Weadditionally measured bankfull width and depth in the wetbasin We did not measure channel characteristics for sitesstratified by floodplain type in the wet basin since they didnot correspond to a single reach of channel as sites stratifiedby slope did in much more confined valleys

In the wet basin we also categorized channels by planformand dominant bedform (Montgomery and Buffington 1997)We defined planforms as follows straight in which the chan-nel was generally confined and significant lateral migrationwas not evident meandering in which lateral migration wasevident but only a single channel existed anastomosing inwhich vegetated islands separated multiple channels and

anabranching in which a single dominant channel existedwith relict channels separated by vegetated islands We fur-ther classified channels as being either multithread (anasto-mosing or anabranching) or single thread (straight or mean-dering) Because logging records are inconsistent and likelyinaccurate in the wet basin (based on the frequent observa-tion of past logging activity that was not recorded in ForestService records) we noted whether signs of logging suchas cut stumps cable decommissioned roads or railroads orother logging-associated tools were found near the reach

We chose a representative location on the floodplain foreach sampled site based on visual examination of vegeta-tion type soil surface texture surface water presence andelevation relative to the bankfull channel elevation (flood-plain sites stratified by type in the wet basin were sampledas close to the randomly sampled point as possible) Oncea location was chosen we extracted a 32 mm diameter soilcore using an open-sided corer (JMC Large Diameter Sam-pling Tube) Due to our adaptive methodology we sampledsoil OC slightly differently in the semiarid versus the wetbasin In the semiarid basin we cored in irregular incrementsgenerally 25ndash30 cm After analyzing data from the semiaridbasin we realized that sampling in regular increments wouldmake analysis more versatile As a result we switched to ex-tracting soil samples at regular 20 cm increments in the wetbasin Cores were taken to refusal (ie coarse gravel or otherobstructions preventing further soil collection) or a depth ofapproximately 1 m Five cores in the semiarid basin 12 coresin the wet basin sites stratified by slope and 11 cores in thewet basin sites stratified by floodplain type did not reach re-fusal When no sand or finer sediment was present in the val-ley bottom (only occurred in headwater channels of the wetbasin) we recorded negligible OC concentration Once soilsamples were removed from the ground they were placed in

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1105

ziplock bags frozen within 72 h (most samples were frozenwithin 8 h) and kept frozen until analysis

24 Measuring soil OC and texture

To measure the concentration of organic carbon in soil sam-ples we used loss on ignition (LOI) We first defrosted sam-ples for 24ndash48 h at room temperature Once defrosted wethoroughly mixed samples to ensure the most homogenoussample possible We then subsampled 10ndash85 g of soil fromeach sample for analysis Using crucibles in a muffle furnacewe dried samples in batches of 30 for 24 h at 105 C to deter-mine moisture content and remove all nonstructurally heldwater Following the guidelines suggested by Hoogsteen etal (2015) we then burned samples for 3 h at 550 C to re-move organic matter By comparing the weight of the burnedsamples with that of the dried samples we obtained an LOIweight

After performing LOI we used burned samples to performtexture by feel to determine the USDA soil texture class andestimated clay content (Thien 1979) To convert LOI weightto OC concentration we used the structural water loss cor-rection of Hoogsteen et al (2015) using clay content esti-mated from soil texture This correction considers water heldby clay that may not evaporate during drying but will evap-orate during burning It also estimates the proportion of theLOI weight that is OC This correction represented as a per-cent of the estimated sample OC content after the correctionranged from 090 to 49576 (95 confidence interval ofthe median between 1617 and 2425 ) for the wet basinand ranged from 554 to 57053 (95 confidence in-terval of the median between 1976 and 3211 ) for thesemiarid basin

One potential confounding factor in LOI is carbonates thatmay burn off during ignition adding to the LOI weight whilenot being organic matter In lithologies where carbonates arerare (eg granitoid rocks like those found in the upper partof the wet basin and entire semiarid basins) this is a rela-tively negligible issue However some of our soil samplescame from parts of the wet basin draining rocks of the west-ern meacutelange belt including argillite greywacke and marbleWe tested samples for the presence of carbonates to deter-mine whether our LOI methods would be sufficient to accu-rately determine OC concentration We randomly chose 10soil samples of a total of 110 drained rocks that could includecarbonates and submitted them to the Colorado State Univer-sity soil testing laboratory for CHN furnace analysis (Sparks1996) which yielded data on the proportion of carbonatesby mass in those samples The median calcium carbonateconcentrations of those 10 samples was 098 (95 con-fidence interval between 070 and 103 ) and the medianpercentage of the total carbon in samples comprised of inor-ganic carbon was 505 (95 confidence interval between197 and 189 ) From this we concluded that the amountof carbonate in the samples draining potentially carbonate-

bearing rocks was low enough that LOI was likely to still beaccurate Consequently we analyzed all soil samples usingLOI to obtain OC concentration The median difference be-tween the LOI and CHN OC estimates for these 10 sampleswas 045 (95 confidence interval between minus083 and147 ) indicating no systematic bias in LOI estimates ofOC concentration

25 GIS and derivative measurements

After fieldwork in each basin we collected the following datafor each reach using a GIS platform elevation drainage arealand cover classification and canopy cover from the NationalLand Cover Database (Homer et al 2015) and the meanslope of the basin draining to each reach (including hill-slopes and channels) Utilizing drainage area at each reachand field-measured channel gradient we calculated an esti-mated stream power as the product of drainage area chan-nel gradient and basin-averaged precipitation We utilizeda 10 m DEM for all GIS topographic measurements To es-timate clay content for each sample we used median val-ues for assigned USDA texture classes To obtain estimatedclay content moisture and OC for each core we calculatedan average weighted by the percentage of core taken up byeach soil sample We categorized wet basin samples stratifiedby floodplain type into those with standing water (wetlandsand oxbow lakes) and those with no standing water (all othertypes)

26 Statistical analyses

All statistical analyses were performed using the R statisti-cal computing software (R Core Team 2017) We conductedall analyses on three modeling groups based on the variablesmeasured in each group In the wet basin we grouped obser-vations by stratification type separating observations strati-fied by channel slope from observations stratified by flood-plain type We separated these two groups from all obser-vations in the semiarid basin which were measured consis-tently We modeled OC concentration and soil texture with amixed-effects linear regression using individual soil samples(ie the individual samples that make up a core) as sam-ple units (n= 103 for wet basin stratified by slope 89 forwet basin stratified by floodplain type and 101 for semiaridbasin) We modeled the sampled site as a random effect ac-knowledging that individual soil samples within a single coreare likely nonindependent We used profiled 95 confidenceintervals on effect estimates (β) for fixed effects to evaluatevariable importance in mixed-effects models

To gain further insight at the reach scale we also mod-eled average OC concentration and soil moisture at eachmeasured site using multiple linear regression We mod-eled soil moisture at the reach scale because we felt thatour single snapshot of moisture conditions was better rep-resented as a site-level average We first performed uni-

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1106 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

variate analysis between each hypothesized predictor andthe response utilizing mainly comparative Wilcoxon rank-sum tests (Wilcoxon 1945) or correlational Spearman cor-relation coefficient statistics We utilized a Holm multiple-comparison correction (Holm 1979) for pairwise compar-isons During this filtering we also viewed box plots or scat-terplots as appropriate to discern which variables appear tohave anything other than a completely random relationshipwith the response We then utilized the multiple linear re-gression of all subsets using the corrected Akaike informa-tion criterion as a model selection criterion (Wagenmakersand Farrell 2004) We iteratively transformed response vari-ables to ensure the homoscedasticity of error terms To selecta single best model we utilized both Akaike-weight-basedimportance and parsimony to select a final reduced modelWe considered sample sizes p values and effect magnitudesin determining variable importance

We also analyzed each core to determine whether therewere buried high-OC-concentration layers at depth We com-pared each buried soil sample to the sample above it usingthe criterion that a peak in OC at depth should have an OCconcentration 15 times that of the overlying sample and beabove 05 (Appling et al 2014)

3 Results

Model results are presented in Table 1 Comparisons betweenbasins and summaries of OC concentration moisture andestimated clay content are shown in Fig 2

31 OC concentration

Most cores display a decrease in OC concentration withdepth (Fig 3) Of cores with more than a single sample32 (722) of cores stratified by slope in the wet basin 32 (825) of cores stratified by floodplain type in the wet basinand 6 (231) of cores in the semiarid basin exhibit OC con-centration peaks at depth Whether a soil sample was classi-fied as an OC peak has no relation to estimated clay contentin sites stratified by floodplain type (p = 028) or those strat-ified by slope (p = 089) in the wet basin In the semiaridbasin soil samples classified as buried OC peaks have signif-icantly higher estimated clay contents (p = 005) than thosethat were not classified as peaks

In general the floodplain-stratified sites in the wet basinstore higher densities of OC than the semiarid basin (Fig 2ab) Figure 2a includes zero values (ie sites with no OC-bearing sediment only present in the wet basin slope-stratified group) whereas Fig 2b does not because sampleunits in Fig 2b are individual soil samples Comparing thesetwo groups it appears that soils in the wet basin exhibit muchhigher OC concentrations than those in the semiarid basinbut in general there are many more reaches with no fine sed-iment available to store OC in the wet basin

At the scale of individual soil samples the depth belowground surface is by far the dominant control on OC con-centration across all modeling groups We used a cube roottransform for all three mixed-effects models of OC concen-tration For wet basin sites stratified by slope deeper soilsamples contain less OC (β =minus00084plusmn 00042) whereassoil samples at higher elevations tend to contain more OC(β = 00010plusmn000099) Depth is the only significant predic-tor of OC content for both wet basin soil samples stratifiedby floodplain type (β =minus00084plusmn00042) and soil samplesin the semiarid basin (β =minus00037plusmn 00019)

Modeling wet basin slope-stratified sites at the reach scalewe found that moisture (β = 00078plusmn 00031) and whetherthe reach was unconfined (β = 077plusmn 049) control soil OC(cube root transformation model adjusted R2

= 054 p lt00001) Modeling wet basin floodplain-stratified sites at thesite scale we found that canopy cover (β = 0012plusmn 0011)and moisture (β = 00040plusmn 00011) are controls on soilOC (cube root transformation model adjusted R2

= 067p lt 00001) Modeling Big Sandy sites at the reach scalewe found that soil depth (β =minus0012plusmn 00071) and mois-ture (β = 0014plusmn 00026) are dominant controls on soil OCconcentration (no transformation model adjustedR2

= 069p lt 00001)

In general moister deeper soils store more OC at the reachscale whereas OC tends to vary dominantly with depth atthe scale of individual soil samples Although estimated claycontent did not emerge as a significant predictor of OC con-centration it is used to calculate clay-held water to correctour LOI-based OC concentration measurements making itimportant in determining OC for each sample

32 Soil texture

In general soil texture follows a predictable trend withriver size between model groups (Fig 2d) Floodplain type-stratified sites in the wet basin store the most clay followedby slope-stratified sites and then sites in the semiarid basin

Modeling soil texture at the individual soil sample scaleacross slope-stratified sites in the wet basin we foundwhether the reach was confined (β = 542plusmn 532) andwhether the bed material was dominantly sand (β = 1047plusmn613) to be dominant controls on estimated clay contentModeling soil texture for sites stratified by floodplain typein the wet basin yielded no significant trends In the semi-arid basin we found that either valley width (β = 00050plusmn00032) or whether the stream was unconfined (β = 041plusmn033) as well as depth below ground surface (β =minus00069plusmn00033 for model with valley width but not confinement) sig-nificantly control soil texture

To summarize sites from unconfined lower-energyreaches in the wet basin and sites from reaches with widervalley bottoms and at lower depths in the semiarid basin ex-hibited more finely textured soils

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1107

Tabl

e1

Mat

rix

ofal

lmod

els

pres

ente

din

the

text

Eac

hm

odel

islis

ted

bym

odel

grou

pre

spon

seva

riab

lea

ndsc

ale

Scal

ere

fers

toth

esa

mpl

eun

itof

the

mod

ela

ndsi

tere

fers

toa

core

with

the

resp

onse

aver

aged

over

allt

hein

divi

dual

soil

sam

ples

inth

eco

reF

orea

chva

riab

lean

dm

odel

an

aste

risk

()i

ndic

ates

that

the

vari

able

was

incl

uded

inm

odel

sele

ctio

nbu

tnot

the

full

mix

ed-e

ffec

tsm

odel

Am

inus

(minus)

indi

cate

sth

atth

eva

riab

lew

asse

lect

edas

impo

rtan

tin

pred

ictin

gth

ere

spon

sean

dde

note

san

indi

rect

corr

elat

ion

whe

reas

apl

us(+

)ind

icat

esa

dire

ctco

rrel

atio

nIn

the

case

ofco

nfine

men

ta

plus

indi

cate

sth

atun

confi

ned

stre

ams

disp

lay

ahi

gher

-mag

nitu

dere

spon

seva

riab

leI

nth

eca

seof

bed

mat

eria

la

plus

indi

cate

sth

atsa

mpl

esw

ithsa

ndex

hibi

tahi

gher

valu

eof

the

resp

onse

NA

indi

cate

sth

atei

ther

the

vari

able

was

notm

easu

red

for

that

basi

nor

mod

elgr

oup

orth

atit

isth

em

odel

resp

onse

Mod

elgr

oup

Res

pons

eSc

ale

(sam

ple

unit)

Var

iabl

esConfinement

Bedform

Channelslope

Bedmaterial

Multithread

Valleywidth

Bankfullwidth

Bankfulldepth

Streampower

Floodplaintype

Standingwater

Depth1

Claycontent

Moisture

Loggingnearby

Grassespresent

Shrubspresent

Treespresent

Elevation

Basinslope

Canopycover

NLCD

Drainagearea

Wet

basi

nst

rat-

ified

bysl

ope

OC

()

Soil

sam

ple

NA

NAminus

+

O

C(

)Si

te+

NA

NA

+

Moi

stur

e(

)Si

te+

minus

NA

NA

N

A+

Te

xtur

e(

)So

ilsa

mpl

e+

+

N

AN

A

NA

Wet

basi

nst

rat-

ified

byflo

od-

plai

nty

pe

OC

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

minus

N

AO

C(

)Si

teN

AN

AN

AN

AN

AN

AN

AN

AN

A

+

+N

AM

oist

ure

()

Site

NA

NA

NA

NA

NA

NA

NA

NA

NA

+

+

NA

NA

Text

ure2

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

N

AN

A

Sem

iari

dba

sin

OC

()

Soil

sam

ple

N

A

NA

NA

N

AN

A

NA

NAminus

NA

NA

NA

NA

OC

()

Site

N

A

NA

NA

NA

NA

N

AN

Aminus

+N

AN

AN

AN

A

M

oist

ure

()

Site

+N

A

NA

NA

NA

NA

N

AN

A+

NA

NA

NA

NA

NA+

Te

xtur

e3(

)So

ilsa

mpl

e+

NA

N

AN

A+

NA

NA

N

AN

Aminus

NA

NA

NA

NA

NA

1D

epth

refe

rsto

eith

erth

eso

ilsa

mpl

ede

pth

belo

wth

egr

ound

orth

eto

tald

epth

ofth

eco

red

epen

ding

onth

esa

mpl

eun

it2

No

sign

ifica

ntre

sults

wer

eob

serv

edfo

rthi

sm

odel

3Fo

rthi

sm

odel

bot

hva

lley

wid

than

dco

nfine

men

tpre

dict

text

ure

and

can

bein

terp

rete

din

terc

hang

eabl

yH

owev

eri

nclu

ding

both

inth

esa

me

mod

elw

ould

yiel

dpr

oble

ms

due

tom

ultic

ollin

eari

ty

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1108 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

Figure 2 Box plots showing comparisons between model groups of OC concentration at the reach scale (a) OC concentration at the scaleof individual soil samples (b) moisture at the reach scale (c) and estimated clay content at the scale of individual soil samples (d) Endsof dotted lines represent 15 times the interquartile range which is represented by boxes Bold line represents median Circles representoutliers Letters indicate probable differences between groups based on pairwise Wilcoxon (andashc) or t tests (d) with a Holm correctionRanges in parentheses below letters show the 95 confidence interval of the median value for the group (andashc) or the mean value for thegroup (d) in which median confidence intervals were overly constrained due to the categorical nature of our estimated clay content data

33 Soil moisture

Soil moisture is less variable between basins than either tex-ture or OC concentration (Fig 2c) All model groups exhibitsimilar soil moisture conditions although there was signifi-cant variability within each model group

Soil moisture at wet basin sites stratified by slope is domi-nantly controlled by channel slope (β =minus1315plusmn911) ele-vation (β = 00046plusmn 00038) and whether the stream is un-confined (β = 389plusmn 267 model adjusted R2

= 038 p lt00001) At wet basin sites stratified by floodplain type es-timated clay content (β = 0060plusmn 0042) and whether thefloodplain unit had standing water (β = 115plusmn 071) sig-nificantly control soil moisture (model adjusted R2

= 046p lt 00001) In the semiarid basin soil depth (β = 0012plusmn0012) elevation (β = 00023plusmn 00014) and whether thereach was unconfined (β = 091plusmn075) significantly controlsoil moisture (model adjusted R2

= 035 p lt 00001)

4 Discussion

41 Understanding spatial variability in OCconcentration in floodplain soils (H1 and H2)

Comparing the wet basin to the semiarid basin shows thatthe wetter higher primary productivity basin is capable ofstoring greater concentrations of OC in floodplain soils butthat both regions generally store similar OC concentrations infloodplain soils This result partially agrees with the exami-nation of subalpine lake deltas by Scott and Wohl (2017) Inthat study subalpine lake deltas in the wet basin were com-pared to deltas in the drier Colorado Front Range Subalpinelake deltas displayed similar OC concentrations likely dueto competing but complementary OC stabilization and lossmechanisms in each region Those deltas represent a subsetof the broader valley bottom soils studied here This moreexpansive study points to both geomorphic controls such asvalley bottom geometry and factors influenced by climatesuch as canopy cover as controls on OC storage in valley bot-toms These results also agree with the results of Lininger et

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1109

Figure 3 Box plots of sample OC concentration binned by meansample depth for the wet (a) and semiarid (b) basins Ends of dot-ted lines represent 15 times the interquartile range which is rep-resented by boxes Bold line represents median Black circles rep-resent outliers Transparent grey points show all data for each binSample size for each bin is denoted by n

al (2018) which indicate that geomorphic context and vege-tation dynamics control OC concentration on floodplain soilsalong large lowland rivers in Alaska USA

At the reach or site scale wetter soil profiles consistentlyyield higher OC concentrations in all model groups How-ever moisture does not differ significantly between modelgroups (Fig 2c) indicating that this alone cannot explaindifferences between basins Soils tend to be finer in the wetbasin but clay content was not an important predictor of OCconcentration in studied soils Clay content likely influencesOC concentration based on previous research (Hoffmann etal 2009) although the inclusion of coarse soil material (in-cluding particulate organic matter) in our samples may ex-plain the lack of an observed correlation here Variabilityin soil redox conditions (ie the presence of anaerobic mi-crosites) may also introduce variability in respiration rate be-tween basins and among individual soil samples (Keiluweit

et al 2017) That is variability in groundwater hydrologymay result in a heterogeneous mix of aerobic (favoring mi-crobial activity) and anaerobic (suppressing microbial activ-ity) conditions within floodplain soils We did not record re-doximorphic features of our soil samples so we are unableto determine whether this influence contributes to the errorin our models Although confinement plays a strong role indetermining OC concentration in wet basin sites stratified byslope it does not differ significantly between basins (52 of semiarid basin reaches are unconfined compared to 63 of wet basin reaches) The major differences between thesebasins are their hydroclimatic and disturbance regimes Thewet basin is at a lower elevation and has denser and higherbiomass forests (Smithwick et al 2002) compared to thesparser parkland forests of the semiarid basin which likelyalso experiences more frequent fires based on fire historiesof nearby regions (recurrence interval on the order of 101ndash102 years Houston 1973 Loope and Gruell 1973) In addi-tion the volcanic soils present in the wet basin may suppresssoil OC respiration (Matus et al 2014) leading to a highersoil OC storage capacity there

Between basins it is likely that the hydroclimatic regimeinfluencing primary production plays some role in the wetbasinrsquos higher maximum OC concentrations in floodplainsoils compared to those of the semiarid basin Howeversmaller-scale factors such as soil texture and moisture alsolikely play a role and are not related to drainage area (Ta-ble 1) indicating that neither OC concentration nor its con-trolling factors vary continuously along a river network thussupporting H2a and H2b This also indicates that local fac-tors set largely by geomorphic and hydrologic dynamicsplay a significant role in modulating the effect of climate onOC concentrations If the wet and semiarid basins displayedsignificantly different OC concentrations our first hypothesisregarding the interbasin controls on OC concentration wouldbe supported However we instead found that climate andprimary productivity only partially determine OC concentra-tions especially when viewed in the context of geomorphicand hydrologic variability Thus the results do not supportH1

Each basin (or model group) is slightly different in termsof the controls on soil OC concentration moisture andtexture In the wet basin sites stratified by slope higher-elevation sites display higher OC concentrations This is con-trary to the general trend in primary productivity which de-creases with increasing elevation However it is importantto note that the headwaters of the wet basin are dominatedby lakes deltas and other depositional features in relativelybroad glacially carved valleys Subalpine lake deltas havebeen shown to store high OC concentrations in this basin(Scott and Wohl 2017) and many of the highest OC concen-trations we measured were located in broad wet meadowssubalpine lake deltas or other unconfined high-elevationreaches Such unconfined sites also likely have significantlycooler temperatures and tend to have higher soil moisture

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1110 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

contents as shown by our modeling (Table 1) As such al-though high-elevation wet basin sites may receive less OCinput they likely have a low rate of OC respiration resultingin higher OC concentrations on the whole which agrees withthe result of Bao et al (2017) In the semiarid basin our mod-eling suggests that the lower temperatures and higher mois-ture (Table 1) at higher elevations do not compensate for thelower primary productivity as elevation does not correlatewith OC concentration

In both basins unconfined reaches contain wetter andfiner-textured soils which may result in a higher soil OCcapacity Although confinement only relates directly to OCcontent in wet basin sites it does play a strong role in deter-mining moisture which in turn plays a role in regulating OCconcentration in both basins likely via inhibiting microbialactivity (Howard and Howard 1993) The relevance of chan-nel slope in determining soil moisture in the wet basin but notsemiarid basin may reflect the prevalence of high-gradientdebris-flow-dominated channels in the wet basin that largelyexhibit only gravel to boulder substrate which we assumestores minimal fine sediment moisture or OC

In the semiarid basin higher soil depths correlate withmoister and finer-textured soils but less OC concentrationThis indicates the trend in OC with depth likely dominatesthe signal of OC concentration with deeper sites containinga higher proportion of OC-depleted deep samples

42 Inferring sources of OC to floodplain soils (H3)

OC can be input to floodplain soils by two primary mech-anisms First dissolved and particulate OC (including largewood) can be deposited on floodplain surfaces by overbankdeposition thus integrating fluvial sedimentary OC into thefloodplain soil profile or in the case of large wood deposit-ing discrete but potentially large concentrations of OC thatcan later be integrated into the soil profile Second litter anddecomposing vegetation on the floodplain surface can act asautochthonous inputs of OC to floodplain soil

Our modeling of OC concentration yielded results consis-tent with previous investigations of controls on soil OC stor-age capacity (Jobbaacutegy and Jackson 2000 Sutfin and Wohl2017) Sites in the heterogeneous floodplain of the wet basindisplay a direct correlation between canopy cover and OCconcentration indicating that increased litter inputs lead toincreased floodplain soil OC concentration Sediment inputslikely differ between floodplain depositional unit types (egcoarser sediment may deposit on point bars compared tofilled secondary channels) although floodplain type does notpredict OC concentration This indicates that vegetation in-puts may be more dominant than fluvial sediment inputs atthese sites

The finding that buried OC peaks in the wet basin do nothave abnormally high clay contents supports the interpre-tation that wood and litter inputs to soil are the dominantsource of OC in the floodplain soils we examined Buried

peaks can be layers created by overbank deposition and sub-sequent burial of fine OC-bearing sediments (Blazejewski etal 2009 Ricker et al 2013) buried pieces of wood (Wohl2013) or buried organic horizons that are now capped bysediments that prevent OC respiration If overbank deposi-tion of fine sediment caused OC peaks we would expect tosee the soil samples classified as peaks exhibiting high claycontents indicating finer sediment Instead our results sug-gest that in the wet basin buried peaks are likely the result ofeither buried organic horizons or buried wood We observedlarge pieces of decaying buried wood in floodplain cut banksin the wet basin supporting this inference Overbank depo-sition of wood on the floodplain was only observed rarely inthis basin indicating that the OC measured in these soils islikely dominantly autochthonous

In the semiarid basin the two cores that exhibit peaks werecollected from the same meadow just downstream of a now-filled former lake that is a potential source of fine sedimentThe channels draining this meadow exhibit an anabranchingplanform indicating the potential to deposit and bury packetsof potentially OC-rich fine sediments However the majorityof cores do not exhibit OC peaks indicating OC input mainlyfrom vegetation at the surface and continuing OC respirationat depth

OC variation within each core is dominantly a function ofdepth We observe a negative correlation between depth be-low ground surface and OC concentration which has beenobserved in other studies including mountain wetlands andfloodplains (Jobbaacutegy and Jackson 2000 Scott and Wohl2017 Sutfin and Wohl 2017 Zhao et al 2017) In generalthis indicates that at least in mountain river floodplains OCis enriched at the surface and decomposes with depth sim-ilar to upland soils This fits with our finding that the ma-jority of our cores do not exhibit significant OC peaks atdepth and supports the dominance of litter and wood OC in-puts to floodplain soils These results support our hypothesisthat decaying litter and wood not overbank sediment depo-sition dominate the input of OC to floodplain soils in ourstudy basins (H3) We note that floodplain wood may alsoact as a trapping site for overbank fine organic matter facil-itating the deposition and input of OC from decaying vege-tation While these study basins likely accumulate soil OCmainly autochthonously other basins that experience over-bank flows accompanying deposition of fine sediment andburial of organic layers exhibit OC storage that is likely dom-inated by fluvial sediment deposition (eg Blazejewski etal 2009 DrsquoElia et al 2017 Ricker et al 2013) Thus itis likely that flow regime lateral connectivity and sedimenttransport dynamics regulate whether floodplain soil OC isdominantly input by the overbank deposition of fine mate-rial or litter and wood decay

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1111

Figure 4 Conceptual model of physical processes that influenceOC concentration in floodplain soils based on this study and otherliterature Each box corresponds to a major factor that influencesOC concentration Colored text within each box denotes factors thatinfluence OC inputs sorption capacity OC export or respirationrate (note that some processes influence both OC export and respi-ration rate) As sorption capacity increases so does the OC capacityof the soil Conversely as the rate of respiration andor OC exportincreases the soil OC capacity decreases Floodplain soils can onlydevelop high concentrations of OC if there are high rates of OCinput However the capacity of the soils to store OC regulates thatinput and is determined by the competing influences of sorption ca-pacity and the combination of respiration rate and OC export Seethe text for further details

43 Conceptual model of soil OC concentration infloodplain soils

We present a conceptual model to summarize our results andplace them in the context of work examining the controlson OC storage in soils (Fig 4) OC is input to floodplainseither through the decay of vegetation or the deposition offine OC-rich sediment This input of OC only determinesOC concentrations insofar as floodplain soils are capable ofstoring OC That storage is effectively determined by a bal-ance between processes that remove OC from floodplainsnamely respiration or erosion followed by respiration (Berheet al 2007) and processes that regulate OC availability tomicrobes namely the capability of the mineral fraction ofthe soil to sorb OC

OC sorption capacity reflects a few specific processes Al-though soil texture generally relates to the ability of OCto sorb to mineral grains and resulting OC availability soilchemistry also plays a strong and potentially dominant rolein regulating OC sorption capacity (Rasmussen et al 2018)Soil texture is largely determined by valley morphology ac-cording to our modeling (Table 1) placing valley morphol-ogy and resulting sediment transport dynamics (Gran andCzuba 2017 Wohl et al 2017b) as indirect controls on sorp-tion capacity

Respiration rate is largely determined by microbial activityand the availability of OC to microbes Erosion can rapidlyexpose soil OC to microbial respiration (Berhe et al 2007)whereas soils that reside in largely anoxic conditions can ex-hibit low rates of microbial respiration (Boye et al 2017)In addition soil mineralogy chemistry and redox condi-tions (Keiluweit et al 2017) can regulate microbial activity(eg andic soils may limit microbial respiration Matus et al2014) Our results suggesting that moisture controls OC con-tent support the idea that drier soils likely have higher ratesof microbial respiration of OC Moisture is a function of tex-ture valley bottom morphology and elevation (a proxy fortemperature) in our modeling (Table 1) Comparing flood-plain types in the wet basin we find that types with standingwater exhibit significantly higher soil moisture contents thanthose without standing water This indicates spatial variabil-ity in moisture content and likely microbial activity (Howardand Howard 1993) In our modeling this effect translates tospatial variability in OC concentration within floodplains andacross entire basins

OC export from soils refers specifically to leaching of dis-solved OC mainly from shallow soils While we do not di-rectly consider OC export in this study it likely introducesvariability in soil OC concentration and is regulated dom-inantly by hydrology (Aringgren et al 2014 Mcdowell andLikens 1988) soil pore water chemistry (including pH andionic strength Brooks et al 1999 Evans et al 2006) andredox conditions (Knorr 2013) OC export as DOC can actin conjunction with respiration to remove OC from soils andlike respiration rate is countered by sorption capacity

To summarize OC inputs regulated by the capacity ofsoils to store OC and suppress microbial respiration andOC export determine OC concentrations in floodplain soils(Fig 4) OC inputs to floodplain soils come from either au-tochthonous litter accumulation on the floodplain surface al-lochthonous wood deposition or allochthonous deposition offine OC-bearing sediments Our results from the mountain-ous basins we studied suggest that deposition of fine mate-rial in overbank flows is rare leading us to infer that au-tochthonous litter and allochthonous wood inputs to flood-plains dominate OC input in mountain rivers Where soils aremore moist microbial respiration is inhibited and more OCis stored Although soil texture is likely not a limiting factoron OC concentration in these floodplains finer-textured soilslikely have a higher sorption capacity retaining more of theOC input from decaying plant material Our results indicatethat geomorphic and hydrologic characteristics act as bound-ary conditions that regulate soil texture and moisture in turnregulating sorption capacity respiration rate and resultingOC concentrations in floodplain soils

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1112 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

5 Conclusions

We present floodplain soil OC concentration data from twodisparate watersheds to compare how interbasin variabilitybetween the two watersheds compares with intra-basin vari-ability in geomorphic and hydrologic characteristics in de-termining OC concentration Our results indicate that OCconcentration in mountain floodplain soils does not vary pre-dictably along a longitudinal gradient nor does it vary sub-stantially between basins with differing climatic and veg-etation characteristics Instead geomorphic and hydrologiccharacteristics such as valley bottom morphology and soilmoisture dominantly determine floodplain OC concentra-tion

In our study basins decaying litter and wood and not over-bank deposition of fine OC-bearing sediment is the mainsource of OC to floodplain soils It is unclear whether thatdecaying vegetation is dominated by autochthonous litter in-puts or transported downed wood In comparing our basinto other studied floodplain soils it seems that vegetation dy-namics play a strong role in determining OC concentrationswhen fine sediment is not regularly deposited on floodplainsurfaces However we suggest that floodplain soil character-istics set by geomorphic and hydrologic conditions regu-late how OC inputs translate to the spatial distribution of OCalong a river network

This implies that OC storage in floodplains likely cannotbe predicted using consistent downstream trends and thatmanagement prioritization designed to facilitate floodplainOC storage should be based on local geomorphic and hy-drologic process variability within each basin For instancemanagement to increase OC sequestration in floodplain soilswill likely be more effective where floodplains are uncon-fined and soils already experience high moisture conditionsfor much of the year Along these lines our results showthat modeling the floodplain biospheric OC pool to predictits response to warming and subsequent effects on climatebased on regional factors such as climate and net primaryproductivity likely misses the substantial interbasin variabil-ity in OC concentration and storage resulting from variabilityin valley bottom geometry and both geomorphic and hydro-logic processes (eg Doetterl et al 2015)

Although our results provide some insights the questionof whether OC stored in floodplain soil comes dominantlyfrom allochthonous versus autochthonous sources remainsopen Our results imply that more productive spatially het-erogeneous floodplains likely input more OC to soils Flood-plain OC concentration while mediated largely by moisturedynamics likely depends mainly on OC inputs from produc-tive riparian forests This implies that management of OCstorage in mountain river floodplains should focus on therestoration of riparian zones to maintain OC input to soil(eg Bullinger-Weber et al 2014) More detailed studiesin regions with varying sediment transport and hydrologicregimes are needed to determine what conditions favor au-

tochthonous versus allochthonous OC inputs but our resultssuggest that autochthonous sources dominate floodplain OCstorage in basins with relatively low rates of vertical accre-tion and high channelndashfloodplain connectivity that promotesfloodplain wetlands

Data availability Data to support the analyses presented here canbe found in Table S1 and in the Colorado State University DigitalRepository (Scott and Wohl 2018a)

Supplement The supplement related to this article is availableonline at httpsdoiorg105194esurf-6-1101-2018-supplement

Author contributions Field data collection was conducted pri-marily by DNS with assistance from EEW at some sites in theMiddle Fork Snoqualmie basin DNS performed data analysis anddrafted the initial version of the paper which was then edited andrewritten by DNS and EEW

Competing interests The authors declare that they have no con-flict of interest

Acknowledgements This work was funded by NSF grantEAR-1562713 We thank Ellen Daugherty for extensive assistancein fieldwork and discussion that improved the conceptual modelWe thank Katherine Lininger for stimulating discussion thatimproved the paper Detailed and constructive comments from twoanonymous reviewers and Robert Hilton improved the paper Wethank Sara Lowe for assistance in processing and analyzing soilsamples

Edited by Robert HiltonReviewed by two anonymous referees

References

Aringgren A M Buffam I Cooper D M Tiwari T Evans C Dand Laudon H Can the heterogeneity in stream dissolved or-ganic carbon be explained by contributing landscape elementsBiogeosciences 11 1199ndash1213 httpsdoiorg105194bg-11-1199-2014 2014

Appling A P Bernhardt E S and Stanford J A Flood-plain biogeochemical mosaics A multi-dimensional view ofalluvial soils J Geophys Res-Biogeo 119 1538ndash1553httpsdoiorg1010022013JG002543 2014

Aufdenkampe A K Mayorga E Raymond P A MelackJ M Doney S C Alin S R Aalto R E and YooK Riverine coupling of biogeochemical cycles between landoceans and atmosphere Front Ecol Environ 9 53ndash60httpsdoiorg101890100014 2011

Ballantyne A P Alden C B Miller J B Tans P P and WhiteJ W C Increase in observed net carbon dioxide uptake by

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

1102 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

ous drainages can store significant quantities of OC that canbe mobilized during floods (Rathburn et al 2017) Howeverwe lack a comprehensive characterization of how floodplainsoil OC varies throughout watersheds It is important to un-derstand the spatial distribution of OC to predict its fate dur-ing floods and inform management to increase floodplain OCstorage (Bullinger-Weber et al 2014)

Floodplain OC enters river corridor soils via litterfall fromvegetation and the erosion of OC-bearing bedrock (Hilton etal 2011 Leithold et al 2016 Sutfin et al 2016) OC inputsare either allochthonous from upstream deposition of soilparticulate and dissolved OC or autochthonous from riparianvegetation (Omengo et al 2016 Ricker et al 2013 Sutfinet al 2016) As such OC input can be regulated by vegeta-tion dynamics and resulting litter input hydrologic and sed-iment transport regimes and water chemistry However it isunclear which OC inputs dominate under various conditionshampering the prediction of changes in OC delivery to soilsunder a changing climate that may have significant effects onvegetation dynamics and hydrology

OC concentration in soil is controlled by processes actingat multiple spatial scales At broad intra-basin scales OCconcentration in soil is regulated by the ability of carbon tosorb to soil particles and the ability of microbes and other or-ganisms to respire soil OC which can be controlled by rhizo-sphere dynamics moisture and temperature Sorption of OCto soil particles reduces OC lability and is controlled by grainsize and resulting available surface area as well as the avail-ability of calcium iron and aluminum (Kaiser and Guggen-berger 2000 Rasmussen et al 2018) Microbial respirationrepresents the primary pathway by which soil OC returns tothe atmosphere In general low temperatures and frequentsaturation inhibit microbial activity and promote OC stor-age (Falloon et al 2011 Jobbaacutegy and Jackson 2000 Sutfinet al 2016) At smaller interbasin scales the hydroclimaticregime controls vegetation dynamics moisture and temper-ature such that soil OC concentration in disparate regionscan be approximately characterized by these predictors (Auf-denkampe et al 2011 Schimel and Braswell 2005) How-ever at the scale of a single watershed hydrology ecologyand geomorphology play strong roles in determining soil tex-ture moisture and microbial dynamics in turn controllingOC storage in valley bottoms (Scott and Wohl 2017 Sutfinand Wohl 2017 Wohl and Pfeiffer 2018) We currently lacka comprehensive field-based examination of how processesacting at interbasin and intra-basin scales interact to regulatefloodplain soil OC concentrations

To address this knowledge gap we quantify spatial varia-tions in the OC concentration of floodplain soil across the en-tirety of two disparate mountain river networks This allowsus to examine interbasin hydroclimatic variation and intra-basin geomorphic and vegetation variation to understand themulti-scale controls on OC concentration We use this multi-scale approach to draw inferences regarding the spatial distri-

bution of floodplain OC controls on that spatial distributionand the dominant source of OC to mountain river floodplains

Objectives and hypotheses

Across a basin it is uncertain whether OC concentration infloodplain soils follows predictable longitudinal variation oris controlled by local factors Similarly in a vertical flood-plain soil profile it is uncertain whether OC concentrationfollows a trend similar to uplands with declining OC con-centration with depth or exhibits vertical heterogeneity as aresult of OC-rich layers deposited by floods It is also unclearwhether OC in floodplain soils is dominantly autochthonousor allochthonous Floodplain soil OC source may be evidentfrom the vertical heterogeneity of OC concentration domi-nantly autochthonous OC profiles should decline with depthwhereas dominantly allochthonous OC profiles should ex-hibit vertical heterogeneity reflecting episodic depositionOur primary objective here is to understand spatial variationsin OC concentration both with depth in a soil profile andacross a basin By quantifying these variations we hope toinfer the processes that regulate OC deposition in floodplainsoil

By examining two mountain river basins that differ interms of hydroclimatic regime and vegetation characteristicswe can quantify both interbasin variation in OC storage andvariation within each basin We hypothesize that at an in-terbasin scale the hydroclimatic regime and resulting rateof litterfall inputs in the riparian zone (Benfield 1997) willdominantly regulate OC concentration (H1) We define hy-droclimatic regime as the combination of precipitation andtemperature dynamics that result in the vegetation character-istics of a basin At an intra-basin scale we expect that val-ley bottom geometry and river lateral mobility will regulatefloodplain sediment characteristics and vegetation dynamicsThus we hypothesize that soil OC concentration does notvary along predictable longitudinal trends within mountainriver basins instead being more dominantly controlled by lo-cal fluvial processes and valley bottom form (H2a) We hy-pothesize that geomorphic process and form determine soiltexture and moisture which in turn set the boundary condi-tions that regulate the sorption of OC to mineral grains (pro-moting stabilization) and the potential of OC to be respiredby microbes (H2b) In terms of OC inputs to floodplain soilswe hypothesize that the source of OC is dominated by au-tochthonous vegetation and litter inputs in these basins (H3)As such we expect OC to dominantly decline with depthonly rarely exhibiting vertical heterogeneity that would rep-resent allochthonous deposition from flooding

2 Methods

This work was done alongside work presented in Scott andWohl (2018b) and hence shares field sites study design GISand sampling techniques

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1103

21 Field sites

We quantified soil organic carbon concentrations to a depthof approximately 1 m in the Big Sandy basin in the WindRiver Range of Wyoming and the Middle Fork Snoqualmiebasin in the central Cascade Mountains of Washington(Fig 1) These basins represent distinct bioclimatic and ge-omorphologic regions ranging from the wet high-relief Cas-cades to the semiarid moderate-relief Middle Rockies

The MF Snoqualmie has a mean annual precipitation of304 m (Oregon State University 2004) 2079 m of total re-lief over a 407 km2 drainage area and a mean basin slope of60 Topography in the MF Snoqualmie is largely glacio-genic with wide unconfined valleys at both high and lowelevations Streams range from steep debris-flow-dominatedheadwater channels to lower-gradient wide laterally uncon-fined channels in its lower reaches The lower reaches of theMF Snoqualmie have been clear-cut extensively since theearly 1900s although there is little logging activity todayVegetation follows an elevation gradient The talus activeglaciers and alpine tundra at the highest elevations transitionto subalpine forests dominated by mountain hemlock (Tsugamertensiana) (above approximately 1500 m) but also includ-ing Pacific silver fir (Abies amabilis) and noble fir (Abies pro-cera) in the lower subalpine and montane zones (above ap-proximately 900 m) Below the montane zone uplands andterraces are covered by Douglas fir (Pseudotsuga menziesii)and western hemlock (Tsuga heterophylla) whereas activeriparian zones are dominated by red alder (Alnus rubra) andbigleaf maple (Acer macrophyllum)

The semiarid Big Sandy is considerably drier than the MFSnoqualmie but also exhibits broad glacially carved valleysespecially in headwater reaches It has a mean annual precip-itation of 072 m (Oregon State University 2004) 1630 m oftotal relief over a 114 km2 drainage area and a mean basinslope of 25 The lower reaches of the Big Sandy are an-thropogenically impacted by moderate grazing use and an ac-cess road that crosses through part of the basin Herbaceousalpine tundra dominates higher elevations (above approxi-mately 3100 m) while the subalpine zone (approximately2900 to 3100 m) is characterized by forests of whitebarkpine (Pinus albicaulis) Engelmann spruce (Picea engelman-nii) and subalpine fir (Abies lasiocarpa) The montane zone(approximately 2600 to 2900 m) is comprised dominantlyof lodgepole pine (Pinus contorta) Only a small portion ofthis basin (approximately 1 ) resides below 2500 m whereshrub steppe begins to dominate (Fall 1994) Parklands andmeadows are abundant in this basin creating a patchy for-est structure Comparing this basin to the MF Snoqualmieprovides bioclimatic contrast that allows us to examine howfloodplain soil OC concentrations vary across a range ofstream morphologies and floodplain morphologic types in re-gions with differing precipitation forest characteristics andbasin morphology

To simplify comparison of these two basins we hence-forth refer to them by their dominant climate The MiddleFork Snoqualmie is the wet basin and the Big Sandy is thesemiarid basin (Fig 1)

22 Study design and sampling

We sampled the semiarid basin in summer 2016 and the MFSnoqualmie in summer 2017 During each sampling cam-paign no large floods occurred and we observed no flood-plain erosion or deposition Across both basins we cored atotal of 128 floodplain sites to determine soil OC concentra-tion Cores were collected as a series of individual soil sam-ples at both regular and irregular depth increments

221 Semiarid basin (Big Sandy)

The sparse vegetation in the semiarid basin enabled us touse a combination of a 10 m DEM and satellite imagery tomanually map the extent of the valley bottom along the en-tire stream network and delineate valley bottoms based onconfinement We defined unconfined valley bottoms as thosein which channel width occupied no more than half the val-ley bottom and confined valley bottoms as those in whichchannel width occupied greater than half the valley bottomWithin each confinement stratum we stratified the streamnetwork by five drainage area classes to produce a total of 10strata ensuring even sampling across the basin Within eachof the resulting 10 strata we randomly selected 5 reachesproducing a total of 50 sample sites throughout the basinDue to access issues we sampled 48 out of the 50 randomlylocated sites We supplemented these with 4 subjectively lo-cated sites that we felt enhanced our ability to capture vari-ation throughout the drainage based on observations in thefield resulting in a total of 52 sampled sites

222 Wet basin (Middle Fork Snoqualmie)

The wet basin is larger than the semiarid basin and has ex-tensive low-gradient floodplains in its downstream reachesThese extensive floodplains display high spatial variability invegetation surface water grain size and estimated surfaceage based on aerial imagery and ground reconnaissance Toensure an unbiased characterization of these heterogeneousfloodplains we used aerial imagery a 10 m DEM and pic-tures from field reconnaissance to delineate the floodplaininto patch categories fill channels (abandoned channels thathave had enough sediment deposited to prevent an oxbowlake from forming) point bars (actively accreting surfaceson the inside of bends) wetlands (areas with standing wa-ter in imagery that are not obviously oxbows) oxbow lakes(abandoned channels dammed at the upstream and down-stream ends to form a lake) and general floodplain surfaces(surfaces that cannot be classified into any of the above cat-

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1104 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

Figure 1 Hillshade showing the location sampling sites and stream network of the sampled basins Big Sandy Wyoming (4269minus10925) is on the left and MF Snoqualmie Washington (4753 minus12152) is on the right Circles represent sampling locations atwhich floodplain soil OC was measured Sample sites are colored by OC concentration Mean annual precipitation (MAP) drainage area(DA) and relief are given for each basin

egories) Within each of these five categories we randomlyselected six points at which to take soil cores

We also stratified the entire wet basin stream network bychannel slope into four strata Within each channel slopestrata (hereafter referred to as slope strata) we randomly se-lected 10 reaches to collect a single floodplain soil core re-sulting in 40 randomly located sample sites

To supplement randomly sampled sites and accommodatethe infeasibility of accessing two of the randomly sampledsites along the stream network we also subjectively selectedsample sites in places that we felt enhanced the degree towhich our sampling captured the variability present amongstreams in the basin This resulted in a total of 46 sites strat-ified by slope 38 of which were randomly sampled in addi-tion to 30 sites stratified by floodplain type

23 Reach-scale field measurements

At each sampled reach (100 m or 10 channel widthswhichever was shorter) we measured channel geometry andother characteristics although our measurements were notconsistent across all basins because field protocol evolvedduring the course of the study In both basins we measuredconfinement valley bottom width and channel bed slope Weadditionally measured bankfull width and depth in the wetbasin We did not measure channel characteristics for sitesstratified by floodplain type in the wet basin since they didnot correspond to a single reach of channel as sites stratifiedby slope did in much more confined valleys

In the wet basin we also categorized channels by planformand dominant bedform (Montgomery and Buffington 1997)We defined planforms as follows straight in which the chan-nel was generally confined and significant lateral migrationwas not evident meandering in which lateral migration wasevident but only a single channel existed anastomosing inwhich vegetated islands separated multiple channels and

anabranching in which a single dominant channel existedwith relict channels separated by vegetated islands We fur-ther classified channels as being either multithread (anasto-mosing or anabranching) or single thread (straight or mean-dering) Because logging records are inconsistent and likelyinaccurate in the wet basin (based on the frequent observa-tion of past logging activity that was not recorded in ForestService records) we noted whether signs of logging suchas cut stumps cable decommissioned roads or railroads orother logging-associated tools were found near the reach

We chose a representative location on the floodplain foreach sampled site based on visual examination of vegeta-tion type soil surface texture surface water presence andelevation relative to the bankfull channel elevation (flood-plain sites stratified by type in the wet basin were sampledas close to the randomly sampled point as possible) Oncea location was chosen we extracted a 32 mm diameter soilcore using an open-sided corer (JMC Large Diameter Sam-pling Tube) Due to our adaptive methodology we sampledsoil OC slightly differently in the semiarid versus the wetbasin In the semiarid basin we cored in irregular incrementsgenerally 25ndash30 cm After analyzing data from the semiaridbasin we realized that sampling in regular increments wouldmake analysis more versatile As a result we switched to ex-tracting soil samples at regular 20 cm increments in the wetbasin Cores were taken to refusal (ie coarse gravel or otherobstructions preventing further soil collection) or a depth ofapproximately 1 m Five cores in the semiarid basin 12 coresin the wet basin sites stratified by slope and 11 cores in thewet basin sites stratified by floodplain type did not reach re-fusal When no sand or finer sediment was present in the val-ley bottom (only occurred in headwater channels of the wetbasin) we recorded negligible OC concentration Once soilsamples were removed from the ground they were placed in

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1105

ziplock bags frozen within 72 h (most samples were frozenwithin 8 h) and kept frozen until analysis

24 Measuring soil OC and texture

To measure the concentration of organic carbon in soil sam-ples we used loss on ignition (LOI) We first defrosted sam-ples for 24ndash48 h at room temperature Once defrosted wethoroughly mixed samples to ensure the most homogenoussample possible We then subsampled 10ndash85 g of soil fromeach sample for analysis Using crucibles in a muffle furnacewe dried samples in batches of 30 for 24 h at 105 C to deter-mine moisture content and remove all nonstructurally heldwater Following the guidelines suggested by Hoogsteen etal (2015) we then burned samples for 3 h at 550 C to re-move organic matter By comparing the weight of the burnedsamples with that of the dried samples we obtained an LOIweight

After performing LOI we used burned samples to performtexture by feel to determine the USDA soil texture class andestimated clay content (Thien 1979) To convert LOI weightto OC concentration we used the structural water loss cor-rection of Hoogsteen et al (2015) using clay content esti-mated from soil texture This correction considers water heldby clay that may not evaporate during drying but will evap-orate during burning It also estimates the proportion of theLOI weight that is OC This correction represented as a per-cent of the estimated sample OC content after the correctionranged from 090 to 49576 (95 confidence interval ofthe median between 1617 and 2425 ) for the wet basinand ranged from 554 to 57053 (95 confidence in-terval of the median between 1976 and 3211 ) for thesemiarid basin

One potential confounding factor in LOI is carbonates thatmay burn off during ignition adding to the LOI weight whilenot being organic matter In lithologies where carbonates arerare (eg granitoid rocks like those found in the upper partof the wet basin and entire semiarid basins) this is a rela-tively negligible issue However some of our soil samplescame from parts of the wet basin draining rocks of the west-ern meacutelange belt including argillite greywacke and marbleWe tested samples for the presence of carbonates to deter-mine whether our LOI methods would be sufficient to accu-rately determine OC concentration We randomly chose 10soil samples of a total of 110 drained rocks that could includecarbonates and submitted them to the Colorado State Univer-sity soil testing laboratory for CHN furnace analysis (Sparks1996) which yielded data on the proportion of carbonatesby mass in those samples The median calcium carbonateconcentrations of those 10 samples was 098 (95 con-fidence interval between 070 and 103 ) and the medianpercentage of the total carbon in samples comprised of inor-ganic carbon was 505 (95 confidence interval between197 and 189 ) From this we concluded that the amountof carbonate in the samples draining potentially carbonate-

bearing rocks was low enough that LOI was likely to still beaccurate Consequently we analyzed all soil samples usingLOI to obtain OC concentration The median difference be-tween the LOI and CHN OC estimates for these 10 sampleswas 045 (95 confidence interval between minus083 and147 ) indicating no systematic bias in LOI estimates ofOC concentration

25 GIS and derivative measurements

After fieldwork in each basin we collected the following datafor each reach using a GIS platform elevation drainage arealand cover classification and canopy cover from the NationalLand Cover Database (Homer et al 2015) and the meanslope of the basin draining to each reach (including hill-slopes and channels) Utilizing drainage area at each reachand field-measured channel gradient we calculated an esti-mated stream power as the product of drainage area chan-nel gradient and basin-averaged precipitation We utilizeda 10 m DEM for all GIS topographic measurements To es-timate clay content for each sample we used median val-ues for assigned USDA texture classes To obtain estimatedclay content moisture and OC for each core we calculatedan average weighted by the percentage of core taken up byeach soil sample We categorized wet basin samples stratifiedby floodplain type into those with standing water (wetlandsand oxbow lakes) and those with no standing water (all othertypes)

26 Statistical analyses

All statistical analyses were performed using the R statisti-cal computing software (R Core Team 2017) We conductedall analyses on three modeling groups based on the variablesmeasured in each group In the wet basin we grouped obser-vations by stratification type separating observations strati-fied by channel slope from observations stratified by flood-plain type We separated these two groups from all obser-vations in the semiarid basin which were measured consis-tently We modeled OC concentration and soil texture with amixed-effects linear regression using individual soil samples(ie the individual samples that make up a core) as sam-ple units (n= 103 for wet basin stratified by slope 89 forwet basin stratified by floodplain type and 101 for semiaridbasin) We modeled the sampled site as a random effect ac-knowledging that individual soil samples within a single coreare likely nonindependent We used profiled 95 confidenceintervals on effect estimates (β) for fixed effects to evaluatevariable importance in mixed-effects models

To gain further insight at the reach scale we also mod-eled average OC concentration and soil moisture at eachmeasured site using multiple linear regression We mod-eled soil moisture at the reach scale because we felt thatour single snapshot of moisture conditions was better rep-resented as a site-level average We first performed uni-

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1106 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

variate analysis between each hypothesized predictor andthe response utilizing mainly comparative Wilcoxon rank-sum tests (Wilcoxon 1945) or correlational Spearman cor-relation coefficient statistics We utilized a Holm multiple-comparison correction (Holm 1979) for pairwise compar-isons During this filtering we also viewed box plots or scat-terplots as appropriate to discern which variables appear tohave anything other than a completely random relationshipwith the response We then utilized the multiple linear re-gression of all subsets using the corrected Akaike informa-tion criterion as a model selection criterion (Wagenmakersand Farrell 2004) We iteratively transformed response vari-ables to ensure the homoscedasticity of error terms To selecta single best model we utilized both Akaike-weight-basedimportance and parsimony to select a final reduced modelWe considered sample sizes p values and effect magnitudesin determining variable importance

We also analyzed each core to determine whether therewere buried high-OC-concentration layers at depth We com-pared each buried soil sample to the sample above it usingthe criterion that a peak in OC at depth should have an OCconcentration 15 times that of the overlying sample and beabove 05 (Appling et al 2014)

3 Results

Model results are presented in Table 1 Comparisons betweenbasins and summaries of OC concentration moisture andestimated clay content are shown in Fig 2

31 OC concentration

Most cores display a decrease in OC concentration withdepth (Fig 3) Of cores with more than a single sample32 (722) of cores stratified by slope in the wet basin 32 (825) of cores stratified by floodplain type in the wet basinand 6 (231) of cores in the semiarid basin exhibit OC con-centration peaks at depth Whether a soil sample was classi-fied as an OC peak has no relation to estimated clay contentin sites stratified by floodplain type (p = 028) or those strat-ified by slope (p = 089) in the wet basin In the semiaridbasin soil samples classified as buried OC peaks have signif-icantly higher estimated clay contents (p = 005) than thosethat were not classified as peaks

In general the floodplain-stratified sites in the wet basinstore higher densities of OC than the semiarid basin (Fig 2ab) Figure 2a includes zero values (ie sites with no OC-bearing sediment only present in the wet basin slope-stratified group) whereas Fig 2b does not because sampleunits in Fig 2b are individual soil samples Comparing thesetwo groups it appears that soils in the wet basin exhibit muchhigher OC concentrations than those in the semiarid basinbut in general there are many more reaches with no fine sed-iment available to store OC in the wet basin

At the scale of individual soil samples the depth belowground surface is by far the dominant control on OC con-centration across all modeling groups We used a cube roottransform for all three mixed-effects models of OC concen-tration For wet basin sites stratified by slope deeper soilsamples contain less OC (β =minus00084plusmn 00042) whereassoil samples at higher elevations tend to contain more OC(β = 00010plusmn000099) Depth is the only significant predic-tor of OC content for both wet basin soil samples stratifiedby floodplain type (β =minus00084plusmn00042) and soil samplesin the semiarid basin (β =minus00037plusmn 00019)

Modeling wet basin slope-stratified sites at the reach scalewe found that moisture (β = 00078plusmn 00031) and whetherthe reach was unconfined (β = 077plusmn 049) control soil OC(cube root transformation model adjusted R2

= 054 p lt00001) Modeling wet basin floodplain-stratified sites at thesite scale we found that canopy cover (β = 0012plusmn 0011)and moisture (β = 00040plusmn 00011) are controls on soilOC (cube root transformation model adjusted R2

= 067p lt 00001) Modeling Big Sandy sites at the reach scalewe found that soil depth (β =minus0012plusmn 00071) and mois-ture (β = 0014plusmn 00026) are dominant controls on soil OCconcentration (no transformation model adjustedR2

= 069p lt 00001)

In general moister deeper soils store more OC at the reachscale whereas OC tends to vary dominantly with depth atthe scale of individual soil samples Although estimated claycontent did not emerge as a significant predictor of OC con-centration it is used to calculate clay-held water to correctour LOI-based OC concentration measurements making itimportant in determining OC for each sample

32 Soil texture

In general soil texture follows a predictable trend withriver size between model groups (Fig 2d) Floodplain type-stratified sites in the wet basin store the most clay followedby slope-stratified sites and then sites in the semiarid basin

Modeling soil texture at the individual soil sample scaleacross slope-stratified sites in the wet basin we foundwhether the reach was confined (β = 542plusmn 532) andwhether the bed material was dominantly sand (β = 1047plusmn613) to be dominant controls on estimated clay contentModeling soil texture for sites stratified by floodplain typein the wet basin yielded no significant trends In the semi-arid basin we found that either valley width (β = 00050plusmn00032) or whether the stream was unconfined (β = 041plusmn033) as well as depth below ground surface (β =minus00069plusmn00033 for model with valley width but not confinement) sig-nificantly control soil texture

To summarize sites from unconfined lower-energyreaches in the wet basin and sites from reaches with widervalley bottoms and at lower depths in the semiarid basin ex-hibited more finely textured soils

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1107

Tabl

e1

Mat

rix

ofal

lmod

els

pres

ente

din

the

text

Eac

hm

odel

islis

ted

bym

odel

grou

pre

spon

seva

riab

lea

ndsc

ale

Scal

ere

fers

toth

esa

mpl

eun

itof

the

mod

ela

ndsi

tere

fers

toa

core

with

the

resp

onse

aver

aged

over

allt

hein

divi

dual

soil

sam

ples

inth

eco

reF

orea

chva

riab

lean

dm

odel

an

aste

risk

()i

ndic

ates

that

the

vari

able

was

incl

uded

inm

odel

sele

ctio

nbu

tnot

the

full

mix

ed-e

ffec

tsm

odel

Am

inus

(minus)

indi

cate

sth

atth

eva

riab

lew

asse

lect

edas

impo

rtan

tin

pred

ictin

gth

ere

spon

sean

dde

note

san

indi

rect

corr

elat

ion

whe

reas

apl

us(+

)ind

icat

esa

dire

ctco

rrel

atio

nIn

the

case

ofco

nfine

men

ta

plus

indi

cate

sth

atun

confi

ned

stre

ams

disp

lay

ahi

gher

-mag

nitu

dere

spon

seva

riab

leI

nth

eca

seof

bed

mat

eria

la

plus

indi

cate

sth

atsa

mpl

esw

ithsa

ndex

hibi

tahi

gher

valu

eof

the

resp

onse

NA

indi

cate

sth

atei

ther

the

vari

able

was

notm

easu

red

for

that

basi

nor

mod

elgr

oup

orth

atit

isth

em

odel

resp

onse

Mod

elgr

oup

Res

pons

eSc

ale

(sam

ple

unit)

Var

iabl

esConfinement

Bedform

Channelslope

Bedmaterial

Multithread

Valleywidth

Bankfullwidth

Bankfulldepth

Streampower

Floodplaintype

Standingwater

Depth1

Claycontent

Moisture

Loggingnearby

Grassespresent

Shrubspresent

Treespresent

Elevation

Basinslope

Canopycover

NLCD

Drainagearea

Wet

basi

nst

rat-

ified

bysl

ope

OC

()

Soil

sam

ple

NA

NAminus

+

O

C(

)Si

te+

NA

NA

+

Moi

stur

e(

)Si

te+

minus

NA

NA

N

A+

Te

xtur

e(

)So

ilsa

mpl

e+

+

N

AN

A

NA

Wet

basi

nst

rat-

ified

byflo

od-

plai

nty

pe

OC

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

minus

N

AO

C(

)Si

teN

AN

AN

AN

AN

AN

AN

AN

AN

A

+

+N

AM

oist

ure

()

Site

NA

NA

NA

NA

NA

NA

NA

NA

NA

+

+

NA

NA

Text

ure2

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

N

AN

A

Sem

iari

dba

sin

OC

()

Soil

sam

ple

N

A

NA

NA

N

AN

A

NA

NAminus

NA

NA

NA

NA

OC

()

Site

N

A

NA

NA

NA

NA

N

AN

Aminus

+N

AN

AN

AN

A

M

oist

ure

()

Site

+N

A

NA

NA

NA

NA

N

AN

A+

NA

NA

NA

NA

NA+

Te

xtur

e3(

)So

ilsa

mpl

e+

NA

N

AN

A+

NA

NA

N

AN

Aminus

NA

NA

NA

NA

NA

1D

epth

refe

rsto

eith

erth

eso

ilsa

mpl

ede

pth

belo

wth

egr

ound

orth

eto

tald

epth

ofth

eco

red

epen

ding

onth

esa

mpl

eun

it2

No

sign

ifica

ntre

sults

wer

eob

serv

edfo

rthi

sm

odel

3Fo

rthi

sm

odel

bot

hva

lley

wid

than

dco

nfine

men

tpre

dict

text

ure

and

can

bein

terp

rete

din

terc

hang

eabl

yH

owev

eri

nclu

ding

both

inth

esa

me

mod

elw

ould

yiel

dpr

oble

ms

due

tom

ultic

ollin

eari

ty

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1108 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

Figure 2 Box plots showing comparisons between model groups of OC concentration at the reach scale (a) OC concentration at the scaleof individual soil samples (b) moisture at the reach scale (c) and estimated clay content at the scale of individual soil samples (d) Endsof dotted lines represent 15 times the interquartile range which is represented by boxes Bold line represents median Circles representoutliers Letters indicate probable differences between groups based on pairwise Wilcoxon (andashc) or t tests (d) with a Holm correctionRanges in parentheses below letters show the 95 confidence interval of the median value for the group (andashc) or the mean value for thegroup (d) in which median confidence intervals were overly constrained due to the categorical nature of our estimated clay content data

33 Soil moisture

Soil moisture is less variable between basins than either tex-ture or OC concentration (Fig 2c) All model groups exhibitsimilar soil moisture conditions although there was signifi-cant variability within each model group

Soil moisture at wet basin sites stratified by slope is domi-nantly controlled by channel slope (β =minus1315plusmn911) ele-vation (β = 00046plusmn 00038) and whether the stream is un-confined (β = 389plusmn 267 model adjusted R2

= 038 p lt00001) At wet basin sites stratified by floodplain type es-timated clay content (β = 0060plusmn 0042) and whether thefloodplain unit had standing water (β = 115plusmn 071) sig-nificantly control soil moisture (model adjusted R2

= 046p lt 00001) In the semiarid basin soil depth (β = 0012plusmn0012) elevation (β = 00023plusmn 00014) and whether thereach was unconfined (β = 091plusmn075) significantly controlsoil moisture (model adjusted R2

= 035 p lt 00001)

4 Discussion

41 Understanding spatial variability in OCconcentration in floodplain soils (H1 and H2)

Comparing the wet basin to the semiarid basin shows thatthe wetter higher primary productivity basin is capable ofstoring greater concentrations of OC in floodplain soils butthat both regions generally store similar OC concentrations infloodplain soils This result partially agrees with the exami-nation of subalpine lake deltas by Scott and Wohl (2017) Inthat study subalpine lake deltas in the wet basin were com-pared to deltas in the drier Colorado Front Range Subalpinelake deltas displayed similar OC concentrations likely dueto competing but complementary OC stabilization and lossmechanisms in each region Those deltas represent a subsetof the broader valley bottom soils studied here This moreexpansive study points to both geomorphic controls such asvalley bottom geometry and factors influenced by climatesuch as canopy cover as controls on OC storage in valley bot-toms These results also agree with the results of Lininger et

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1109

Figure 3 Box plots of sample OC concentration binned by meansample depth for the wet (a) and semiarid (b) basins Ends of dot-ted lines represent 15 times the interquartile range which is rep-resented by boxes Bold line represents median Black circles rep-resent outliers Transparent grey points show all data for each binSample size for each bin is denoted by n

al (2018) which indicate that geomorphic context and vege-tation dynamics control OC concentration on floodplain soilsalong large lowland rivers in Alaska USA

At the reach or site scale wetter soil profiles consistentlyyield higher OC concentrations in all model groups How-ever moisture does not differ significantly between modelgroups (Fig 2c) indicating that this alone cannot explaindifferences between basins Soils tend to be finer in the wetbasin but clay content was not an important predictor of OCconcentration in studied soils Clay content likely influencesOC concentration based on previous research (Hoffmann etal 2009) although the inclusion of coarse soil material (in-cluding particulate organic matter) in our samples may ex-plain the lack of an observed correlation here Variabilityin soil redox conditions (ie the presence of anaerobic mi-crosites) may also introduce variability in respiration rate be-tween basins and among individual soil samples (Keiluweit

et al 2017) That is variability in groundwater hydrologymay result in a heterogeneous mix of aerobic (favoring mi-crobial activity) and anaerobic (suppressing microbial activ-ity) conditions within floodplain soils We did not record re-doximorphic features of our soil samples so we are unableto determine whether this influence contributes to the errorin our models Although confinement plays a strong role indetermining OC concentration in wet basin sites stratified byslope it does not differ significantly between basins (52 of semiarid basin reaches are unconfined compared to 63 of wet basin reaches) The major differences between thesebasins are their hydroclimatic and disturbance regimes Thewet basin is at a lower elevation and has denser and higherbiomass forests (Smithwick et al 2002) compared to thesparser parkland forests of the semiarid basin which likelyalso experiences more frequent fires based on fire historiesof nearby regions (recurrence interval on the order of 101ndash102 years Houston 1973 Loope and Gruell 1973) In addi-tion the volcanic soils present in the wet basin may suppresssoil OC respiration (Matus et al 2014) leading to a highersoil OC storage capacity there

Between basins it is likely that the hydroclimatic regimeinfluencing primary production plays some role in the wetbasinrsquos higher maximum OC concentrations in floodplainsoils compared to those of the semiarid basin Howeversmaller-scale factors such as soil texture and moisture alsolikely play a role and are not related to drainage area (Ta-ble 1) indicating that neither OC concentration nor its con-trolling factors vary continuously along a river network thussupporting H2a and H2b This also indicates that local fac-tors set largely by geomorphic and hydrologic dynamicsplay a significant role in modulating the effect of climate onOC concentrations If the wet and semiarid basins displayedsignificantly different OC concentrations our first hypothesisregarding the interbasin controls on OC concentration wouldbe supported However we instead found that climate andprimary productivity only partially determine OC concentra-tions especially when viewed in the context of geomorphicand hydrologic variability Thus the results do not supportH1

Each basin (or model group) is slightly different in termsof the controls on soil OC concentration moisture andtexture In the wet basin sites stratified by slope higher-elevation sites display higher OC concentrations This is con-trary to the general trend in primary productivity which de-creases with increasing elevation However it is importantto note that the headwaters of the wet basin are dominatedby lakes deltas and other depositional features in relativelybroad glacially carved valleys Subalpine lake deltas havebeen shown to store high OC concentrations in this basin(Scott and Wohl 2017) and many of the highest OC concen-trations we measured were located in broad wet meadowssubalpine lake deltas or other unconfined high-elevationreaches Such unconfined sites also likely have significantlycooler temperatures and tend to have higher soil moisture

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1110 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

contents as shown by our modeling (Table 1) As such al-though high-elevation wet basin sites may receive less OCinput they likely have a low rate of OC respiration resultingin higher OC concentrations on the whole which agrees withthe result of Bao et al (2017) In the semiarid basin our mod-eling suggests that the lower temperatures and higher mois-ture (Table 1) at higher elevations do not compensate for thelower primary productivity as elevation does not correlatewith OC concentration

In both basins unconfined reaches contain wetter andfiner-textured soils which may result in a higher soil OCcapacity Although confinement only relates directly to OCcontent in wet basin sites it does play a strong role in deter-mining moisture which in turn plays a role in regulating OCconcentration in both basins likely via inhibiting microbialactivity (Howard and Howard 1993) The relevance of chan-nel slope in determining soil moisture in the wet basin but notsemiarid basin may reflect the prevalence of high-gradientdebris-flow-dominated channels in the wet basin that largelyexhibit only gravel to boulder substrate which we assumestores minimal fine sediment moisture or OC

In the semiarid basin higher soil depths correlate withmoister and finer-textured soils but less OC concentrationThis indicates the trend in OC with depth likely dominatesthe signal of OC concentration with deeper sites containinga higher proportion of OC-depleted deep samples

42 Inferring sources of OC to floodplain soils (H3)

OC can be input to floodplain soils by two primary mech-anisms First dissolved and particulate OC (including largewood) can be deposited on floodplain surfaces by overbankdeposition thus integrating fluvial sedimentary OC into thefloodplain soil profile or in the case of large wood deposit-ing discrete but potentially large concentrations of OC thatcan later be integrated into the soil profile Second litter anddecomposing vegetation on the floodplain surface can act asautochthonous inputs of OC to floodplain soil

Our modeling of OC concentration yielded results consis-tent with previous investigations of controls on soil OC stor-age capacity (Jobbaacutegy and Jackson 2000 Sutfin and Wohl2017) Sites in the heterogeneous floodplain of the wet basindisplay a direct correlation between canopy cover and OCconcentration indicating that increased litter inputs lead toincreased floodplain soil OC concentration Sediment inputslikely differ between floodplain depositional unit types (egcoarser sediment may deposit on point bars compared tofilled secondary channels) although floodplain type does notpredict OC concentration This indicates that vegetation in-puts may be more dominant than fluvial sediment inputs atthese sites

The finding that buried OC peaks in the wet basin do nothave abnormally high clay contents supports the interpre-tation that wood and litter inputs to soil are the dominantsource of OC in the floodplain soils we examined Buried

peaks can be layers created by overbank deposition and sub-sequent burial of fine OC-bearing sediments (Blazejewski etal 2009 Ricker et al 2013) buried pieces of wood (Wohl2013) or buried organic horizons that are now capped bysediments that prevent OC respiration If overbank deposi-tion of fine sediment caused OC peaks we would expect tosee the soil samples classified as peaks exhibiting high claycontents indicating finer sediment Instead our results sug-gest that in the wet basin buried peaks are likely the result ofeither buried organic horizons or buried wood We observedlarge pieces of decaying buried wood in floodplain cut banksin the wet basin supporting this inference Overbank depo-sition of wood on the floodplain was only observed rarely inthis basin indicating that the OC measured in these soils islikely dominantly autochthonous

In the semiarid basin the two cores that exhibit peaks werecollected from the same meadow just downstream of a now-filled former lake that is a potential source of fine sedimentThe channels draining this meadow exhibit an anabranchingplanform indicating the potential to deposit and bury packetsof potentially OC-rich fine sediments However the majorityof cores do not exhibit OC peaks indicating OC input mainlyfrom vegetation at the surface and continuing OC respirationat depth

OC variation within each core is dominantly a function ofdepth We observe a negative correlation between depth be-low ground surface and OC concentration which has beenobserved in other studies including mountain wetlands andfloodplains (Jobbaacutegy and Jackson 2000 Scott and Wohl2017 Sutfin and Wohl 2017 Zhao et al 2017) In generalthis indicates that at least in mountain river floodplains OCis enriched at the surface and decomposes with depth sim-ilar to upland soils This fits with our finding that the ma-jority of our cores do not exhibit significant OC peaks atdepth and supports the dominance of litter and wood OC in-puts to floodplain soils These results support our hypothesisthat decaying litter and wood not overbank sediment depo-sition dominate the input of OC to floodplain soils in ourstudy basins (H3) We note that floodplain wood may alsoact as a trapping site for overbank fine organic matter facil-itating the deposition and input of OC from decaying vege-tation While these study basins likely accumulate soil OCmainly autochthonously other basins that experience over-bank flows accompanying deposition of fine sediment andburial of organic layers exhibit OC storage that is likely dom-inated by fluvial sediment deposition (eg Blazejewski etal 2009 DrsquoElia et al 2017 Ricker et al 2013) Thus itis likely that flow regime lateral connectivity and sedimenttransport dynamics regulate whether floodplain soil OC isdominantly input by the overbank deposition of fine mate-rial or litter and wood decay

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1111

Figure 4 Conceptual model of physical processes that influenceOC concentration in floodplain soils based on this study and otherliterature Each box corresponds to a major factor that influencesOC concentration Colored text within each box denotes factors thatinfluence OC inputs sorption capacity OC export or respirationrate (note that some processes influence both OC export and respi-ration rate) As sorption capacity increases so does the OC capacityof the soil Conversely as the rate of respiration andor OC exportincreases the soil OC capacity decreases Floodplain soils can onlydevelop high concentrations of OC if there are high rates of OCinput However the capacity of the soils to store OC regulates thatinput and is determined by the competing influences of sorption ca-pacity and the combination of respiration rate and OC export Seethe text for further details

43 Conceptual model of soil OC concentration infloodplain soils

We present a conceptual model to summarize our results andplace them in the context of work examining the controlson OC storage in soils (Fig 4) OC is input to floodplainseither through the decay of vegetation or the deposition offine OC-rich sediment This input of OC only determinesOC concentrations insofar as floodplain soils are capable ofstoring OC That storage is effectively determined by a bal-ance between processes that remove OC from floodplainsnamely respiration or erosion followed by respiration (Berheet al 2007) and processes that regulate OC availability tomicrobes namely the capability of the mineral fraction ofthe soil to sorb OC

OC sorption capacity reflects a few specific processes Al-though soil texture generally relates to the ability of OCto sorb to mineral grains and resulting OC availability soilchemistry also plays a strong and potentially dominant rolein regulating OC sorption capacity (Rasmussen et al 2018)Soil texture is largely determined by valley morphology ac-cording to our modeling (Table 1) placing valley morphol-ogy and resulting sediment transport dynamics (Gran andCzuba 2017 Wohl et al 2017b) as indirect controls on sorp-tion capacity

Respiration rate is largely determined by microbial activityand the availability of OC to microbes Erosion can rapidlyexpose soil OC to microbial respiration (Berhe et al 2007)whereas soils that reside in largely anoxic conditions can ex-hibit low rates of microbial respiration (Boye et al 2017)In addition soil mineralogy chemistry and redox condi-tions (Keiluweit et al 2017) can regulate microbial activity(eg andic soils may limit microbial respiration Matus et al2014) Our results suggesting that moisture controls OC con-tent support the idea that drier soils likely have higher ratesof microbial respiration of OC Moisture is a function of tex-ture valley bottom morphology and elevation (a proxy fortemperature) in our modeling (Table 1) Comparing flood-plain types in the wet basin we find that types with standingwater exhibit significantly higher soil moisture contents thanthose without standing water This indicates spatial variabil-ity in moisture content and likely microbial activity (Howardand Howard 1993) In our modeling this effect translates tospatial variability in OC concentration within floodplains andacross entire basins

OC export from soils refers specifically to leaching of dis-solved OC mainly from shallow soils While we do not di-rectly consider OC export in this study it likely introducesvariability in soil OC concentration and is regulated dom-inantly by hydrology (Aringgren et al 2014 Mcdowell andLikens 1988) soil pore water chemistry (including pH andionic strength Brooks et al 1999 Evans et al 2006) andredox conditions (Knorr 2013) OC export as DOC can actin conjunction with respiration to remove OC from soils andlike respiration rate is countered by sorption capacity

To summarize OC inputs regulated by the capacity ofsoils to store OC and suppress microbial respiration andOC export determine OC concentrations in floodplain soils(Fig 4) OC inputs to floodplain soils come from either au-tochthonous litter accumulation on the floodplain surface al-lochthonous wood deposition or allochthonous deposition offine OC-bearing sediments Our results from the mountain-ous basins we studied suggest that deposition of fine mate-rial in overbank flows is rare leading us to infer that au-tochthonous litter and allochthonous wood inputs to flood-plains dominate OC input in mountain rivers Where soils aremore moist microbial respiration is inhibited and more OCis stored Although soil texture is likely not a limiting factoron OC concentration in these floodplains finer-textured soilslikely have a higher sorption capacity retaining more of theOC input from decaying plant material Our results indicatethat geomorphic and hydrologic characteristics act as bound-ary conditions that regulate soil texture and moisture in turnregulating sorption capacity respiration rate and resultingOC concentrations in floodplain soils

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1112 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

5 Conclusions

We present floodplain soil OC concentration data from twodisparate watersheds to compare how interbasin variabilitybetween the two watersheds compares with intra-basin vari-ability in geomorphic and hydrologic characteristics in de-termining OC concentration Our results indicate that OCconcentration in mountain floodplain soils does not vary pre-dictably along a longitudinal gradient nor does it vary sub-stantially between basins with differing climatic and veg-etation characteristics Instead geomorphic and hydrologiccharacteristics such as valley bottom morphology and soilmoisture dominantly determine floodplain OC concentra-tion

In our study basins decaying litter and wood and not over-bank deposition of fine OC-bearing sediment is the mainsource of OC to floodplain soils It is unclear whether thatdecaying vegetation is dominated by autochthonous litter in-puts or transported downed wood In comparing our basinto other studied floodplain soils it seems that vegetation dy-namics play a strong role in determining OC concentrationswhen fine sediment is not regularly deposited on floodplainsurfaces However we suggest that floodplain soil character-istics set by geomorphic and hydrologic conditions regu-late how OC inputs translate to the spatial distribution of OCalong a river network

This implies that OC storage in floodplains likely cannotbe predicted using consistent downstream trends and thatmanagement prioritization designed to facilitate floodplainOC storage should be based on local geomorphic and hy-drologic process variability within each basin For instancemanagement to increase OC sequestration in floodplain soilswill likely be more effective where floodplains are uncon-fined and soils already experience high moisture conditionsfor much of the year Along these lines our results showthat modeling the floodplain biospheric OC pool to predictits response to warming and subsequent effects on climatebased on regional factors such as climate and net primaryproductivity likely misses the substantial interbasin variabil-ity in OC concentration and storage resulting from variabilityin valley bottom geometry and both geomorphic and hydro-logic processes (eg Doetterl et al 2015)

Although our results provide some insights the questionof whether OC stored in floodplain soil comes dominantlyfrom allochthonous versus autochthonous sources remainsopen Our results imply that more productive spatially het-erogeneous floodplains likely input more OC to soils Flood-plain OC concentration while mediated largely by moisturedynamics likely depends mainly on OC inputs from produc-tive riparian forests This implies that management of OCstorage in mountain river floodplains should focus on therestoration of riparian zones to maintain OC input to soil(eg Bullinger-Weber et al 2014) More detailed studiesin regions with varying sediment transport and hydrologicregimes are needed to determine what conditions favor au-

tochthonous versus allochthonous OC inputs but our resultssuggest that autochthonous sources dominate floodplain OCstorage in basins with relatively low rates of vertical accre-tion and high channelndashfloodplain connectivity that promotesfloodplain wetlands

Data availability Data to support the analyses presented here canbe found in Table S1 and in the Colorado State University DigitalRepository (Scott and Wohl 2018a)

Supplement The supplement related to this article is availableonline at httpsdoiorg105194esurf-6-1101-2018-supplement

Author contributions Field data collection was conducted pri-marily by DNS with assistance from EEW at some sites in theMiddle Fork Snoqualmie basin DNS performed data analysis anddrafted the initial version of the paper which was then edited andrewritten by DNS and EEW

Competing interests The authors declare that they have no con-flict of interest

Acknowledgements This work was funded by NSF grantEAR-1562713 We thank Ellen Daugherty for extensive assistancein fieldwork and discussion that improved the conceptual modelWe thank Katherine Lininger for stimulating discussion thatimproved the paper Detailed and constructive comments from twoanonymous reviewers and Robert Hilton improved the paper Wethank Sara Lowe for assistance in processing and analyzing soilsamples

Edited by Robert HiltonReviewed by two anonymous referees

References

Aringgren A M Buffam I Cooper D M Tiwari T Evans C Dand Laudon H Can the heterogeneity in stream dissolved or-ganic carbon be explained by contributing landscape elementsBiogeosciences 11 1199ndash1213 httpsdoiorg105194bg-11-1199-2014 2014

Appling A P Bernhardt E S and Stanford J A Flood-plain biogeochemical mosaics A multi-dimensional view ofalluvial soils J Geophys Res-Biogeo 119 1538ndash1553httpsdoiorg1010022013JG002543 2014

Aufdenkampe A K Mayorga E Raymond P A MelackJ M Doney S C Alin S R Aalto R E and YooK Riverine coupling of biogeochemical cycles between landoceans and atmosphere Front Ecol Environ 9 53ndash60httpsdoiorg101890100014 2011

Ballantyne A P Alden C B Miller J B Tans P P and WhiteJ W C Increase in observed net carbon dioxide uptake by

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1103

21 Field sites

We quantified soil organic carbon concentrations to a depthof approximately 1 m in the Big Sandy basin in the WindRiver Range of Wyoming and the Middle Fork Snoqualmiebasin in the central Cascade Mountains of Washington(Fig 1) These basins represent distinct bioclimatic and ge-omorphologic regions ranging from the wet high-relief Cas-cades to the semiarid moderate-relief Middle Rockies

The MF Snoqualmie has a mean annual precipitation of304 m (Oregon State University 2004) 2079 m of total re-lief over a 407 km2 drainage area and a mean basin slope of60 Topography in the MF Snoqualmie is largely glacio-genic with wide unconfined valleys at both high and lowelevations Streams range from steep debris-flow-dominatedheadwater channels to lower-gradient wide laterally uncon-fined channels in its lower reaches The lower reaches of theMF Snoqualmie have been clear-cut extensively since theearly 1900s although there is little logging activity todayVegetation follows an elevation gradient The talus activeglaciers and alpine tundra at the highest elevations transitionto subalpine forests dominated by mountain hemlock (Tsugamertensiana) (above approximately 1500 m) but also includ-ing Pacific silver fir (Abies amabilis) and noble fir (Abies pro-cera) in the lower subalpine and montane zones (above ap-proximately 900 m) Below the montane zone uplands andterraces are covered by Douglas fir (Pseudotsuga menziesii)and western hemlock (Tsuga heterophylla) whereas activeriparian zones are dominated by red alder (Alnus rubra) andbigleaf maple (Acer macrophyllum)

The semiarid Big Sandy is considerably drier than the MFSnoqualmie but also exhibits broad glacially carved valleysespecially in headwater reaches It has a mean annual precip-itation of 072 m (Oregon State University 2004) 1630 m oftotal relief over a 114 km2 drainage area and a mean basinslope of 25 The lower reaches of the Big Sandy are an-thropogenically impacted by moderate grazing use and an ac-cess road that crosses through part of the basin Herbaceousalpine tundra dominates higher elevations (above approxi-mately 3100 m) while the subalpine zone (approximately2900 to 3100 m) is characterized by forests of whitebarkpine (Pinus albicaulis) Engelmann spruce (Picea engelman-nii) and subalpine fir (Abies lasiocarpa) The montane zone(approximately 2600 to 2900 m) is comprised dominantlyof lodgepole pine (Pinus contorta) Only a small portion ofthis basin (approximately 1 ) resides below 2500 m whereshrub steppe begins to dominate (Fall 1994) Parklands andmeadows are abundant in this basin creating a patchy for-est structure Comparing this basin to the MF Snoqualmieprovides bioclimatic contrast that allows us to examine howfloodplain soil OC concentrations vary across a range ofstream morphologies and floodplain morphologic types in re-gions with differing precipitation forest characteristics andbasin morphology

To simplify comparison of these two basins we hence-forth refer to them by their dominant climate The MiddleFork Snoqualmie is the wet basin and the Big Sandy is thesemiarid basin (Fig 1)

22 Study design and sampling

We sampled the semiarid basin in summer 2016 and the MFSnoqualmie in summer 2017 During each sampling cam-paign no large floods occurred and we observed no flood-plain erosion or deposition Across both basins we cored atotal of 128 floodplain sites to determine soil OC concentra-tion Cores were collected as a series of individual soil sam-ples at both regular and irregular depth increments

221 Semiarid basin (Big Sandy)

The sparse vegetation in the semiarid basin enabled us touse a combination of a 10 m DEM and satellite imagery tomanually map the extent of the valley bottom along the en-tire stream network and delineate valley bottoms based onconfinement We defined unconfined valley bottoms as thosein which channel width occupied no more than half the val-ley bottom and confined valley bottoms as those in whichchannel width occupied greater than half the valley bottomWithin each confinement stratum we stratified the streamnetwork by five drainage area classes to produce a total of 10strata ensuring even sampling across the basin Within eachof the resulting 10 strata we randomly selected 5 reachesproducing a total of 50 sample sites throughout the basinDue to access issues we sampled 48 out of the 50 randomlylocated sites We supplemented these with 4 subjectively lo-cated sites that we felt enhanced our ability to capture vari-ation throughout the drainage based on observations in thefield resulting in a total of 52 sampled sites

222 Wet basin (Middle Fork Snoqualmie)

The wet basin is larger than the semiarid basin and has ex-tensive low-gradient floodplains in its downstream reachesThese extensive floodplains display high spatial variability invegetation surface water grain size and estimated surfaceage based on aerial imagery and ground reconnaissance Toensure an unbiased characterization of these heterogeneousfloodplains we used aerial imagery a 10 m DEM and pic-tures from field reconnaissance to delineate the floodplaininto patch categories fill channels (abandoned channels thathave had enough sediment deposited to prevent an oxbowlake from forming) point bars (actively accreting surfaceson the inside of bends) wetlands (areas with standing wa-ter in imagery that are not obviously oxbows) oxbow lakes(abandoned channels dammed at the upstream and down-stream ends to form a lake) and general floodplain surfaces(surfaces that cannot be classified into any of the above cat-

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1104 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

Figure 1 Hillshade showing the location sampling sites and stream network of the sampled basins Big Sandy Wyoming (4269minus10925) is on the left and MF Snoqualmie Washington (4753 minus12152) is on the right Circles represent sampling locations atwhich floodplain soil OC was measured Sample sites are colored by OC concentration Mean annual precipitation (MAP) drainage area(DA) and relief are given for each basin

egories) Within each of these five categories we randomlyselected six points at which to take soil cores

We also stratified the entire wet basin stream network bychannel slope into four strata Within each channel slopestrata (hereafter referred to as slope strata) we randomly se-lected 10 reaches to collect a single floodplain soil core re-sulting in 40 randomly located sample sites

To supplement randomly sampled sites and accommodatethe infeasibility of accessing two of the randomly sampledsites along the stream network we also subjectively selectedsample sites in places that we felt enhanced the degree towhich our sampling captured the variability present amongstreams in the basin This resulted in a total of 46 sites strat-ified by slope 38 of which were randomly sampled in addi-tion to 30 sites stratified by floodplain type

23 Reach-scale field measurements

At each sampled reach (100 m or 10 channel widthswhichever was shorter) we measured channel geometry andother characteristics although our measurements were notconsistent across all basins because field protocol evolvedduring the course of the study In both basins we measuredconfinement valley bottom width and channel bed slope Weadditionally measured bankfull width and depth in the wetbasin We did not measure channel characteristics for sitesstratified by floodplain type in the wet basin since they didnot correspond to a single reach of channel as sites stratifiedby slope did in much more confined valleys

In the wet basin we also categorized channels by planformand dominant bedform (Montgomery and Buffington 1997)We defined planforms as follows straight in which the chan-nel was generally confined and significant lateral migrationwas not evident meandering in which lateral migration wasevident but only a single channel existed anastomosing inwhich vegetated islands separated multiple channels and

anabranching in which a single dominant channel existedwith relict channels separated by vegetated islands We fur-ther classified channels as being either multithread (anasto-mosing or anabranching) or single thread (straight or mean-dering) Because logging records are inconsistent and likelyinaccurate in the wet basin (based on the frequent observa-tion of past logging activity that was not recorded in ForestService records) we noted whether signs of logging suchas cut stumps cable decommissioned roads or railroads orother logging-associated tools were found near the reach

We chose a representative location on the floodplain foreach sampled site based on visual examination of vegeta-tion type soil surface texture surface water presence andelevation relative to the bankfull channel elevation (flood-plain sites stratified by type in the wet basin were sampledas close to the randomly sampled point as possible) Oncea location was chosen we extracted a 32 mm diameter soilcore using an open-sided corer (JMC Large Diameter Sam-pling Tube) Due to our adaptive methodology we sampledsoil OC slightly differently in the semiarid versus the wetbasin In the semiarid basin we cored in irregular incrementsgenerally 25ndash30 cm After analyzing data from the semiaridbasin we realized that sampling in regular increments wouldmake analysis more versatile As a result we switched to ex-tracting soil samples at regular 20 cm increments in the wetbasin Cores were taken to refusal (ie coarse gravel or otherobstructions preventing further soil collection) or a depth ofapproximately 1 m Five cores in the semiarid basin 12 coresin the wet basin sites stratified by slope and 11 cores in thewet basin sites stratified by floodplain type did not reach re-fusal When no sand or finer sediment was present in the val-ley bottom (only occurred in headwater channels of the wetbasin) we recorded negligible OC concentration Once soilsamples were removed from the ground they were placed in

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1105

ziplock bags frozen within 72 h (most samples were frozenwithin 8 h) and kept frozen until analysis

24 Measuring soil OC and texture

To measure the concentration of organic carbon in soil sam-ples we used loss on ignition (LOI) We first defrosted sam-ples for 24ndash48 h at room temperature Once defrosted wethoroughly mixed samples to ensure the most homogenoussample possible We then subsampled 10ndash85 g of soil fromeach sample for analysis Using crucibles in a muffle furnacewe dried samples in batches of 30 for 24 h at 105 C to deter-mine moisture content and remove all nonstructurally heldwater Following the guidelines suggested by Hoogsteen etal (2015) we then burned samples for 3 h at 550 C to re-move organic matter By comparing the weight of the burnedsamples with that of the dried samples we obtained an LOIweight

After performing LOI we used burned samples to performtexture by feel to determine the USDA soil texture class andestimated clay content (Thien 1979) To convert LOI weightto OC concentration we used the structural water loss cor-rection of Hoogsteen et al (2015) using clay content esti-mated from soil texture This correction considers water heldby clay that may not evaporate during drying but will evap-orate during burning It also estimates the proportion of theLOI weight that is OC This correction represented as a per-cent of the estimated sample OC content after the correctionranged from 090 to 49576 (95 confidence interval ofthe median between 1617 and 2425 ) for the wet basinand ranged from 554 to 57053 (95 confidence in-terval of the median between 1976 and 3211 ) for thesemiarid basin

One potential confounding factor in LOI is carbonates thatmay burn off during ignition adding to the LOI weight whilenot being organic matter In lithologies where carbonates arerare (eg granitoid rocks like those found in the upper partof the wet basin and entire semiarid basins) this is a rela-tively negligible issue However some of our soil samplescame from parts of the wet basin draining rocks of the west-ern meacutelange belt including argillite greywacke and marbleWe tested samples for the presence of carbonates to deter-mine whether our LOI methods would be sufficient to accu-rately determine OC concentration We randomly chose 10soil samples of a total of 110 drained rocks that could includecarbonates and submitted them to the Colorado State Univer-sity soil testing laboratory for CHN furnace analysis (Sparks1996) which yielded data on the proportion of carbonatesby mass in those samples The median calcium carbonateconcentrations of those 10 samples was 098 (95 con-fidence interval between 070 and 103 ) and the medianpercentage of the total carbon in samples comprised of inor-ganic carbon was 505 (95 confidence interval between197 and 189 ) From this we concluded that the amountof carbonate in the samples draining potentially carbonate-

bearing rocks was low enough that LOI was likely to still beaccurate Consequently we analyzed all soil samples usingLOI to obtain OC concentration The median difference be-tween the LOI and CHN OC estimates for these 10 sampleswas 045 (95 confidence interval between minus083 and147 ) indicating no systematic bias in LOI estimates ofOC concentration

25 GIS and derivative measurements

After fieldwork in each basin we collected the following datafor each reach using a GIS platform elevation drainage arealand cover classification and canopy cover from the NationalLand Cover Database (Homer et al 2015) and the meanslope of the basin draining to each reach (including hill-slopes and channels) Utilizing drainage area at each reachand field-measured channel gradient we calculated an esti-mated stream power as the product of drainage area chan-nel gradient and basin-averaged precipitation We utilizeda 10 m DEM for all GIS topographic measurements To es-timate clay content for each sample we used median val-ues for assigned USDA texture classes To obtain estimatedclay content moisture and OC for each core we calculatedan average weighted by the percentage of core taken up byeach soil sample We categorized wet basin samples stratifiedby floodplain type into those with standing water (wetlandsand oxbow lakes) and those with no standing water (all othertypes)

26 Statistical analyses

All statistical analyses were performed using the R statisti-cal computing software (R Core Team 2017) We conductedall analyses on three modeling groups based on the variablesmeasured in each group In the wet basin we grouped obser-vations by stratification type separating observations strati-fied by channel slope from observations stratified by flood-plain type We separated these two groups from all obser-vations in the semiarid basin which were measured consis-tently We modeled OC concentration and soil texture with amixed-effects linear regression using individual soil samples(ie the individual samples that make up a core) as sam-ple units (n= 103 for wet basin stratified by slope 89 forwet basin stratified by floodplain type and 101 for semiaridbasin) We modeled the sampled site as a random effect ac-knowledging that individual soil samples within a single coreare likely nonindependent We used profiled 95 confidenceintervals on effect estimates (β) for fixed effects to evaluatevariable importance in mixed-effects models

To gain further insight at the reach scale we also mod-eled average OC concentration and soil moisture at eachmeasured site using multiple linear regression We mod-eled soil moisture at the reach scale because we felt thatour single snapshot of moisture conditions was better rep-resented as a site-level average We first performed uni-

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1106 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

variate analysis between each hypothesized predictor andthe response utilizing mainly comparative Wilcoxon rank-sum tests (Wilcoxon 1945) or correlational Spearman cor-relation coefficient statistics We utilized a Holm multiple-comparison correction (Holm 1979) for pairwise compar-isons During this filtering we also viewed box plots or scat-terplots as appropriate to discern which variables appear tohave anything other than a completely random relationshipwith the response We then utilized the multiple linear re-gression of all subsets using the corrected Akaike informa-tion criterion as a model selection criterion (Wagenmakersand Farrell 2004) We iteratively transformed response vari-ables to ensure the homoscedasticity of error terms To selecta single best model we utilized both Akaike-weight-basedimportance and parsimony to select a final reduced modelWe considered sample sizes p values and effect magnitudesin determining variable importance

We also analyzed each core to determine whether therewere buried high-OC-concentration layers at depth We com-pared each buried soil sample to the sample above it usingthe criterion that a peak in OC at depth should have an OCconcentration 15 times that of the overlying sample and beabove 05 (Appling et al 2014)

3 Results

Model results are presented in Table 1 Comparisons betweenbasins and summaries of OC concentration moisture andestimated clay content are shown in Fig 2

31 OC concentration

Most cores display a decrease in OC concentration withdepth (Fig 3) Of cores with more than a single sample32 (722) of cores stratified by slope in the wet basin 32 (825) of cores stratified by floodplain type in the wet basinand 6 (231) of cores in the semiarid basin exhibit OC con-centration peaks at depth Whether a soil sample was classi-fied as an OC peak has no relation to estimated clay contentin sites stratified by floodplain type (p = 028) or those strat-ified by slope (p = 089) in the wet basin In the semiaridbasin soil samples classified as buried OC peaks have signif-icantly higher estimated clay contents (p = 005) than thosethat were not classified as peaks

In general the floodplain-stratified sites in the wet basinstore higher densities of OC than the semiarid basin (Fig 2ab) Figure 2a includes zero values (ie sites with no OC-bearing sediment only present in the wet basin slope-stratified group) whereas Fig 2b does not because sampleunits in Fig 2b are individual soil samples Comparing thesetwo groups it appears that soils in the wet basin exhibit muchhigher OC concentrations than those in the semiarid basinbut in general there are many more reaches with no fine sed-iment available to store OC in the wet basin

At the scale of individual soil samples the depth belowground surface is by far the dominant control on OC con-centration across all modeling groups We used a cube roottransform for all three mixed-effects models of OC concen-tration For wet basin sites stratified by slope deeper soilsamples contain less OC (β =minus00084plusmn 00042) whereassoil samples at higher elevations tend to contain more OC(β = 00010plusmn000099) Depth is the only significant predic-tor of OC content for both wet basin soil samples stratifiedby floodplain type (β =minus00084plusmn00042) and soil samplesin the semiarid basin (β =minus00037plusmn 00019)

Modeling wet basin slope-stratified sites at the reach scalewe found that moisture (β = 00078plusmn 00031) and whetherthe reach was unconfined (β = 077plusmn 049) control soil OC(cube root transformation model adjusted R2

= 054 p lt00001) Modeling wet basin floodplain-stratified sites at thesite scale we found that canopy cover (β = 0012plusmn 0011)and moisture (β = 00040plusmn 00011) are controls on soilOC (cube root transformation model adjusted R2

= 067p lt 00001) Modeling Big Sandy sites at the reach scalewe found that soil depth (β =minus0012plusmn 00071) and mois-ture (β = 0014plusmn 00026) are dominant controls on soil OCconcentration (no transformation model adjustedR2

= 069p lt 00001)

In general moister deeper soils store more OC at the reachscale whereas OC tends to vary dominantly with depth atthe scale of individual soil samples Although estimated claycontent did not emerge as a significant predictor of OC con-centration it is used to calculate clay-held water to correctour LOI-based OC concentration measurements making itimportant in determining OC for each sample

32 Soil texture

In general soil texture follows a predictable trend withriver size between model groups (Fig 2d) Floodplain type-stratified sites in the wet basin store the most clay followedby slope-stratified sites and then sites in the semiarid basin

Modeling soil texture at the individual soil sample scaleacross slope-stratified sites in the wet basin we foundwhether the reach was confined (β = 542plusmn 532) andwhether the bed material was dominantly sand (β = 1047plusmn613) to be dominant controls on estimated clay contentModeling soil texture for sites stratified by floodplain typein the wet basin yielded no significant trends In the semi-arid basin we found that either valley width (β = 00050plusmn00032) or whether the stream was unconfined (β = 041plusmn033) as well as depth below ground surface (β =minus00069plusmn00033 for model with valley width but not confinement) sig-nificantly control soil texture

To summarize sites from unconfined lower-energyreaches in the wet basin and sites from reaches with widervalley bottoms and at lower depths in the semiarid basin ex-hibited more finely textured soils

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1107

Tabl

e1

Mat

rix

ofal

lmod

els

pres

ente

din

the

text

Eac

hm

odel

islis

ted

bym

odel

grou

pre

spon

seva

riab

lea

ndsc

ale

Scal

ere

fers

toth

esa

mpl

eun

itof

the

mod

ela

ndsi

tere

fers

toa

core

with

the

resp

onse

aver

aged

over

allt

hein

divi

dual

soil

sam

ples

inth

eco

reF

orea

chva

riab

lean

dm

odel

an

aste

risk

()i

ndic

ates

that

the

vari

able

was

incl

uded

inm

odel

sele

ctio

nbu

tnot

the

full

mix

ed-e

ffec

tsm

odel

Am

inus

(minus)

indi

cate

sth

atth

eva

riab

lew

asse

lect

edas

impo

rtan

tin

pred

ictin

gth

ere

spon

sean

dde

note

san

indi

rect

corr

elat

ion

whe

reas

apl

us(+

)ind

icat

esa

dire

ctco

rrel

atio

nIn

the

case

ofco

nfine

men

ta

plus

indi

cate

sth

atun

confi

ned

stre

ams

disp

lay

ahi

gher

-mag

nitu

dere

spon

seva

riab

leI

nth

eca

seof

bed

mat

eria

la

plus

indi

cate

sth

atsa

mpl

esw

ithsa

ndex

hibi

tahi

gher

valu

eof

the

resp

onse

NA

indi

cate

sth

atei

ther

the

vari

able

was

notm

easu

red

for

that

basi

nor

mod

elgr

oup

orth

atit

isth

em

odel

resp

onse

Mod

elgr

oup

Res

pons

eSc

ale

(sam

ple

unit)

Var

iabl

esConfinement

Bedform

Channelslope

Bedmaterial

Multithread

Valleywidth

Bankfullwidth

Bankfulldepth

Streampower

Floodplaintype

Standingwater

Depth1

Claycontent

Moisture

Loggingnearby

Grassespresent

Shrubspresent

Treespresent

Elevation

Basinslope

Canopycover

NLCD

Drainagearea

Wet

basi

nst

rat-

ified

bysl

ope

OC

()

Soil

sam

ple

NA

NAminus

+

O

C(

)Si

te+

NA

NA

+

Moi

stur

e(

)Si

te+

minus

NA

NA

N

A+

Te

xtur

e(

)So

ilsa

mpl

e+

+

N

AN

A

NA

Wet

basi

nst

rat-

ified

byflo

od-

plai

nty

pe

OC

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

minus

N

AO

C(

)Si

teN

AN

AN

AN

AN

AN

AN

AN

AN

A

+

+N

AM

oist

ure

()

Site

NA

NA

NA

NA

NA

NA

NA

NA

NA

+

+

NA

NA

Text

ure2

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

N

AN

A

Sem

iari

dba

sin

OC

()

Soil

sam

ple

N

A

NA

NA

N

AN

A

NA

NAminus

NA

NA

NA

NA

OC

()

Site

N

A

NA

NA

NA

NA

N

AN

Aminus

+N

AN

AN

AN

A

M

oist

ure

()

Site

+N

A

NA

NA

NA

NA

N

AN

A+

NA

NA

NA

NA

NA+

Te

xtur

e3(

)So

ilsa

mpl

e+

NA

N

AN

A+

NA

NA

N

AN

Aminus

NA

NA

NA

NA

NA

1D

epth

refe

rsto

eith

erth

eso

ilsa

mpl

ede

pth

belo

wth

egr

ound

orth

eto

tald

epth

ofth

eco

red

epen

ding

onth

esa

mpl

eun

it2

No

sign

ifica

ntre

sults

wer

eob

serv

edfo

rthi

sm

odel

3Fo

rthi

sm

odel

bot

hva

lley

wid

than

dco

nfine

men

tpre

dict

text

ure

and

can

bein

terp

rete

din

terc

hang

eabl

yH

owev

eri

nclu

ding

both

inth

esa

me

mod

elw

ould

yiel

dpr

oble

ms

due

tom

ultic

ollin

eari

ty

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1108 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

Figure 2 Box plots showing comparisons between model groups of OC concentration at the reach scale (a) OC concentration at the scaleof individual soil samples (b) moisture at the reach scale (c) and estimated clay content at the scale of individual soil samples (d) Endsof dotted lines represent 15 times the interquartile range which is represented by boxes Bold line represents median Circles representoutliers Letters indicate probable differences between groups based on pairwise Wilcoxon (andashc) or t tests (d) with a Holm correctionRanges in parentheses below letters show the 95 confidence interval of the median value for the group (andashc) or the mean value for thegroup (d) in which median confidence intervals were overly constrained due to the categorical nature of our estimated clay content data

33 Soil moisture

Soil moisture is less variable between basins than either tex-ture or OC concentration (Fig 2c) All model groups exhibitsimilar soil moisture conditions although there was signifi-cant variability within each model group

Soil moisture at wet basin sites stratified by slope is domi-nantly controlled by channel slope (β =minus1315plusmn911) ele-vation (β = 00046plusmn 00038) and whether the stream is un-confined (β = 389plusmn 267 model adjusted R2

= 038 p lt00001) At wet basin sites stratified by floodplain type es-timated clay content (β = 0060plusmn 0042) and whether thefloodplain unit had standing water (β = 115plusmn 071) sig-nificantly control soil moisture (model adjusted R2

= 046p lt 00001) In the semiarid basin soil depth (β = 0012plusmn0012) elevation (β = 00023plusmn 00014) and whether thereach was unconfined (β = 091plusmn075) significantly controlsoil moisture (model adjusted R2

= 035 p lt 00001)

4 Discussion

41 Understanding spatial variability in OCconcentration in floodplain soils (H1 and H2)

Comparing the wet basin to the semiarid basin shows thatthe wetter higher primary productivity basin is capable ofstoring greater concentrations of OC in floodplain soils butthat both regions generally store similar OC concentrations infloodplain soils This result partially agrees with the exami-nation of subalpine lake deltas by Scott and Wohl (2017) Inthat study subalpine lake deltas in the wet basin were com-pared to deltas in the drier Colorado Front Range Subalpinelake deltas displayed similar OC concentrations likely dueto competing but complementary OC stabilization and lossmechanisms in each region Those deltas represent a subsetof the broader valley bottom soils studied here This moreexpansive study points to both geomorphic controls such asvalley bottom geometry and factors influenced by climatesuch as canopy cover as controls on OC storage in valley bot-toms These results also agree with the results of Lininger et

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1109

Figure 3 Box plots of sample OC concentration binned by meansample depth for the wet (a) and semiarid (b) basins Ends of dot-ted lines represent 15 times the interquartile range which is rep-resented by boxes Bold line represents median Black circles rep-resent outliers Transparent grey points show all data for each binSample size for each bin is denoted by n

al (2018) which indicate that geomorphic context and vege-tation dynamics control OC concentration on floodplain soilsalong large lowland rivers in Alaska USA

At the reach or site scale wetter soil profiles consistentlyyield higher OC concentrations in all model groups How-ever moisture does not differ significantly between modelgroups (Fig 2c) indicating that this alone cannot explaindifferences between basins Soils tend to be finer in the wetbasin but clay content was not an important predictor of OCconcentration in studied soils Clay content likely influencesOC concentration based on previous research (Hoffmann etal 2009) although the inclusion of coarse soil material (in-cluding particulate organic matter) in our samples may ex-plain the lack of an observed correlation here Variabilityin soil redox conditions (ie the presence of anaerobic mi-crosites) may also introduce variability in respiration rate be-tween basins and among individual soil samples (Keiluweit

et al 2017) That is variability in groundwater hydrologymay result in a heterogeneous mix of aerobic (favoring mi-crobial activity) and anaerobic (suppressing microbial activ-ity) conditions within floodplain soils We did not record re-doximorphic features of our soil samples so we are unableto determine whether this influence contributes to the errorin our models Although confinement plays a strong role indetermining OC concentration in wet basin sites stratified byslope it does not differ significantly between basins (52 of semiarid basin reaches are unconfined compared to 63 of wet basin reaches) The major differences between thesebasins are their hydroclimatic and disturbance regimes Thewet basin is at a lower elevation and has denser and higherbiomass forests (Smithwick et al 2002) compared to thesparser parkland forests of the semiarid basin which likelyalso experiences more frequent fires based on fire historiesof nearby regions (recurrence interval on the order of 101ndash102 years Houston 1973 Loope and Gruell 1973) In addi-tion the volcanic soils present in the wet basin may suppresssoil OC respiration (Matus et al 2014) leading to a highersoil OC storage capacity there

Between basins it is likely that the hydroclimatic regimeinfluencing primary production plays some role in the wetbasinrsquos higher maximum OC concentrations in floodplainsoils compared to those of the semiarid basin Howeversmaller-scale factors such as soil texture and moisture alsolikely play a role and are not related to drainage area (Ta-ble 1) indicating that neither OC concentration nor its con-trolling factors vary continuously along a river network thussupporting H2a and H2b This also indicates that local fac-tors set largely by geomorphic and hydrologic dynamicsplay a significant role in modulating the effect of climate onOC concentrations If the wet and semiarid basins displayedsignificantly different OC concentrations our first hypothesisregarding the interbasin controls on OC concentration wouldbe supported However we instead found that climate andprimary productivity only partially determine OC concentra-tions especially when viewed in the context of geomorphicand hydrologic variability Thus the results do not supportH1

Each basin (or model group) is slightly different in termsof the controls on soil OC concentration moisture andtexture In the wet basin sites stratified by slope higher-elevation sites display higher OC concentrations This is con-trary to the general trend in primary productivity which de-creases with increasing elevation However it is importantto note that the headwaters of the wet basin are dominatedby lakes deltas and other depositional features in relativelybroad glacially carved valleys Subalpine lake deltas havebeen shown to store high OC concentrations in this basin(Scott and Wohl 2017) and many of the highest OC concen-trations we measured were located in broad wet meadowssubalpine lake deltas or other unconfined high-elevationreaches Such unconfined sites also likely have significantlycooler temperatures and tend to have higher soil moisture

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1110 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

contents as shown by our modeling (Table 1) As such al-though high-elevation wet basin sites may receive less OCinput they likely have a low rate of OC respiration resultingin higher OC concentrations on the whole which agrees withthe result of Bao et al (2017) In the semiarid basin our mod-eling suggests that the lower temperatures and higher mois-ture (Table 1) at higher elevations do not compensate for thelower primary productivity as elevation does not correlatewith OC concentration

In both basins unconfined reaches contain wetter andfiner-textured soils which may result in a higher soil OCcapacity Although confinement only relates directly to OCcontent in wet basin sites it does play a strong role in deter-mining moisture which in turn plays a role in regulating OCconcentration in both basins likely via inhibiting microbialactivity (Howard and Howard 1993) The relevance of chan-nel slope in determining soil moisture in the wet basin but notsemiarid basin may reflect the prevalence of high-gradientdebris-flow-dominated channels in the wet basin that largelyexhibit only gravel to boulder substrate which we assumestores minimal fine sediment moisture or OC

In the semiarid basin higher soil depths correlate withmoister and finer-textured soils but less OC concentrationThis indicates the trend in OC with depth likely dominatesthe signal of OC concentration with deeper sites containinga higher proportion of OC-depleted deep samples

42 Inferring sources of OC to floodplain soils (H3)

OC can be input to floodplain soils by two primary mech-anisms First dissolved and particulate OC (including largewood) can be deposited on floodplain surfaces by overbankdeposition thus integrating fluvial sedimentary OC into thefloodplain soil profile or in the case of large wood deposit-ing discrete but potentially large concentrations of OC thatcan later be integrated into the soil profile Second litter anddecomposing vegetation on the floodplain surface can act asautochthonous inputs of OC to floodplain soil

Our modeling of OC concentration yielded results consis-tent with previous investigations of controls on soil OC stor-age capacity (Jobbaacutegy and Jackson 2000 Sutfin and Wohl2017) Sites in the heterogeneous floodplain of the wet basindisplay a direct correlation between canopy cover and OCconcentration indicating that increased litter inputs lead toincreased floodplain soil OC concentration Sediment inputslikely differ between floodplain depositional unit types (egcoarser sediment may deposit on point bars compared tofilled secondary channels) although floodplain type does notpredict OC concentration This indicates that vegetation in-puts may be more dominant than fluvial sediment inputs atthese sites

The finding that buried OC peaks in the wet basin do nothave abnormally high clay contents supports the interpre-tation that wood and litter inputs to soil are the dominantsource of OC in the floodplain soils we examined Buried

peaks can be layers created by overbank deposition and sub-sequent burial of fine OC-bearing sediments (Blazejewski etal 2009 Ricker et al 2013) buried pieces of wood (Wohl2013) or buried organic horizons that are now capped bysediments that prevent OC respiration If overbank deposi-tion of fine sediment caused OC peaks we would expect tosee the soil samples classified as peaks exhibiting high claycontents indicating finer sediment Instead our results sug-gest that in the wet basin buried peaks are likely the result ofeither buried organic horizons or buried wood We observedlarge pieces of decaying buried wood in floodplain cut banksin the wet basin supporting this inference Overbank depo-sition of wood on the floodplain was only observed rarely inthis basin indicating that the OC measured in these soils islikely dominantly autochthonous

In the semiarid basin the two cores that exhibit peaks werecollected from the same meadow just downstream of a now-filled former lake that is a potential source of fine sedimentThe channels draining this meadow exhibit an anabranchingplanform indicating the potential to deposit and bury packetsof potentially OC-rich fine sediments However the majorityof cores do not exhibit OC peaks indicating OC input mainlyfrom vegetation at the surface and continuing OC respirationat depth

OC variation within each core is dominantly a function ofdepth We observe a negative correlation between depth be-low ground surface and OC concentration which has beenobserved in other studies including mountain wetlands andfloodplains (Jobbaacutegy and Jackson 2000 Scott and Wohl2017 Sutfin and Wohl 2017 Zhao et al 2017) In generalthis indicates that at least in mountain river floodplains OCis enriched at the surface and decomposes with depth sim-ilar to upland soils This fits with our finding that the ma-jority of our cores do not exhibit significant OC peaks atdepth and supports the dominance of litter and wood OC in-puts to floodplain soils These results support our hypothesisthat decaying litter and wood not overbank sediment depo-sition dominate the input of OC to floodplain soils in ourstudy basins (H3) We note that floodplain wood may alsoact as a trapping site for overbank fine organic matter facil-itating the deposition and input of OC from decaying vege-tation While these study basins likely accumulate soil OCmainly autochthonously other basins that experience over-bank flows accompanying deposition of fine sediment andburial of organic layers exhibit OC storage that is likely dom-inated by fluvial sediment deposition (eg Blazejewski etal 2009 DrsquoElia et al 2017 Ricker et al 2013) Thus itis likely that flow regime lateral connectivity and sedimenttransport dynamics regulate whether floodplain soil OC isdominantly input by the overbank deposition of fine mate-rial or litter and wood decay

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1111

Figure 4 Conceptual model of physical processes that influenceOC concentration in floodplain soils based on this study and otherliterature Each box corresponds to a major factor that influencesOC concentration Colored text within each box denotes factors thatinfluence OC inputs sorption capacity OC export or respirationrate (note that some processes influence both OC export and respi-ration rate) As sorption capacity increases so does the OC capacityof the soil Conversely as the rate of respiration andor OC exportincreases the soil OC capacity decreases Floodplain soils can onlydevelop high concentrations of OC if there are high rates of OCinput However the capacity of the soils to store OC regulates thatinput and is determined by the competing influences of sorption ca-pacity and the combination of respiration rate and OC export Seethe text for further details

43 Conceptual model of soil OC concentration infloodplain soils

We present a conceptual model to summarize our results andplace them in the context of work examining the controlson OC storage in soils (Fig 4) OC is input to floodplainseither through the decay of vegetation or the deposition offine OC-rich sediment This input of OC only determinesOC concentrations insofar as floodplain soils are capable ofstoring OC That storage is effectively determined by a bal-ance between processes that remove OC from floodplainsnamely respiration or erosion followed by respiration (Berheet al 2007) and processes that regulate OC availability tomicrobes namely the capability of the mineral fraction ofthe soil to sorb OC

OC sorption capacity reflects a few specific processes Al-though soil texture generally relates to the ability of OCto sorb to mineral grains and resulting OC availability soilchemistry also plays a strong and potentially dominant rolein regulating OC sorption capacity (Rasmussen et al 2018)Soil texture is largely determined by valley morphology ac-cording to our modeling (Table 1) placing valley morphol-ogy and resulting sediment transport dynamics (Gran andCzuba 2017 Wohl et al 2017b) as indirect controls on sorp-tion capacity

Respiration rate is largely determined by microbial activityand the availability of OC to microbes Erosion can rapidlyexpose soil OC to microbial respiration (Berhe et al 2007)whereas soils that reside in largely anoxic conditions can ex-hibit low rates of microbial respiration (Boye et al 2017)In addition soil mineralogy chemistry and redox condi-tions (Keiluweit et al 2017) can regulate microbial activity(eg andic soils may limit microbial respiration Matus et al2014) Our results suggesting that moisture controls OC con-tent support the idea that drier soils likely have higher ratesof microbial respiration of OC Moisture is a function of tex-ture valley bottom morphology and elevation (a proxy fortemperature) in our modeling (Table 1) Comparing flood-plain types in the wet basin we find that types with standingwater exhibit significantly higher soil moisture contents thanthose without standing water This indicates spatial variabil-ity in moisture content and likely microbial activity (Howardand Howard 1993) In our modeling this effect translates tospatial variability in OC concentration within floodplains andacross entire basins

OC export from soils refers specifically to leaching of dis-solved OC mainly from shallow soils While we do not di-rectly consider OC export in this study it likely introducesvariability in soil OC concentration and is regulated dom-inantly by hydrology (Aringgren et al 2014 Mcdowell andLikens 1988) soil pore water chemistry (including pH andionic strength Brooks et al 1999 Evans et al 2006) andredox conditions (Knorr 2013) OC export as DOC can actin conjunction with respiration to remove OC from soils andlike respiration rate is countered by sorption capacity

To summarize OC inputs regulated by the capacity ofsoils to store OC and suppress microbial respiration andOC export determine OC concentrations in floodplain soils(Fig 4) OC inputs to floodplain soils come from either au-tochthonous litter accumulation on the floodplain surface al-lochthonous wood deposition or allochthonous deposition offine OC-bearing sediments Our results from the mountain-ous basins we studied suggest that deposition of fine mate-rial in overbank flows is rare leading us to infer that au-tochthonous litter and allochthonous wood inputs to flood-plains dominate OC input in mountain rivers Where soils aremore moist microbial respiration is inhibited and more OCis stored Although soil texture is likely not a limiting factoron OC concentration in these floodplains finer-textured soilslikely have a higher sorption capacity retaining more of theOC input from decaying plant material Our results indicatethat geomorphic and hydrologic characteristics act as bound-ary conditions that regulate soil texture and moisture in turnregulating sorption capacity respiration rate and resultingOC concentrations in floodplain soils

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1112 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

5 Conclusions

We present floodplain soil OC concentration data from twodisparate watersheds to compare how interbasin variabilitybetween the two watersheds compares with intra-basin vari-ability in geomorphic and hydrologic characteristics in de-termining OC concentration Our results indicate that OCconcentration in mountain floodplain soils does not vary pre-dictably along a longitudinal gradient nor does it vary sub-stantially between basins with differing climatic and veg-etation characteristics Instead geomorphic and hydrologiccharacteristics such as valley bottom morphology and soilmoisture dominantly determine floodplain OC concentra-tion

In our study basins decaying litter and wood and not over-bank deposition of fine OC-bearing sediment is the mainsource of OC to floodplain soils It is unclear whether thatdecaying vegetation is dominated by autochthonous litter in-puts or transported downed wood In comparing our basinto other studied floodplain soils it seems that vegetation dy-namics play a strong role in determining OC concentrationswhen fine sediment is not regularly deposited on floodplainsurfaces However we suggest that floodplain soil character-istics set by geomorphic and hydrologic conditions regu-late how OC inputs translate to the spatial distribution of OCalong a river network

This implies that OC storage in floodplains likely cannotbe predicted using consistent downstream trends and thatmanagement prioritization designed to facilitate floodplainOC storage should be based on local geomorphic and hy-drologic process variability within each basin For instancemanagement to increase OC sequestration in floodplain soilswill likely be more effective where floodplains are uncon-fined and soils already experience high moisture conditionsfor much of the year Along these lines our results showthat modeling the floodplain biospheric OC pool to predictits response to warming and subsequent effects on climatebased on regional factors such as climate and net primaryproductivity likely misses the substantial interbasin variabil-ity in OC concentration and storage resulting from variabilityin valley bottom geometry and both geomorphic and hydro-logic processes (eg Doetterl et al 2015)

Although our results provide some insights the questionof whether OC stored in floodplain soil comes dominantlyfrom allochthonous versus autochthonous sources remainsopen Our results imply that more productive spatially het-erogeneous floodplains likely input more OC to soils Flood-plain OC concentration while mediated largely by moisturedynamics likely depends mainly on OC inputs from produc-tive riparian forests This implies that management of OCstorage in mountain river floodplains should focus on therestoration of riparian zones to maintain OC input to soil(eg Bullinger-Weber et al 2014) More detailed studiesin regions with varying sediment transport and hydrologicregimes are needed to determine what conditions favor au-

tochthonous versus allochthonous OC inputs but our resultssuggest that autochthonous sources dominate floodplain OCstorage in basins with relatively low rates of vertical accre-tion and high channelndashfloodplain connectivity that promotesfloodplain wetlands

Data availability Data to support the analyses presented here canbe found in Table S1 and in the Colorado State University DigitalRepository (Scott and Wohl 2018a)

Supplement The supplement related to this article is availableonline at httpsdoiorg105194esurf-6-1101-2018-supplement

Author contributions Field data collection was conducted pri-marily by DNS with assistance from EEW at some sites in theMiddle Fork Snoqualmie basin DNS performed data analysis anddrafted the initial version of the paper which was then edited andrewritten by DNS and EEW

Competing interests The authors declare that they have no con-flict of interest

Acknowledgements This work was funded by NSF grantEAR-1562713 We thank Ellen Daugherty for extensive assistancein fieldwork and discussion that improved the conceptual modelWe thank Katherine Lininger for stimulating discussion thatimproved the paper Detailed and constructive comments from twoanonymous reviewers and Robert Hilton improved the paper Wethank Sara Lowe for assistance in processing and analyzing soilsamples

Edited by Robert HiltonReviewed by two anonymous referees

References

Aringgren A M Buffam I Cooper D M Tiwari T Evans C Dand Laudon H Can the heterogeneity in stream dissolved or-ganic carbon be explained by contributing landscape elementsBiogeosciences 11 1199ndash1213 httpsdoiorg105194bg-11-1199-2014 2014

Appling A P Bernhardt E S and Stanford J A Flood-plain biogeochemical mosaics A multi-dimensional view ofalluvial soils J Geophys Res-Biogeo 119 1538ndash1553httpsdoiorg1010022013JG002543 2014

Aufdenkampe A K Mayorga E Raymond P A MelackJ M Doney S C Alin S R Aalto R E and YooK Riverine coupling of biogeochemical cycles between landoceans and atmosphere Front Ecol Environ 9 53ndash60httpsdoiorg101890100014 2011

Ballantyne A P Alden C B Miller J B Tans P P and WhiteJ W C Increase in observed net carbon dioxide uptake by

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

1104 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

Figure 1 Hillshade showing the location sampling sites and stream network of the sampled basins Big Sandy Wyoming (4269minus10925) is on the left and MF Snoqualmie Washington (4753 minus12152) is on the right Circles represent sampling locations atwhich floodplain soil OC was measured Sample sites are colored by OC concentration Mean annual precipitation (MAP) drainage area(DA) and relief are given for each basin

egories) Within each of these five categories we randomlyselected six points at which to take soil cores

We also stratified the entire wet basin stream network bychannel slope into four strata Within each channel slopestrata (hereafter referred to as slope strata) we randomly se-lected 10 reaches to collect a single floodplain soil core re-sulting in 40 randomly located sample sites

To supplement randomly sampled sites and accommodatethe infeasibility of accessing two of the randomly sampledsites along the stream network we also subjectively selectedsample sites in places that we felt enhanced the degree towhich our sampling captured the variability present amongstreams in the basin This resulted in a total of 46 sites strat-ified by slope 38 of which were randomly sampled in addi-tion to 30 sites stratified by floodplain type

23 Reach-scale field measurements

At each sampled reach (100 m or 10 channel widthswhichever was shorter) we measured channel geometry andother characteristics although our measurements were notconsistent across all basins because field protocol evolvedduring the course of the study In both basins we measuredconfinement valley bottom width and channel bed slope Weadditionally measured bankfull width and depth in the wetbasin We did not measure channel characteristics for sitesstratified by floodplain type in the wet basin since they didnot correspond to a single reach of channel as sites stratifiedby slope did in much more confined valleys

In the wet basin we also categorized channels by planformand dominant bedform (Montgomery and Buffington 1997)We defined planforms as follows straight in which the chan-nel was generally confined and significant lateral migrationwas not evident meandering in which lateral migration wasevident but only a single channel existed anastomosing inwhich vegetated islands separated multiple channels and

anabranching in which a single dominant channel existedwith relict channels separated by vegetated islands We fur-ther classified channels as being either multithread (anasto-mosing or anabranching) or single thread (straight or mean-dering) Because logging records are inconsistent and likelyinaccurate in the wet basin (based on the frequent observa-tion of past logging activity that was not recorded in ForestService records) we noted whether signs of logging suchas cut stumps cable decommissioned roads or railroads orother logging-associated tools were found near the reach

We chose a representative location on the floodplain foreach sampled site based on visual examination of vegeta-tion type soil surface texture surface water presence andelevation relative to the bankfull channel elevation (flood-plain sites stratified by type in the wet basin were sampledas close to the randomly sampled point as possible) Oncea location was chosen we extracted a 32 mm diameter soilcore using an open-sided corer (JMC Large Diameter Sam-pling Tube) Due to our adaptive methodology we sampledsoil OC slightly differently in the semiarid versus the wetbasin In the semiarid basin we cored in irregular incrementsgenerally 25ndash30 cm After analyzing data from the semiaridbasin we realized that sampling in regular increments wouldmake analysis more versatile As a result we switched to ex-tracting soil samples at regular 20 cm increments in the wetbasin Cores were taken to refusal (ie coarse gravel or otherobstructions preventing further soil collection) or a depth ofapproximately 1 m Five cores in the semiarid basin 12 coresin the wet basin sites stratified by slope and 11 cores in thewet basin sites stratified by floodplain type did not reach re-fusal When no sand or finer sediment was present in the val-ley bottom (only occurred in headwater channels of the wetbasin) we recorded negligible OC concentration Once soilsamples were removed from the ground they were placed in

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1105

ziplock bags frozen within 72 h (most samples were frozenwithin 8 h) and kept frozen until analysis

24 Measuring soil OC and texture

To measure the concentration of organic carbon in soil sam-ples we used loss on ignition (LOI) We first defrosted sam-ples for 24ndash48 h at room temperature Once defrosted wethoroughly mixed samples to ensure the most homogenoussample possible We then subsampled 10ndash85 g of soil fromeach sample for analysis Using crucibles in a muffle furnacewe dried samples in batches of 30 for 24 h at 105 C to deter-mine moisture content and remove all nonstructurally heldwater Following the guidelines suggested by Hoogsteen etal (2015) we then burned samples for 3 h at 550 C to re-move organic matter By comparing the weight of the burnedsamples with that of the dried samples we obtained an LOIweight

After performing LOI we used burned samples to performtexture by feel to determine the USDA soil texture class andestimated clay content (Thien 1979) To convert LOI weightto OC concentration we used the structural water loss cor-rection of Hoogsteen et al (2015) using clay content esti-mated from soil texture This correction considers water heldby clay that may not evaporate during drying but will evap-orate during burning It also estimates the proportion of theLOI weight that is OC This correction represented as a per-cent of the estimated sample OC content after the correctionranged from 090 to 49576 (95 confidence interval ofthe median between 1617 and 2425 ) for the wet basinand ranged from 554 to 57053 (95 confidence in-terval of the median between 1976 and 3211 ) for thesemiarid basin

One potential confounding factor in LOI is carbonates thatmay burn off during ignition adding to the LOI weight whilenot being organic matter In lithologies where carbonates arerare (eg granitoid rocks like those found in the upper partof the wet basin and entire semiarid basins) this is a rela-tively negligible issue However some of our soil samplescame from parts of the wet basin draining rocks of the west-ern meacutelange belt including argillite greywacke and marbleWe tested samples for the presence of carbonates to deter-mine whether our LOI methods would be sufficient to accu-rately determine OC concentration We randomly chose 10soil samples of a total of 110 drained rocks that could includecarbonates and submitted them to the Colorado State Univer-sity soil testing laboratory for CHN furnace analysis (Sparks1996) which yielded data on the proportion of carbonatesby mass in those samples The median calcium carbonateconcentrations of those 10 samples was 098 (95 con-fidence interval between 070 and 103 ) and the medianpercentage of the total carbon in samples comprised of inor-ganic carbon was 505 (95 confidence interval between197 and 189 ) From this we concluded that the amountof carbonate in the samples draining potentially carbonate-

bearing rocks was low enough that LOI was likely to still beaccurate Consequently we analyzed all soil samples usingLOI to obtain OC concentration The median difference be-tween the LOI and CHN OC estimates for these 10 sampleswas 045 (95 confidence interval between minus083 and147 ) indicating no systematic bias in LOI estimates ofOC concentration

25 GIS and derivative measurements

After fieldwork in each basin we collected the following datafor each reach using a GIS platform elevation drainage arealand cover classification and canopy cover from the NationalLand Cover Database (Homer et al 2015) and the meanslope of the basin draining to each reach (including hill-slopes and channels) Utilizing drainage area at each reachand field-measured channel gradient we calculated an esti-mated stream power as the product of drainage area chan-nel gradient and basin-averaged precipitation We utilizeda 10 m DEM for all GIS topographic measurements To es-timate clay content for each sample we used median val-ues for assigned USDA texture classes To obtain estimatedclay content moisture and OC for each core we calculatedan average weighted by the percentage of core taken up byeach soil sample We categorized wet basin samples stratifiedby floodplain type into those with standing water (wetlandsand oxbow lakes) and those with no standing water (all othertypes)

26 Statistical analyses

All statistical analyses were performed using the R statisti-cal computing software (R Core Team 2017) We conductedall analyses on three modeling groups based on the variablesmeasured in each group In the wet basin we grouped obser-vations by stratification type separating observations strati-fied by channel slope from observations stratified by flood-plain type We separated these two groups from all obser-vations in the semiarid basin which were measured consis-tently We modeled OC concentration and soil texture with amixed-effects linear regression using individual soil samples(ie the individual samples that make up a core) as sam-ple units (n= 103 for wet basin stratified by slope 89 forwet basin stratified by floodplain type and 101 for semiaridbasin) We modeled the sampled site as a random effect ac-knowledging that individual soil samples within a single coreare likely nonindependent We used profiled 95 confidenceintervals on effect estimates (β) for fixed effects to evaluatevariable importance in mixed-effects models

To gain further insight at the reach scale we also mod-eled average OC concentration and soil moisture at eachmeasured site using multiple linear regression We mod-eled soil moisture at the reach scale because we felt thatour single snapshot of moisture conditions was better rep-resented as a site-level average We first performed uni-

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1106 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

variate analysis between each hypothesized predictor andthe response utilizing mainly comparative Wilcoxon rank-sum tests (Wilcoxon 1945) or correlational Spearman cor-relation coefficient statistics We utilized a Holm multiple-comparison correction (Holm 1979) for pairwise compar-isons During this filtering we also viewed box plots or scat-terplots as appropriate to discern which variables appear tohave anything other than a completely random relationshipwith the response We then utilized the multiple linear re-gression of all subsets using the corrected Akaike informa-tion criterion as a model selection criterion (Wagenmakersand Farrell 2004) We iteratively transformed response vari-ables to ensure the homoscedasticity of error terms To selecta single best model we utilized both Akaike-weight-basedimportance and parsimony to select a final reduced modelWe considered sample sizes p values and effect magnitudesin determining variable importance

We also analyzed each core to determine whether therewere buried high-OC-concentration layers at depth We com-pared each buried soil sample to the sample above it usingthe criterion that a peak in OC at depth should have an OCconcentration 15 times that of the overlying sample and beabove 05 (Appling et al 2014)

3 Results

Model results are presented in Table 1 Comparisons betweenbasins and summaries of OC concentration moisture andestimated clay content are shown in Fig 2

31 OC concentration

Most cores display a decrease in OC concentration withdepth (Fig 3) Of cores with more than a single sample32 (722) of cores stratified by slope in the wet basin 32 (825) of cores stratified by floodplain type in the wet basinand 6 (231) of cores in the semiarid basin exhibit OC con-centration peaks at depth Whether a soil sample was classi-fied as an OC peak has no relation to estimated clay contentin sites stratified by floodplain type (p = 028) or those strat-ified by slope (p = 089) in the wet basin In the semiaridbasin soil samples classified as buried OC peaks have signif-icantly higher estimated clay contents (p = 005) than thosethat were not classified as peaks

In general the floodplain-stratified sites in the wet basinstore higher densities of OC than the semiarid basin (Fig 2ab) Figure 2a includes zero values (ie sites with no OC-bearing sediment only present in the wet basin slope-stratified group) whereas Fig 2b does not because sampleunits in Fig 2b are individual soil samples Comparing thesetwo groups it appears that soils in the wet basin exhibit muchhigher OC concentrations than those in the semiarid basinbut in general there are many more reaches with no fine sed-iment available to store OC in the wet basin

At the scale of individual soil samples the depth belowground surface is by far the dominant control on OC con-centration across all modeling groups We used a cube roottransform for all three mixed-effects models of OC concen-tration For wet basin sites stratified by slope deeper soilsamples contain less OC (β =minus00084plusmn 00042) whereassoil samples at higher elevations tend to contain more OC(β = 00010plusmn000099) Depth is the only significant predic-tor of OC content for both wet basin soil samples stratifiedby floodplain type (β =minus00084plusmn00042) and soil samplesin the semiarid basin (β =minus00037plusmn 00019)

Modeling wet basin slope-stratified sites at the reach scalewe found that moisture (β = 00078plusmn 00031) and whetherthe reach was unconfined (β = 077plusmn 049) control soil OC(cube root transformation model adjusted R2

= 054 p lt00001) Modeling wet basin floodplain-stratified sites at thesite scale we found that canopy cover (β = 0012plusmn 0011)and moisture (β = 00040plusmn 00011) are controls on soilOC (cube root transformation model adjusted R2

= 067p lt 00001) Modeling Big Sandy sites at the reach scalewe found that soil depth (β =minus0012plusmn 00071) and mois-ture (β = 0014plusmn 00026) are dominant controls on soil OCconcentration (no transformation model adjustedR2

= 069p lt 00001)

In general moister deeper soils store more OC at the reachscale whereas OC tends to vary dominantly with depth atthe scale of individual soil samples Although estimated claycontent did not emerge as a significant predictor of OC con-centration it is used to calculate clay-held water to correctour LOI-based OC concentration measurements making itimportant in determining OC for each sample

32 Soil texture

In general soil texture follows a predictable trend withriver size between model groups (Fig 2d) Floodplain type-stratified sites in the wet basin store the most clay followedby slope-stratified sites and then sites in the semiarid basin

Modeling soil texture at the individual soil sample scaleacross slope-stratified sites in the wet basin we foundwhether the reach was confined (β = 542plusmn 532) andwhether the bed material was dominantly sand (β = 1047plusmn613) to be dominant controls on estimated clay contentModeling soil texture for sites stratified by floodplain typein the wet basin yielded no significant trends In the semi-arid basin we found that either valley width (β = 00050plusmn00032) or whether the stream was unconfined (β = 041plusmn033) as well as depth below ground surface (β =minus00069plusmn00033 for model with valley width but not confinement) sig-nificantly control soil texture

To summarize sites from unconfined lower-energyreaches in the wet basin and sites from reaches with widervalley bottoms and at lower depths in the semiarid basin ex-hibited more finely textured soils

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1107

Tabl

e1

Mat

rix

ofal

lmod

els

pres

ente

din

the

text

Eac

hm

odel

islis

ted

bym

odel

grou

pre

spon

seva

riab

lea

ndsc

ale

Scal

ere

fers

toth

esa

mpl

eun

itof

the

mod

ela

ndsi

tere

fers

toa

core

with

the

resp

onse

aver

aged

over

allt

hein

divi

dual

soil

sam

ples

inth

eco

reF

orea

chva

riab

lean

dm

odel

an

aste

risk

()i

ndic

ates

that

the

vari

able

was

incl

uded

inm

odel

sele

ctio

nbu

tnot

the

full

mix

ed-e

ffec

tsm

odel

Am

inus

(minus)

indi

cate

sth

atth

eva

riab

lew

asse

lect

edas

impo

rtan

tin

pred

ictin

gth

ere

spon

sean

dde

note

san

indi

rect

corr

elat

ion

whe

reas

apl

us(+

)ind

icat

esa

dire

ctco

rrel

atio

nIn

the

case

ofco

nfine

men

ta

plus

indi

cate

sth

atun

confi

ned

stre

ams

disp

lay

ahi

gher

-mag

nitu

dere

spon

seva

riab

leI

nth

eca

seof

bed

mat

eria

la

plus

indi

cate

sth

atsa

mpl

esw

ithsa

ndex

hibi

tahi

gher

valu

eof

the

resp

onse

NA

indi

cate

sth

atei

ther

the

vari

able

was

notm

easu

red

for

that

basi

nor

mod

elgr

oup

orth

atit

isth

em

odel

resp

onse

Mod

elgr

oup

Res

pons

eSc

ale

(sam

ple

unit)

Var

iabl

esConfinement

Bedform

Channelslope

Bedmaterial

Multithread

Valleywidth

Bankfullwidth

Bankfulldepth

Streampower

Floodplaintype

Standingwater

Depth1

Claycontent

Moisture

Loggingnearby

Grassespresent

Shrubspresent

Treespresent

Elevation

Basinslope

Canopycover

NLCD

Drainagearea

Wet

basi

nst

rat-

ified

bysl

ope

OC

()

Soil

sam

ple

NA

NAminus

+

O

C(

)Si

te+

NA

NA

+

Moi

stur

e(

)Si

te+

minus

NA

NA

N

A+

Te

xtur

e(

)So

ilsa

mpl

e+

+

N

AN

A

NA

Wet

basi

nst

rat-

ified

byflo

od-

plai

nty

pe

OC

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

minus

N

AO

C(

)Si

teN

AN

AN

AN

AN

AN

AN

AN

AN

A

+

+N

AM

oist

ure

()

Site

NA

NA

NA

NA

NA

NA

NA

NA

NA

+

+

NA

NA

Text

ure2

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

N

AN

A

Sem

iari

dba

sin

OC

()

Soil

sam

ple

N

A

NA

NA

N

AN

A

NA

NAminus

NA

NA

NA

NA

OC

()

Site

N

A

NA

NA

NA

NA

N

AN

Aminus

+N

AN

AN

AN

A

M

oist

ure

()

Site

+N

A

NA

NA

NA

NA

N

AN

A+

NA

NA

NA

NA

NA+

Te

xtur

e3(

)So

ilsa

mpl

e+

NA

N

AN

A+

NA

NA

N

AN

Aminus

NA

NA

NA

NA

NA

1D

epth

refe

rsto

eith

erth

eso

ilsa

mpl

ede

pth

belo

wth

egr

ound

orth

eto

tald

epth

ofth

eco

red

epen

ding

onth

esa

mpl

eun

it2

No

sign

ifica

ntre

sults

wer

eob

serv

edfo

rthi

sm

odel

3Fo

rthi

sm

odel

bot

hva

lley

wid

than

dco

nfine

men

tpre

dict

text

ure

and

can

bein

terp

rete

din

terc

hang

eabl

yH

owev

eri

nclu

ding

both

inth

esa

me

mod

elw

ould

yiel

dpr

oble

ms

due

tom

ultic

ollin

eari

ty

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1108 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

Figure 2 Box plots showing comparisons between model groups of OC concentration at the reach scale (a) OC concentration at the scaleof individual soil samples (b) moisture at the reach scale (c) and estimated clay content at the scale of individual soil samples (d) Endsof dotted lines represent 15 times the interquartile range which is represented by boxes Bold line represents median Circles representoutliers Letters indicate probable differences between groups based on pairwise Wilcoxon (andashc) or t tests (d) with a Holm correctionRanges in parentheses below letters show the 95 confidence interval of the median value for the group (andashc) or the mean value for thegroup (d) in which median confidence intervals were overly constrained due to the categorical nature of our estimated clay content data

33 Soil moisture

Soil moisture is less variable between basins than either tex-ture or OC concentration (Fig 2c) All model groups exhibitsimilar soil moisture conditions although there was signifi-cant variability within each model group

Soil moisture at wet basin sites stratified by slope is domi-nantly controlled by channel slope (β =minus1315plusmn911) ele-vation (β = 00046plusmn 00038) and whether the stream is un-confined (β = 389plusmn 267 model adjusted R2

= 038 p lt00001) At wet basin sites stratified by floodplain type es-timated clay content (β = 0060plusmn 0042) and whether thefloodplain unit had standing water (β = 115plusmn 071) sig-nificantly control soil moisture (model adjusted R2

= 046p lt 00001) In the semiarid basin soil depth (β = 0012plusmn0012) elevation (β = 00023plusmn 00014) and whether thereach was unconfined (β = 091plusmn075) significantly controlsoil moisture (model adjusted R2

= 035 p lt 00001)

4 Discussion

41 Understanding spatial variability in OCconcentration in floodplain soils (H1 and H2)

Comparing the wet basin to the semiarid basin shows thatthe wetter higher primary productivity basin is capable ofstoring greater concentrations of OC in floodplain soils butthat both regions generally store similar OC concentrations infloodplain soils This result partially agrees with the exami-nation of subalpine lake deltas by Scott and Wohl (2017) Inthat study subalpine lake deltas in the wet basin were com-pared to deltas in the drier Colorado Front Range Subalpinelake deltas displayed similar OC concentrations likely dueto competing but complementary OC stabilization and lossmechanisms in each region Those deltas represent a subsetof the broader valley bottom soils studied here This moreexpansive study points to both geomorphic controls such asvalley bottom geometry and factors influenced by climatesuch as canopy cover as controls on OC storage in valley bot-toms These results also agree with the results of Lininger et

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1109

Figure 3 Box plots of sample OC concentration binned by meansample depth for the wet (a) and semiarid (b) basins Ends of dot-ted lines represent 15 times the interquartile range which is rep-resented by boxes Bold line represents median Black circles rep-resent outliers Transparent grey points show all data for each binSample size for each bin is denoted by n

al (2018) which indicate that geomorphic context and vege-tation dynamics control OC concentration on floodplain soilsalong large lowland rivers in Alaska USA

At the reach or site scale wetter soil profiles consistentlyyield higher OC concentrations in all model groups How-ever moisture does not differ significantly between modelgroups (Fig 2c) indicating that this alone cannot explaindifferences between basins Soils tend to be finer in the wetbasin but clay content was not an important predictor of OCconcentration in studied soils Clay content likely influencesOC concentration based on previous research (Hoffmann etal 2009) although the inclusion of coarse soil material (in-cluding particulate organic matter) in our samples may ex-plain the lack of an observed correlation here Variabilityin soil redox conditions (ie the presence of anaerobic mi-crosites) may also introduce variability in respiration rate be-tween basins and among individual soil samples (Keiluweit

et al 2017) That is variability in groundwater hydrologymay result in a heterogeneous mix of aerobic (favoring mi-crobial activity) and anaerobic (suppressing microbial activ-ity) conditions within floodplain soils We did not record re-doximorphic features of our soil samples so we are unableto determine whether this influence contributes to the errorin our models Although confinement plays a strong role indetermining OC concentration in wet basin sites stratified byslope it does not differ significantly between basins (52 of semiarid basin reaches are unconfined compared to 63 of wet basin reaches) The major differences between thesebasins are their hydroclimatic and disturbance regimes Thewet basin is at a lower elevation and has denser and higherbiomass forests (Smithwick et al 2002) compared to thesparser parkland forests of the semiarid basin which likelyalso experiences more frequent fires based on fire historiesof nearby regions (recurrence interval on the order of 101ndash102 years Houston 1973 Loope and Gruell 1973) In addi-tion the volcanic soils present in the wet basin may suppresssoil OC respiration (Matus et al 2014) leading to a highersoil OC storage capacity there

Between basins it is likely that the hydroclimatic regimeinfluencing primary production plays some role in the wetbasinrsquos higher maximum OC concentrations in floodplainsoils compared to those of the semiarid basin Howeversmaller-scale factors such as soil texture and moisture alsolikely play a role and are not related to drainage area (Ta-ble 1) indicating that neither OC concentration nor its con-trolling factors vary continuously along a river network thussupporting H2a and H2b This also indicates that local fac-tors set largely by geomorphic and hydrologic dynamicsplay a significant role in modulating the effect of climate onOC concentrations If the wet and semiarid basins displayedsignificantly different OC concentrations our first hypothesisregarding the interbasin controls on OC concentration wouldbe supported However we instead found that climate andprimary productivity only partially determine OC concentra-tions especially when viewed in the context of geomorphicand hydrologic variability Thus the results do not supportH1

Each basin (or model group) is slightly different in termsof the controls on soil OC concentration moisture andtexture In the wet basin sites stratified by slope higher-elevation sites display higher OC concentrations This is con-trary to the general trend in primary productivity which de-creases with increasing elevation However it is importantto note that the headwaters of the wet basin are dominatedby lakes deltas and other depositional features in relativelybroad glacially carved valleys Subalpine lake deltas havebeen shown to store high OC concentrations in this basin(Scott and Wohl 2017) and many of the highest OC concen-trations we measured were located in broad wet meadowssubalpine lake deltas or other unconfined high-elevationreaches Such unconfined sites also likely have significantlycooler temperatures and tend to have higher soil moisture

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1110 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

contents as shown by our modeling (Table 1) As such al-though high-elevation wet basin sites may receive less OCinput they likely have a low rate of OC respiration resultingin higher OC concentrations on the whole which agrees withthe result of Bao et al (2017) In the semiarid basin our mod-eling suggests that the lower temperatures and higher mois-ture (Table 1) at higher elevations do not compensate for thelower primary productivity as elevation does not correlatewith OC concentration

In both basins unconfined reaches contain wetter andfiner-textured soils which may result in a higher soil OCcapacity Although confinement only relates directly to OCcontent in wet basin sites it does play a strong role in deter-mining moisture which in turn plays a role in regulating OCconcentration in both basins likely via inhibiting microbialactivity (Howard and Howard 1993) The relevance of chan-nel slope in determining soil moisture in the wet basin but notsemiarid basin may reflect the prevalence of high-gradientdebris-flow-dominated channels in the wet basin that largelyexhibit only gravel to boulder substrate which we assumestores minimal fine sediment moisture or OC

In the semiarid basin higher soil depths correlate withmoister and finer-textured soils but less OC concentrationThis indicates the trend in OC with depth likely dominatesthe signal of OC concentration with deeper sites containinga higher proportion of OC-depleted deep samples

42 Inferring sources of OC to floodplain soils (H3)

OC can be input to floodplain soils by two primary mech-anisms First dissolved and particulate OC (including largewood) can be deposited on floodplain surfaces by overbankdeposition thus integrating fluvial sedimentary OC into thefloodplain soil profile or in the case of large wood deposit-ing discrete but potentially large concentrations of OC thatcan later be integrated into the soil profile Second litter anddecomposing vegetation on the floodplain surface can act asautochthonous inputs of OC to floodplain soil

Our modeling of OC concentration yielded results consis-tent with previous investigations of controls on soil OC stor-age capacity (Jobbaacutegy and Jackson 2000 Sutfin and Wohl2017) Sites in the heterogeneous floodplain of the wet basindisplay a direct correlation between canopy cover and OCconcentration indicating that increased litter inputs lead toincreased floodplain soil OC concentration Sediment inputslikely differ between floodplain depositional unit types (egcoarser sediment may deposit on point bars compared tofilled secondary channels) although floodplain type does notpredict OC concentration This indicates that vegetation in-puts may be more dominant than fluvial sediment inputs atthese sites

The finding that buried OC peaks in the wet basin do nothave abnormally high clay contents supports the interpre-tation that wood and litter inputs to soil are the dominantsource of OC in the floodplain soils we examined Buried

peaks can be layers created by overbank deposition and sub-sequent burial of fine OC-bearing sediments (Blazejewski etal 2009 Ricker et al 2013) buried pieces of wood (Wohl2013) or buried organic horizons that are now capped bysediments that prevent OC respiration If overbank deposi-tion of fine sediment caused OC peaks we would expect tosee the soil samples classified as peaks exhibiting high claycontents indicating finer sediment Instead our results sug-gest that in the wet basin buried peaks are likely the result ofeither buried organic horizons or buried wood We observedlarge pieces of decaying buried wood in floodplain cut banksin the wet basin supporting this inference Overbank depo-sition of wood on the floodplain was only observed rarely inthis basin indicating that the OC measured in these soils islikely dominantly autochthonous

In the semiarid basin the two cores that exhibit peaks werecollected from the same meadow just downstream of a now-filled former lake that is a potential source of fine sedimentThe channels draining this meadow exhibit an anabranchingplanform indicating the potential to deposit and bury packetsof potentially OC-rich fine sediments However the majorityof cores do not exhibit OC peaks indicating OC input mainlyfrom vegetation at the surface and continuing OC respirationat depth

OC variation within each core is dominantly a function ofdepth We observe a negative correlation between depth be-low ground surface and OC concentration which has beenobserved in other studies including mountain wetlands andfloodplains (Jobbaacutegy and Jackson 2000 Scott and Wohl2017 Sutfin and Wohl 2017 Zhao et al 2017) In generalthis indicates that at least in mountain river floodplains OCis enriched at the surface and decomposes with depth sim-ilar to upland soils This fits with our finding that the ma-jority of our cores do not exhibit significant OC peaks atdepth and supports the dominance of litter and wood OC in-puts to floodplain soils These results support our hypothesisthat decaying litter and wood not overbank sediment depo-sition dominate the input of OC to floodplain soils in ourstudy basins (H3) We note that floodplain wood may alsoact as a trapping site for overbank fine organic matter facil-itating the deposition and input of OC from decaying vege-tation While these study basins likely accumulate soil OCmainly autochthonously other basins that experience over-bank flows accompanying deposition of fine sediment andburial of organic layers exhibit OC storage that is likely dom-inated by fluvial sediment deposition (eg Blazejewski etal 2009 DrsquoElia et al 2017 Ricker et al 2013) Thus itis likely that flow regime lateral connectivity and sedimenttransport dynamics regulate whether floodplain soil OC isdominantly input by the overbank deposition of fine mate-rial or litter and wood decay

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1111

Figure 4 Conceptual model of physical processes that influenceOC concentration in floodplain soils based on this study and otherliterature Each box corresponds to a major factor that influencesOC concentration Colored text within each box denotes factors thatinfluence OC inputs sorption capacity OC export or respirationrate (note that some processes influence both OC export and respi-ration rate) As sorption capacity increases so does the OC capacityof the soil Conversely as the rate of respiration andor OC exportincreases the soil OC capacity decreases Floodplain soils can onlydevelop high concentrations of OC if there are high rates of OCinput However the capacity of the soils to store OC regulates thatinput and is determined by the competing influences of sorption ca-pacity and the combination of respiration rate and OC export Seethe text for further details

43 Conceptual model of soil OC concentration infloodplain soils

We present a conceptual model to summarize our results andplace them in the context of work examining the controlson OC storage in soils (Fig 4) OC is input to floodplainseither through the decay of vegetation or the deposition offine OC-rich sediment This input of OC only determinesOC concentrations insofar as floodplain soils are capable ofstoring OC That storage is effectively determined by a bal-ance between processes that remove OC from floodplainsnamely respiration or erosion followed by respiration (Berheet al 2007) and processes that regulate OC availability tomicrobes namely the capability of the mineral fraction ofthe soil to sorb OC

OC sorption capacity reflects a few specific processes Al-though soil texture generally relates to the ability of OCto sorb to mineral grains and resulting OC availability soilchemistry also plays a strong and potentially dominant rolein regulating OC sorption capacity (Rasmussen et al 2018)Soil texture is largely determined by valley morphology ac-cording to our modeling (Table 1) placing valley morphol-ogy and resulting sediment transport dynamics (Gran andCzuba 2017 Wohl et al 2017b) as indirect controls on sorp-tion capacity

Respiration rate is largely determined by microbial activityand the availability of OC to microbes Erosion can rapidlyexpose soil OC to microbial respiration (Berhe et al 2007)whereas soils that reside in largely anoxic conditions can ex-hibit low rates of microbial respiration (Boye et al 2017)In addition soil mineralogy chemistry and redox condi-tions (Keiluweit et al 2017) can regulate microbial activity(eg andic soils may limit microbial respiration Matus et al2014) Our results suggesting that moisture controls OC con-tent support the idea that drier soils likely have higher ratesof microbial respiration of OC Moisture is a function of tex-ture valley bottom morphology and elevation (a proxy fortemperature) in our modeling (Table 1) Comparing flood-plain types in the wet basin we find that types with standingwater exhibit significantly higher soil moisture contents thanthose without standing water This indicates spatial variabil-ity in moisture content and likely microbial activity (Howardand Howard 1993) In our modeling this effect translates tospatial variability in OC concentration within floodplains andacross entire basins

OC export from soils refers specifically to leaching of dis-solved OC mainly from shallow soils While we do not di-rectly consider OC export in this study it likely introducesvariability in soil OC concentration and is regulated dom-inantly by hydrology (Aringgren et al 2014 Mcdowell andLikens 1988) soil pore water chemistry (including pH andionic strength Brooks et al 1999 Evans et al 2006) andredox conditions (Knorr 2013) OC export as DOC can actin conjunction with respiration to remove OC from soils andlike respiration rate is countered by sorption capacity

To summarize OC inputs regulated by the capacity ofsoils to store OC and suppress microbial respiration andOC export determine OC concentrations in floodplain soils(Fig 4) OC inputs to floodplain soils come from either au-tochthonous litter accumulation on the floodplain surface al-lochthonous wood deposition or allochthonous deposition offine OC-bearing sediments Our results from the mountain-ous basins we studied suggest that deposition of fine mate-rial in overbank flows is rare leading us to infer that au-tochthonous litter and allochthonous wood inputs to flood-plains dominate OC input in mountain rivers Where soils aremore moist microbial respiration is inhibited and more OCis stored Although soil texture is likely not a limiting factoron OC concentration in these floodplains finer-textured soilslikely have a higher sorption capacity retaining more of theOC input from decaying plant material Our results indicatethat geomorphic and hydrologic characteristics act as bound-ary conditions that regulate soil texture and moisture in turnregulating sorption capacity respiration rate and resultingOC concentrations in floodplain soils

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1112 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

5 Conclusions

We present floodplain soil OC concentration data from twodisparate watersheds to compare how interbasin variabilitybetween the two watersheds compares with intra-basin vari-ability in geomorphic and hydrologic characteristics in de-termining OC concentration Our results indicate that OCconcentration in mountain floodplain soils does not vary pre-dictably along a longitudinal gradient nor does it vary sub-stantially between basins with differing climatic and veg-etation characteristics Instead geomorphic and hydrologiccharacteristics such as valley bottom morphology and soilmoisture dominantly determine floodplain OC concentra-tion

In our study basins decaying litter and wood and not over-bank deposition of fine OC-bearing sediment is the mainsource of OC to floodplain soils It is unclear whether thatdecaying vegetation is dominated by autochthonous litter in-puts or transported downed wood In comparing our basinto other studied floodplain soils it seems that vegetation dy-namics play a strong role in determining OC concentrationswhen fine sediment is not regularly deposited on floodplainsurfaces However we suggest that floodplain soil character-istics set by geomorphic and hydrologic conditions regu-late how OC inputs translate to the spatial distribution of OCalong a river network

This implies that OC storage in floodplains likely cannotbe predicted using consistent downstream trends and thatmanagement prioritization designed to facilitate floodplainOC storage should be based on local geomorphic and hy-drologic process variability within each basin For instancemanagement to increase OC sequestration in floodplain soilswill likely be more effective where floodplains are uncon-fined and soils already experience high moisture conditionsfor much of the year Along these lines our results showthat modeling the floodplain biospheric OC pool to predictits response to warming and subsequent effects on climatebased on regional factors such as climate and net primaryproductivity likely misses the substantial interbasin variabil-ity in OC concentration and storage resulting from variabilityin valley bottom geometry and both geomorphic and hydro-logic processes (eg Doetterl et al 2015)

Although our results provide some insights the questionof whether OC stored in floodplain soil comes dominantlyfrom allochthonous versus autochthonous sources remainsopen Our results imply that more productive spatially het-erogeneous floodplains likely input more OC to soils Flood-plain OC concentration while mediated largely by moisturedynamics likely depends mainly on OC inputs from produc-tive riparian forests This implies that management of OCstorage in mountain river floodplains should focus on therestoration of riparian zones to maintain OC input to soil(eg Bullinger-Weber et al 2014) More detailed studiesin regions with varying sediment transport and hydrologicregimes are needed to determine what conditions favor au-

tochthonous versus allochthonous OC inputs but our resultssuggest that autochthonous sources dominate floodplain OCstorage in basins with relatively low rates of vertical accre-tion and high channelndashfloodplain connectivity that promotesfloodplain wetlands

Data availability Data to support the analyses presented here canbe found in Table S1 and in the Colorado State University DigitalRepository (Scott and Wohl 2018a)

Supplement The supplement related to this article is availableonline at httpsdoiorg105194esurf-6-1101-2018-supplement

Author contributions Field data collection was conducted pri-marily by DNS with assistance from EEW at some sites in theMiddle Fork Snoqualmie basin DNS performed data analysis anddrafted the initial version of the paper which was then edited andrewritten by DNS and EEW

Competing interests The authors declare that they have no con-flict of interest

Acknowledgements This work was funded by NSF grantEAR-1562713 We thank Ellen Daugherty for extensive assistancein fieldwork and discussion that improved the conceptual modelWe thank Katherine Lininger for stimulating discussion thatimproved the paper Detailed and constructive comments from twoanonymous reviewers and Robert Hilton improved the paper Wethank Sara Lowe for assistance in processing and analyzing soilsamples

Edited by Robert HiltonReviewed by two anonymous referees

References

Aringgren A M Buffam I Cooper D M Tiwari T Evans C Dand Laudon H Can the heterogeneity in stream dissolved or-ganic carbon be explained by contributing landscape elementsBiogeosciences 11 1199ndash1213 httpsdoiorg105194bg-11-1199-2014 2014

Appling A P Bernhardt E S and Stanford J A Flood-plain biogeochemical mosaics A multi-dimensional view ofalluvial soils J Geophys Res-Biogeo 119 1538ndash1553httpsdoiorg1010022013JG002543 2014

Aufdenkampe A K Mayorga E Raymond P A MelackJ M Doney S C Alin S R Aalto R E and YooK Riverine coupling of biogeochemical cycles between landoceans and atmosphere Front Ecol Environ 9 53ndash60httpsdoiorg101890100014 2011

Ballantyne A P Alden C B Miller J B Tans P P and WhiteJ W C Increase in observed net carbon dioxide uptake by

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1105

ziplock bags frozen within 72 h (most samples were frozenwithin 8 h) and kept frozen until analysis

24 Measuring soil OC and texture

To measure the concentration of organic carbon in soil sam-ples we used loss on ignition (LOI) We first defrosted sam-ples for 24ndash48 h at room temperature Once defrosted wethoroughly mixed samples to ensure the most homogenoussample possible We then subsampled 10ndash85 g of soil fromeach sample for analysis Using crucibles in a muffle furnacewe dried samples in batches of 30 for 24 h at 105 C to deter-mine moisture content and remove all nonstructurally heldwater Following the guidelines suggested by Hoogsteen etal (2015) we then burned samples for 3 h at 550 C to re-move organic matter By comparing the weight of the burnedsamples with that of the dried samples we obtained an LOIweight

After performing LOI we used burned samples to performtexture by feel to determine the USDA soil texture class andestimated clay content (Thien 1979) To convert LOI weightto OC concentration we used the structural water loss cor-rection of Hoogsteen et al (2015) using clay content esti-mated from soil texture This correction considers water heldby clay that may not evaporate during drying but will evap-orate during burning It also estimates the proportion of theLOI weight that is OC This correction represented as a per-cent of the estimated sample OC content after the correctionranged from 090 to 49576 (95 confidence interval ofthe median between 1617 and 2425 ) for the wet basinand ranged from 554 to 57053 (95 confidence in-terval of the median between 1976 and 3211 ) for thesemiarid basin

One potential confounding factor in LOI is carbonates thatmay burn off during ignition adding to the LOI weight whilenot being organic matter In lithologies where carbonates arerare (eg granitoid rocks like those found in the upper partof the wet basin and entire semiarid basins) this is a rela-tively negligible issue However some of our soil samplescame from parts of the wet basin draining rocks of the west-ern meacutelange belt including argillite greywacke and marbleWe tested samples for the presence of carbonates to deter-mine whether our LOI methods would be sufficient to accu-rately determine OC concentration We randomly chose 10soil samples of a total of 110 drained rocks that could includecarbonates and submitted them to the Colorado State Univer-sity soil testing laboratory for CHN furnace analysis (Sparks1996) which yielded data on the proportion of carbonatesby mass in those samples The median calcium carbonateconcentrations of those 10 samples was 098 (95 con-fidence interval between 070 and 103 ) and the medianpercentage of the total carbon in samples comprised of inor-ganic carbon was 505 (95 confidence interval between197 and 189 ) From this we concluded that the amountof carbonate in the samples draining potentially carbonate-

bearing rocks was low enough that LOI was likely to still beaccurate Consequently we analyzed all soil samples usingLOI to obtain OC concentration The median difference be-tween the LOI and CHN OC estimates for these 10 sampleswas 045 (95 confidence interval between minus083 and147 ) indicating no systematic bias in LOI estimates ofOC concentration

25 GIS and derivative measurements

After fieldwork in each basin we collected the following datafor each reach using a GIS platform elevation drainage arealand cover classification and canopy cover from the NationalLand Cover Database (Homer et al 2015) and the meanslope of the basin draining to each reach (including hill-slopes and channels) Utilizing drainage area at each reachand field-measured channel gradient we calculated an esti-mated stream power as the product of drainage area chan-nel gradient and basin-averaged precipitation We utilizeda 10 m DEM for all GIS topographic measurements To es-timate clay content for each sample we used median val-ues for assigned USDA texture classes To obtain estimatedclay content moisture and OC for each core we calculatedan average weighted by the percentage of core taken up byeach soil sample We categorized wet basin samples stratifiedby floodplain type into those with standing water (wetlandsand oxbow lakes) and those with no standing water (all othertypes)

26 Statistical analyses

All statistical analyses were performed using the R statisti-cal computing software (R Core Team 2017) We conductedall analyses on three modeling groups based on the variablesmeasured in each group In the wet basin we grouped obser-vations by stratification type separating observations strati-fied by channel slope from observations stratified by flood-plain type We separated these two groups from all obser-vations in the semiarid basin which were measured consis-tently We modeled OC concentration and soil texture with amixed-effects linear regression using individual soil samples(ie the individual samples that make up a core) as sam-ple units (n= 103 for wet basin stratified by slope 89 forwet basin stratified by floodplain type and 101 for semiaridbasin) We modeled the sampled site as a random effect ac-knowledging that individual soil samples within a single coreare likely nonindependent We used profiled 95 confidenceintervals on effect estimates (β) for fixed effects to evaluatevariable importance in mixed-effects models

To gain further insight at the reach scale we also mod-eled average OC concentration and soil moisture at eachmeasured site using multiple linear regression We mod-eled soil moisture at the reach scale because we felt thatour single snapshot of moisture conditions was better rep-resented as a site-level average We first performed uni-

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1106 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

variate analysis between each hypothesized predictor andthe response utilizing mainly comparative Wilcoxon rank-sum tests (Wilcoxon 1945) or correlational Spearman cor-relation coefficient statistics We utilized a Holm multiple-comparison correction (Holm 1979) for pairwise compar-isons During this filtering we also viewed box plots or scat-terplots as appropriate to discern which variables appear tohave anything other than a completely random relationshipwith the response We then utilized the multiple linear re-gression of all subsets using the corrected Akaike informa-tion criterion as a model selection criterion (Wagenmakersand Farrell 2004) We iteratively transformed response vari-ables to ensure the homoscedasticity of error terms To selecta single best model we utilized both Akaike-weight-basedimportance and parsimony to select a final reduced modelWe considered sample sizes p values and effect magnitudesin determining variable importance

We also analyzed each core to determine whether therewere buried high-OC-concentration layers at depth We com-pared each buried soil sample to the sample above it usingthe criterion that a peak in OC at depth should have an OCconcentration 15 times that of the overlying sample and beabove 05 (Appling et al 2014)

3 Results

Model results are presented in Table 1 Comparisons betweenbasins and summaries of OC concentration moisture andestimated clay content are shown in Fig 2

31 OC concentration

Most cores display a decrease in OC concentration withdepth (Fig 3) Of cores with more than a single sample32 (722) of cores stratified by slope in the wet basin 32 (825) of cores stratified by floodplain type in the wet basinand 6 (231) of cores in the semiarid basin exhibit OC con-centration peaks at depth Whether a soil sample was classi-fied as an OC peak has no relation to estimated clay contentin sites stratified by floodplain type (p = 028) or those strat-ified by slope (p = 089) in the wet basin In the semiaridbasin soil samples classified as buried OC peaks have signif-icantly higher estimated clay contents (p = 005) than thosethat were not classified as peaks

In general the floodplain-stratified sites in the wet basinstore higher densities of OC than the semiarid basin (Fig 2ab) Figure 2a includes zero values (ie sites with no OC-bearing sediment only present in the wet basin slope-stratified group) whereas Fig 2b does not because sampleunits in Fig 2b are individual soil samples Comparing thesetwo groups it appears that soils in the wet basin exhibit muchhigher OC concentrations than those in the semiarid basinbut in general there are many more reaches with no fine sed-iment available to store OC in the wet basin

At the scale of individual soil samples the depth belowground surface is by far the dominant control on OC con-centration across all modeling groups We used a cube roottransform for all three mixed-effects models of OC concen-tration For wet basin sites stratified by slope deeper soilsamples contain less OC (β =minus00084plusmn 00042) whereassoil samples at higher elevations tend to contain more OC(β = 00010plusmn000099) Depth is the only significant predic-tor of OC content for both wet basin soil samples stratifiedby floodplain type (β =minus00084plusmn00042) and soil samplesin the semiarid basin (β =minus00037plusmn 00019)

Modeling wet basin slope-stratified sites at the reach scalewe found that moisture (β = 00078plusmn 00031) and whetherthe reach was unconfined (β = 077plusmn 049) control soil OC(cube root transformation model adjusted R2

= 054 p lt00001) Modeling wet basin floodplain-stratified sites at thesite scale we found that canopy cover (β = 0012plusmn 0011)and moisture (β = 00040plusmn 00011) are controls on soilOC (cube root transformation model adjusted R2

= 067p lt 00001) Modeling Big Sandy sites at the reach scalewe found that soil depth (β =minus0012plusmn 00071) and mois-ture (β = 0014plusmn 00026) are dominant controls on soil OCconcentration (no transformation model adjustedR2

= 069p lt 00001)

In general moister deeper soils store more OC at the reachscale whereas OC tends to vary dominantly with depth atthe scale of individual soil samples Although estimated claycontent did not emerge as a significant predictor of OC con-centration it is used to calculate clay-held water to correctour LOI-based OC concentration measurements making itimportant in determining OC for each sample

32 Soil texture

In general soil texture follows a predictable trend withriver size between model groups (Fig 2d) Floodplain type-stratified sites in the wet basin store the most clay followedby slope-stratified sites and then sites in the semiarid basin

Modeling soil texture at the individual soil sample scaleacross slope-stratified sites in the wet basin we foundwhether the reach was confined (β = 542plusmn 532) andwhether the bed material was dominantly sand (β = 1047plusmn613) to be dominant controls on estimated clay contentModeling soil texture for sites stratified by floodplain typein the wet basin yielded no significant trends In the semi-arid basin we found that either valley width (β = 00050plusmn00032) or whether the stream was unconfined (β = 041plusmn033) as well as depth below ground surface (β =minus00069plusmn00033 for model with valley width but not confinement) sig-nificantly control soil texture

To summarize sites from unconfined lower-energyreaches in the wet basin and sites from reaches with widervalley bottoms and at lower depths in the semiarid basin ex-hibited more finely textured soils

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1107

Tabl

e1

Mat

rix

ofal

lmod

els

pres

ente

din

the

text

Eac

hm

odel

islis

ted

bym

odel

grou

pre

spon

seva

riab

lea

ndsc

ale

Scal

ere

fers

toth

esa

mpl

eun

itof

the

mod

ela

ndsi

tere

fers

toa

core

with

the

resp

onse

aver

aged

over

allt

hein

divi

dual

soil

sam

ples

inth

eco

reF

orea

chva

riab

lean

dm

odel

an

aste

risk

()i

ndic

ates

that

the

vari

able

was

incl

uded

inm

odel

sele

ctio

nbu

tnot

the

full

mix

ed-e

ffec

tsm

odel

Am

inus

(minus)

indi

cate

sth

atth

eva

riab

lew

asse

lect

edas

impo

rtan

tin

pred

ictin

gth

ere

spon

sean

dde

note

san

indi

rect

corr

elat

ion

whe

reas

apl

us(+

)ind

icat

esa

dire

ctco

rrel

atio

nIn

the

case

ofco

nfine

men

ta

plus

indi

cate

sth

atun

confi

ned

stre

ams

disp

lay

ahi

gher

-mag

nitu

dere

spon

seva

riab

leI

nth

eca

seof

bed

mat

eria

la

plus

indi

cate

sth

atsa

mpl

esw

ithsa

ndex

hibi

tahi

gher

valu

eof

the

resp

onse

NA

indi

cate

sth

atei

ther

the

vari

able

was

notm

easu

red

for

that

basi

nor

mod

elgr

oup

orth

atit

isth

em

odel

resp

onse

Mod

elgr

oup

Res

pons

eSc

ale

(sam

ple

unit)

Var

iabl

esConfinement

Bedform

Channelslope

Bedmaterial

Multithread

Valleywidth

Bankfullwidth

Bankfulldepth

Streampower

Floodplaintype

Standingwater

Depth1

Claycontent

Moisture

Loggingnearby

Grassespresent

Shrubspresent

Treespresent

Elevation

Basinslope

Canopycover

NLCD

Drainagearea

Wet

basi

nst

rat-

ified

bysl

ope

OC

()

Soil

sam

ple

NA

NAminus

+

O

C(

)Si

te+

NA

NA

+

Moi

stur

e(

)Si

te+

minus

NA

NA

N

A+

Te

xtur

e(

)So

ilsa

mpl

e+

+

N

AN

A

NA

Wet

basi

nst

rat-

ified

byflo

od-

plai

nty

pe

OC

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

minus

N

AO

C(

)Si

teN

AN

AN

AN

AN

AN

AN

AN

AN

A

+

+N

AM

oist

ure

()

Site

NA

NA

NA

NA

NA

NA

NA

NA

NA

+

+

NA

NA

Text

ure2

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

N

AN

A

Sem

iari

dba

sin

OC

()

Soil

sam

ple

N

A

NA

NA

N

AN

A

NA

NAminus

NA

NA

NA

NA

OC

()

Site

N

A

NA

NA

NA

NA

N

AN

Aminus

+N

AN

AN

AN

A

M

oist

ure

()

Site

+N

A

NA

NA

NA

NA

N

AN

A+

NA

NA

NA

NA

NA+

Te

xtur

e3(

)So

ilsa

mpl

e+

NA

N

AN

A+

NA

NA

N

AN

Aminus

NA

NA

NA

NA

NA

1D

epth

refe

rsto

eith

erth

eso

ilsa

mpl

ede

pth

belo

wth

egr

ound

orth

eto

tald

epth

ofth

eco

red

epen

ding

onth

esa

mpl

eun

it2

No

sign

ifica

ntre

sults

wer

eob

serv

edfo

rthi

sm

odel

3Fo

rthi

sm

odel

bot

hva

lley

wid

than

dco

nfine

men

tpre

dict

text

ure

and

can

bein

terp

rete

din

terc

hang

eabl

yH

owev

eri

nclu

ding

both

inth

esa

me

mod

elw

ould

yiel

dpr

oble

ms

due

tom

ultic

ollin

eari

ty

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1108 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

Figure 2 Box plots showing comparisons between model groups of OC concentration at the reach scale (a) OC concentration at the scaleof individual soil samples (b) moisture at the reach scale (c) and estimated clay content at the scale of individual soil samples (d) Endsof dotted lines represent 15 times the interquartile range which is represented by boxes Bold line represents median Circles representoutliers Letters indicate probable differences between groups based on pairwise Wilcoxon (andashc) or t tests (d) with a Holm correctionRanges in parentheses below letters show the 95 confidence interval of the median value for the group (andashc) or the mean value for thegroup (d) in which median confidence intervals were overly constrained due to the categorical nature of our estimated clay content data

33 Soil moisture

Soil moisture is less variable between basins than either tex-ture or OC concentration (Fig 2c) All model groups exhibitsimilar soil moisture conditions although there was signifi-cant variability within each model group

Soil moisture at wet basin sites stratified by slope is domi-nantly controlled by channel slope (β =minus1315plusmn911) ele-vation (β = 00046plusmn 00038) and whether the stream is un-confined (β = 389plusmn 267 model adjusted R2

= 038 p lt00001) At wet basin sites stratified by floodplain type es-timated clay content (β = 0060plusmn 0042) and whether thefloodplain unit had standing water (β = 115plusmn 071) sig-nificantly control soil moisture (model adjusted R2

= 046p lt 00001) In the semiarid basin soil depth (β = 0012plusmn0012) elevation (β = 00023plusmn 00014) and whether thereach was unconfined (β = 091plusmn075) significantly controlsoil moisture (model adjusted R2

= 035 p lt 00001)

4 Discussion

41 Understanding spatial variability in OCconcentration in floodplain soils (H1 and H2)

Comparing the wet basin to the semiarid basin shows thatthe wetter higher primary productivity basin is capable ofstoring greater concentrations of OC in floodplain soils butthat both regions generally store similar OC concentrations infloodplain soils This result partially agrees with the exami-nation of subalpine lake deltas by Scott and Wohl (2017) Inthat study subalpine lake deltas in the wet basin were com-pared to deltas in the drier Colorado Front Range Subalpinelake deltas displayed similar OC concentrations likely dueto competing but complementary OC stabilization and lossmechanisms in each region Those deltas represent a subsetof the broader valley bottom soils studied here This moreexpansive study points to both geomorphic controls such asvalley bottom geometry and factors influenced by climatesuch as canopy cover as controls on OC storage in valley bot-toms These results also agree with the results of Lininger et

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1109

Figure 3 Box plots of sample OC concentration binned by meansample depth for the wet (a) and semiarid (b) basins Ends of dot-ted lines represent 15 times the interquartile range which is rep-resented by boxes Bold line represents median Black circles rep-resent outliers Transparent grey points show all data for each binSample size for each bin is denoted by n

al (2018) which indicate that geomorphic context and vege-tation dynamics control OC concentration on floodplain soilsalong large lowland rivers in Alaska USA

At the reach or site scale wetter soil profiles consistentlyyield higher OC concentrations in all model groups How-ever moisture does not differ significantly between modelgroups (Fig 2c) indicating that this alone cannot explaindifferences between basins Soils tend to be finer in the wetbasin but clay content was not an important predictor of OCconcentration in studied soils Clay content likely influencesOC concentration based on previous research (Hoffmann etal 2009) although the inclusion of coarse soil material (in-cluding particulate organic matter) in our samples may ex-plain the lack of an observed correlation here Variabilityin soil redox conditions (ie the presence of anaerobic mi-crosites) may also introduce variability in respiration rate be-tween basins and among individual soil samples (Keiluweit

et al 2017) That is variability in groundwater hydrologymay result in a heterogeneous mix of aerobic (favoring mi-crobial activity) and anaerobic (suppressing microbial activ-ity) conditions within floodplain soils We did not record re-doximorphic features of our soil samples so we are unableto determine whether this influence contributes to the errorin our models Although confinement plays a strong role indetermining OC concentration in wet basin sites stratified byslope it does not differ significantly between basins (52 of semiarid basin reaches are unconfined compared to 63 of wet basin reaches) The major differences between thesebasins are their hydroclimatic and disturbance regimes Thewet basin is at a lower elevation and has denser and higherbiomass forests (Smithwick et al 2002) compared to thesparser parkland forests of the semiarid basin which likelyalso experiences more frequent fires based on fire historiesof nearby regions (recurrence interval on the order of 101ndash102 years Houston 1973 Loope and Gruell 1973) In addi-tion the volcanic soils present in the wet basin may suppresssoil OC respiration (Matus et al 2014) leading to a highersoil OC storage capacity there

Between basins it is likely that the hydroclimatic regimeinfluencing primary production plays some role in the wetbasinrsquos higher maximum OC concentrations in floodplainsoils compared to those of the semiarid basin Howeversmaller-scale factors such as soil texture and moisture alsolikely play a role and are not related to drainage area (Ta-ble 1) indicating that neither OC concentration nor its con-trolling factors vary continuously along a river network thussupporting H2a and H2b This also indicates that local fac-tors set largely by geomorphic and hydrologic dynamicsplay a significant role in modulating the effect of climate onOC concentrations If the wet and semiarid basins displayedsignificantly different OC concentrations our first hypothesisregarding the interbasin controls on OC concentration wouldbe supported However we instead found that climate andprimary productivity only partially determine OC concentra-tions especially when viewed in the context of geomorphicand hydrologic variability Thus the results do not supportH1

Each basin (or model group) is slightly different in termsof the controls on soil OC concentration moisture andtexture In the wet basin sites stratified by slope higher-elevation sites display higher OC concentrations This is con-trary to the general trend in primary productivity which de-creases with increasing elevation However it is importantto note that the headwaters of the wet basin are dominatedby lakes deltas and other depositional features in relativelybroad glacially carved valleys Subalpine lake deltas havebeen shown to store high OC concentrations in this basin(Scott and Wohl 2017) and many of the highest OC concen-trations we measured were located in broad wet meadowssubalpine lake deltas or other unconfined high-elevationreaches Such unconfined sites also likely have significantlycooler temperatures and tend to have higher soil moisture

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1110 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

contents as shown by our modeling (Table 1) As such al-though high-elevation wet basin sites may receive less OCinput they likely have a low rate of OC respiration resultingin higher OC concentrations on the whole which agrees withthe result of Bao et al (2017) In the semiarid basin our mod-eling suggests that the lower temperatures and higher mois-ture (Table 1) at higher elevations do not compensate for thelower primary productivity as elevation does not correlatewith OC concentration

In both basins unconfined reaches contain wetter andfiner-textured soils which may result in a higher soil OCcapacity Although confinement only relates directly to OCcontent in wet basin sites it does play a strong role in deter-mining moisture which in turn plays a role in regulating OCconcentration in both basins likely via inhibiting microbialactivity (Howard and Howard 1993) The relevance of chan-nel slope in determining soil moisture in the wet basin but notsemiarid basin may reflect the prevalence of high-gradientdebris-flow-dominated channels in the wet basin that largelyexhibit only gravel to boulder substrate which we assumestores minimal fine sediment moisture or OC

In the semiarid basin higher soil depths correlate withmoister and finer-textured soils but less OC concentrationThis indicates the trend in OC with depth likely dominatesthe signal of OC concentration with deeper sites containinga higher proportion of OC-depleted deep samples

42 Inferring sources of OC to floodplain soils (H3)

OC can be input to floodplain soils by two primary mech-anisms First dissolved and particulate OC (including largewood) can be deposited on floodplain surfaces by overbankdeposition thus integrating fluvial sedimentary OC into thefloodplain soil profile or in the case of large wood deposit-ing discrete but potentially large concentrations of OC thatcan later be integrated into the soil profile Second litter anddecomposing vegetation on the floodplain surface can act asautochthonous inputs of OC to floodplain soil

Our modeling of OC concentration yielded results consis-tent with previous investigations of controls on soil OC stor-age capacity (Jobbaacutegy and Jackson 2000 Sutfin and Wohl2017) Sites in the heterogeneous floodplain of the wet basindisplay a direct correlation between canopy cover and OCconcentration indicating that increased litter inputs lead toincreased floodplain soil OC concentration Sediment inputslikely differ between floodplain depositional unit types (egcoarser sediment may deposit on point bars compared tofilled secondary channels) although floodplain type does notpredict OC concentration This indicates that vegetation in-puts may be more dominant than fluvial sediment inputs atthese sites

The finding that buried OC peaks in the wet basin do nothave abnormally high clay contents supports the interpre-tation that wood and litter inputs to soil are the dominantsource of OC in the floodplain soils we examined Buried

peaks can be layers created by overbank deposition and sub-sequent burial of fine OC-bearing sediments (Blazejewski etal 2009 Ricker et al 2013) buried pieces of wood (Wohl2013) or buried organic horizons that are now capped bysediments that prevent OC respiration If overbank deposi-tion of fine sediment caused OC peaks we would expect tosee the soil samples classified as peaks exhibiting high claycontents indicating finer sediment Instead our results sug-gest that in the wet basin buried peaks are likely the result ofeither buried organic horizons or buried wood We observedlarge pieces of decaying buried wood in floodplain cut banksin the wet basin supporting this inference Overbank depo-sition of wood on the floodplain was only observed rarely inthis basin indicating that the OC measured in these soils islikely dominantly autochthonous

In the semiarid basin the two cores that exhibit peaks werecollected from the same meadow just downstream of a now-filled former lake that is a potential source of fine sedimentThe channels draining this meadow exhibit an anabranchingplanform indicating the potential to deposit and bury packetsof potentially OC-rich fine sediments However the majorityof cores do not exhibit OC peaks indicating OC input mainlyfrom vegetation at the surface and continuing OC respirationat depth

OC variation within each core is dominantly a function ofdepth We observe a negative correlation between depth be-low ground surface and OC concentration which has beenobserved in other studies including mountain wetlands andfloodplains (Jobbaacutegy and Jackson 2000 Scott and Wohl2017 Sutfin and Wohl 2017 Zhao et al 2017) In generalthis indicates that at least in mountain river floodplains OCis enriched at the surface and decomposes with depth sim-ilar to upland soils This fits with our finding that the ma-jority of our cores do not exhibit significant OC peaks atdepth and supports the dominance of litter and wood OC in-puts to floodplain soils These results support our hypothesisthat decaying litter and wood not overbank sediment depo-sition dominate the input of OC to floodplain soils in ourstudy basins (H3) We note that floodplain wood may alsoact as a trapping site for overbank fine organic matter facil-itating the deposition and input of OC from decaying vege-tation While these study basins likely accumulate soil OCmainly autochthonously other basins that experience over-bank flows accompanying deposition of fine sediment andburial of organic layers exhibit OC storage that is likely dom-inated by fluvial sediment deposition (eg Blazejewski etal 2009 DrsquoElia et al 2017 Ricker et al 2013) Thus itis likely that flow regime lateral connectivity and sedimenttransport dynamics regulate whether floodplain soil OC isdominantly input by the overbank deposition of fine mate-rial or litter and wood decay

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1111

Figure 4 Conceptual model of physical processes that influenceOC concentration in floodplain soils based on this study and otherliterature Each box corresponds to a major factor that influencesOC concentration Colored text within each box denotes factors thatinfluence OC inputs sorption capacity OC export or respirationrate (note that some processes influence both OC export and respi-ration rate) As sorption capacity increases so does the OC capacityof the soil Conversely as the rate of respiration andor OC exportincreases the soil OC capacity decreases Floodplain soils can onlydevelop high concentrations of OC if there are high rates of OCinput However the capacity of the soils to store OC regulates thatinput and is determined by the competing influences of sorption ca-pacity and the combination of respiration rate and OC export Seethe text for further details

43 Conceptual model of soil OC concentration infloodplain soils

We present a conceptual model to summarize our results andplace them in the context of work examining the controlson OC storage in soils (Fig 4) OC is input to floodplainseither through the decay of vegetation or the deposition offine OC-rich sediment This input of OC only determinesOC concentrations insofar as floodplain soils are capable ofstoring OC That storage is effectively determined by a bal-ance between processes that remove OC from floodplainsnamely respiration or erosion followed by respiration (Berheet al 2007) and processes that regulate OC availability tomicrobes namely the capability of the mineral fraction ofthe soil to sorb OC

OC sorption capacity reflects a few specific processes Al-though soil texture generally relates to the ability of OCto sorb to mineral grains and resulting OC availability soilchemistry also plays a strong and potentially dominant rolein regulating OC sorption capacity (Rasmussen et al 2018)Soil texture is largely determined by valley morphology ac-cording to our modeling (Table 1) placing valley morphol-ogy and resulting sediment transport dynamics (Gran andCzuba 2017 Wohl et al 2017b) as indirect controls on sorp-tion capacity

Respiration rate is largely determined by microbial activityand the availability of OC to microbes Erosion can rapidlyexpose soil OC to microbial respiration (Berhe et al 2007)whereas soils that reside in largely anoxic conditions can ex-hibit low rates of microbial respiration (Boye et al 2017)In addition soil mineralogy chemistry and redox condi-tions (Keiluweit et al 2017) can regulate microbial activity(eg andic soils may limit microbial respiration Matus et al2014) Our results suggesting that moisture controls OC con-tent support the idea that drier soils likely have higher ratesof microbial respiration of OC Moisture is a function of tex-ture valley bottom morphology and elevation (a proxy fortemperature) in our modeling (Table 1) Comparing flood-plain types in the wet basin we find that types with standingwater exhibit significantly higher soil moisture contents thanthose without standing water This indicates spatial variabil-ity in moisture content and likely microbial activity (Howardand Howard 1993) In our modeling this effect translates tospatial variability in OC concentration within floodplains andacross entire basins

OC export from soils refers specifically to leaching of dis-solved OC mainly from shallow soils While we do not di-rectly consider OC export in this study it likely introducesvariability in soil OC concentration and is regulated dom-inantly by hydrology (Aringgren et al 2014 Mcdowell andLikens 1988) soil pore water chemistry (including pH andionic strength Brooks et al 1999 Evans et al 2006) andredox conditions (Knorr 2013) OC export as DOC can actin conjunction with respiration to remove OC from soils andlike respiration rate is countered by sorption capacity

To summarize OC inputs regulated by the capacity ofsoils to store OC and suppress microbial respiration andOC export determine OC concentrations in floodplain soils(Fig 4) OC inputs to floodplain soils come from either au-tochthonous litter accumulation on the floodplain surface al-lochthonous wood deposition or allochthonous deposition offine OC-bearing sediments Our results from the mountain-ous basins we studied suggest that deposition of fine mate-rial in overbank flows is rare leading us to infer that au-tochthonous litter and allochthonous wood inputs to flood-plains dominate OC input in mountain rivers Where soils aremore moist microbial respiration is inhibited and more OCis stored Although soil texture is likely not a limiting factoron OC concentration in these floodplains finer-textured soilslikely have a higher sorption capacity retaining more of theOC input from decaying plant material Our results indicatethat geomorphic and hydrologic characteristics act as bound-ary conditions that regulate soil texture and moisture in turnregulating sorption capacity respiration rate and resultingOC concentrations in floodplain soils

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1112 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

5 Conclusions

We present floodplain soil OC concentration data from twodisparate watersheds to compare how interbasin variabilitybetween the two watersheds compares with intra-basin vari-ability in geomorphic and hydrologic characteristics in de-termining OC concentration Our results indicate that OCconcentration in mountain floodplain soils does not vary pre-dictably along a longitudinal gradient nor does it vary sub-stantially between basins with differing climatic and veg-etation characteristics Instead geomorphic and hydrologiccharacteristics such as valley bottom morphology and soilmoisture dominantly determine floodplain OC concentra-tion

In our study basins decaying litter and wood and not over-bank deposition of fine OC-bearing sediment is the mainsource of OC to floodplain soils It is unclear whether thatdecaying vegetation is dominated by autochthonous litter in-puts or transported downed wood In comparing our basinto other studied floodplain soils it seems that vegetation dy-namics play a strong role in determining OC concentrationswhen fine sediment is not regularly deposited on floodplainsurfaces However we suggest that floodplain soil character-istics set by geomorphic and hydrologic conditions regu-late how OC inputs translate to the spatial distribution of OCalong a river network

This implies that OC storage in floodplains likely cannotbe predicted using consistent downstream trends and thatmanagement prioritization designed to facilitate floodplainOC storage should be based on local geomorphic and hy-drologic process variability within each basin For instancemanagement to increase OC sequestration in floodplain soilswill likely be more effective where floodplains are uncon-fined and soils already experience high moisture conditionsfor much of the year Along these lines our results showthat modeling the floodplain biospheric OC pool to predictits response to warming and subsequent effects on climatebased on regional factors such as climate and net primaryproductivity likely misses the substantial interbasin variabil-ity in OC concentration and storage resulting from variabilityin valley bottom geometry and both geomorphic and hydro-logic processes (eg Doetterl et al 2015)

Although our results provide some insights the questionof whether OC stored in floodplain soil comes dominantlyfrom allochthonous versus autochthonous sources remainsopen Our results imply that more productive spatially het-erogeneous floodplains likely input more OC to soils Flood-plain OC concentration while mediated largely by moisturedynamics likely depends mainly on OC inputs from produc-tive riparian forests This implies that management of OCstorage in mountain river floodplains should focus on therestoration of riparian zones to maintain OC input to soil(eg Bullinger-Weber et al 2014) More detailed studiesin regions with varying sediment transport and hydrologicregimes are needed to determine what conditions favor au-

tochthonous versus allochthonous OC inputs but our resultssuggest that autochthonous sources dominate floodplain OCstorage in basins with relatively low rates of vertical accre-tion and high channelndashfloodplain connectivity that promotesfloodplain wetlands

Data availability Data to support the analyses presented here canbe found in Table S1 and in the Colorado State University DigitalRepository (Scott and Wohl 2018a)

Supplement The supplement related to this article is availableonline at httpsdoiorg105194esurf-6-1101-2018-supplement

Author contributions Field data collection was conducted pri-marily by DNS with assistance from EEW at some sites in theMiddle Fork Snoqualmie basin DNS performed data analysis anddrafted the initial version of the paper which was then edited andrewritten by DNS and EEW

Competing interests The authors declare that they have no con-flict of interest

Acknowledgements This work was funded by NSF grantEAR-1562713 We thank Ellen Daugherty for extensive assistancein fieldwork and discussion that improved the conceptual modelWe thank Katherine Lininger for stimulating discussion thatimproved the paper Detailed and constructive comments from twoanonymous reviewers and Robert Hilton improved the paper Wethank Sara Lowe for assistance in processing and analyzing soilsamples

Edited by Robert HiltonReviewed by two anonymous referees

References

Aringgren A M Buffam I Cooper D M Tiwari T Evans C Dand Laudon H Can the heterogeneity in stream dissolved or-ganic carbon be explained by contributing landscape elementsBiogeosciences 11 1199ndash1213 httpsdoiorg105194bg-11-1199-2014 2014

Appling A P Bernhardt E S and Stanford J A Flood-plain biogeochemical mosaics A multi-dimensional view ofalluvial soils J Geophys Res-Biogeo 119 1538ndash1553httpsdoiorg1010022013JG002543 2014

Aufdenkampe A K Mayorga E Raymond P A MelackJ M Doney S C Alin S R Aalto R E and YooK Riverine coupling of biogeochemical cycles between landoceans and atmosphere Front Ecol Environ 9 53ndash60httpsdoiorg101890100014 2011

Ballantyne A P Alden C B Miller J B Tans P P and WhiteJ W C Increase in observed net carbon dioxide uptake by

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

1106 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

variate analysis between each hypothesized predictor andthe response utilizing mainly comparative Wilcoxon rank-sum tests (Wilcoxon 1945) or correlational Spearman cor-relation coefficient statistics We utilized a Holm multiple-comparison correction (Holm 1979) for pairwise compar-isons During this filtering we also viewed box plots or scat-terplots as appropriate to discern which variables appear tohave anything other than a completely random relationshipwith the response We then utilized the multiple linear re-gression of all subsets using the corrected Akaike informa-tion criterion as a model selection criterion (Wagenmakersand Farrell 2004) We iteratively transformed response vari-ables to ensure the homoscedasticity of error terms To selecta single best model we utilized both Akaike-weight-basedimportance and parsimony to select a final reduced modelWe considered sample sizes p values and effect magnitudesin determining variable importance

We also analyzed each core to determine whether therewere buried high-OC-concentration layers at depth We com-pared each buried soil sample to the sample above it usingthe criterion that a peak in OC at depth should have an OCconcentration 15 times that of the overlying sample and beabove 05 (Appling et al 2014)

3 Results

Model results are presented in Table 1 Comparisons betweenbasins and summaries of OC concentration moisture andestimated clay content are shown in Fig 2

31 OC concentration

Most cores display a decrease in OC concentration withdepth (Fig 3) Of cores with more than a single sample32 (722) of cores stratified by slope in the wet basin 32 (825) of cores stratified by floodplain type in the wet basinand 6 (231) of cores in the semiarid basin exhibit OC con-centration peaks at depth Whether a soil sample was classi-fied as an OC peak has no relation to estimated clay contentin sites stratified by floodplain type (p = 028) or those strat-ified by slope (p = 089) in the wet basin In the semiaridbasin soil samples classified as buried OC peaks have signif-icantly higher estimated clay contents (p = 005) than thosethat were not classified as peaks

In general the floodplain-stratified sites in the wet basinstore higher densities of OC than the semiarid basin (Fig 2ab) Figure 2a includes zero values (ie sites with no OC-bearing sediment only present in the wet basin slope-stratified group) whereas Fig 2b does not because sampleunits in Fig 2b are individual soil samples Comparing thesetwo groups it appears that soils in the wet basin exhibit muchhigher OC concentrations than those in the semiarid basinbut in general there are many more reaches with no fine sed-iment available to store OC in the wet basin

At the scale of individual soil samples the depth belowground surface is by far the dominant control on OC con-centration across all modeling groups We used a cube roottransform for all three mixed-effects models of OC concen-tration For wet basin sites stratified by slope deeper soilsamples contain less OC (β =minus00084plusmn 00042) whereassoil samples at higher elevations tend to contain more OC(β = 00010plusmn000099) Depth is the only significant predic-tor of OC content for both wet basin soil samples stratifiedby floodplain type (β =minus00084plusmn00042) and soil samplesin the semiarid basin (β =minus00037plusmn 00019)

Modeling wet basin slope-stratified sites at the reach scalewe found that moisture (β = 00078plusmn 00031) and whetherthe reach was unconfined (β = 077plusmn 049) control soil OC(cube root transformation model adjusted R2

= 054 p lt00001) Modeling wet basin floodplain-stratified sites at thesite scale we found that canopy cover (β = 0012plusmn 0011)and moisture (β = 00040plusmn 00011) are controls on soilOC (cube root transformation model adjusted R2

= 067p lt 00001) Modeling Big Sandy sites at the reach scalewe found that soil depth (β =minus0012plusmn 00071) and mois-ture (β = 0014plusmn 00026) are dominant controls on soil OCconcentration (no transformation model adjustedR2

= 069p lt 00001)

In general moister deeper soils store more OC at the reachscale whereas OC tends to vary dominantly with depth atthe scale of individual soil samples Although estimated claycontent did not emerge as a significant predictor of OC con-centration it is used to calculate clay-held water to correctour LOI-based OC concentration measurements making itimportant in determining OC for each sample

32 Soil texture

In general soil texture follows a predictable trend withriver size between model groups (Fig 2d) Floodplain type-stratified sites in the wet basin store the most clay followedby slope-stratified sites and then sites in the semiarid basin

Modeling soil texture at the individual soil sample scaleacross slope-stratified sites in the wet basin we foundwhether the reach was confined (β = 542plusmn 532) andwhether the bed material was dominantly sand (β = 1047plusmn613) to be dominant controls on estimated clay contentModeling soil texture for sites stratified by floodplain typein the wet basin yielded no significant trends In the semi-arid basin we found that either valley width (β = 00050plusmn00032) or whether the stream was unconfined (β = 041plusmn033) as well as depth below ground surface (β =minus00069plusmn00033 for model with valley width but not confinement) sig-nificantly control soil texture

To summarize sites from unconfined lower-energyreaches in the wet basin and sites from reaches with widervalley bottoms and at lower depths in the semiarid basin ex-hibited more finely textured soils

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1107

Tabl

e1

Mat

rix

ofal

lmod

els

pres

ente

din

the

text

Eac

hm

odel

islis

ted

bym

odel

grou

pre

spon

seva

riab

lea

ndsc

ale

Scal

ere

fers

toth

esa

mpl

eun

itof

the

mod

ela

ndsi

tere

fers

toa

core

with

the

resp

onse

aver

aged

over

allt

hein

divi

dual

soil

sam

ples

inth

eco

reF

orea

chva

riab

lean

dm

odel

an

aste

risk

()i

ndic

ates

that

the

vari

able

was

incl

uded

inm

odel

sele

ctio

nbu

tnot

the

full

mix

ed-e

ffec

tsm

odel

Am

inus

(minus)

indi

cate

sth

atth

eva

riab

lew

asse

lect

edas

impo

rtan

tin

pred

ictin

gth

ere

spon

sean

dde

note

san

indi

rect

corr

elat

ion

whe

reas

apl

us(+

)ind

icat

esa

dire

ctco

rrel

atio

nIn

the

case

ofco

nfine

men

ta

plus

indi

cate

sth

atun

confi

ned

stre

ams

disp

lay

ahi

gher

-mag

nitu

dere

spon

seva

riab

leI

nth

eca

seof

bed

mat

eria

la

plus

indi

cate

sth

atsa

mpl

esw

ithsa

ndex

hibi

tahi

gher

valu

eof

the

resp

onse

NA

indi

cate

sth

atei

ther

the

vari

able

was

notm

easu

red

for

that

basi

nor

mod

elgr

oup

orth

atit

isth

em

odel

resp

onse

Mod

elgr

oup

Res

pons

eSc

ale

(sam

ple

unit)

Var

iabl

esConfinement

Bedform

Channelslope

Bedmaterial

Multithread

Valleywidth

Bankfullwidth

Bankfulldepth

Streampower

Floodplaintype

Standingwater

Depth1

Claycontent

Moisture

Loggingnearby

Grassespresent

Shrubspresent

Treespresent

Elevation

Basinslope

Canopycover

NLCD

Drainagearea

Wet

basi

nst

rat-

ified

bysl

ope

OC

()

Soil

sam

ple

NA

NAminus

+

O

C(

)Si

te+

NA

NA

+

Moi

stur

e(

)Si

te+

minus

NA

NA

N

A+

Te

xtur

e(

)So

ilsa

mpl

e+

+

N

AN

A

NA

Wet

basi

nst

rat-

ified

byflo

od-

plai

nty

pe

OC

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

minus

N

AO

C(

)Si

teN

AN

AN

AN

AN

AN

AN

AN

AN

A

+

+N

AM

oist

ure

()

Site

NA

NA

NA

NA

NA

NA

NA

NA

NA

+

+

NA

NA

Text

ure2

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

N

AN

A

Sem

iari

dba

sin

OC

()

Soil

sam

ple

N

A

NA

NA

N

AN

A

NA

NAminus

NA

NA

NA

NA

OC

()

Site

N

A

NA

NA

NA

NA

N

AN

Aminus

+N

AN

AN

AN

A

M

oist

ure

()

Site

+N

A

NA

NA

NA

NA

N

AN

A+

NA

NA

NA

NA

NA+

Te

xtur

e3(

)So

ilsa

mpl

e+

NA

N

AN

A+

NA

NA

N

AN

Aminus

NA

NA

NA

NA

NA

1D

epth

refe

rsto

eith

erth

eso

ilsa

mpl

ede

pth

belo

wth

egr

ound

orth

eto

tald

epth

ofth

eco

red

epen

ding

onth

esa

mpl

eun

it2

No

sign

ifica

ntre

sults

wer

eob

serv

edfo

rthi

sm

odel

3Fo

rthi

sm

odel

bot

hva

lley

wid

than

dco

nfine

men

tpre

dict

text

ure

and

can

bein

terp

rete

din

terc

hang

eabl

yH

owev

eri

nclu

ding

both

inth

esa

me

mod

elw

ould

yiel

dpr

oble

ms

due

tom

ultic

ollin

eari

ty

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1108 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

Figure 2 Box plots showing comparisons between model groups of OC concentration at the reach scale (a) OC concentration at the scaleof individual soil samples (b) moisture at the reach scale (c) and estimated clay content at the scale of individual soil samples (d) Endsof dotted lines represent 15 times the interquartile range which is represented by boxes Bold line represents median Circles representoutliers Letters indicate probable differences between groups based on pairwise Wilcoxon (andashc) or t tests (d) with a Holm correctionRanges in parentheses below letters show the 95 confidence interval of the median value for the group (andashc) or the mean value for thegroup (d) in which median confidence intervals were overly constrained due to the categorical nature of our estimated clay content data

33 Soil moisture

Soil moisture is less variable between basins than either tex-ture or OC concentration (Fig 2c) All model groups exhibitsimilar soil moisture conditions although there was signifi-cant variability within each model group

Soil moisture at wet basin sites stratified by slope is domi-nantly controlled by channel slope (β =minus1315plusmn911) ele-vation (β = 00046plusmn 00038) and whether the stream is un-confined (β = 389plusmn 267 model adjusted R2

= 038 p lt00001) At wet basin sites stratified by floodplain type es-timated clay content (β = 0060plusmn 0042) and whether thefloodplain unit had standing water (β = 115plusmn 071) sig-nificantly control soil moisture (model adjusted R2

= 046p lt 00001) In the semiarid basin soil depth (β = 0012plusmn0012) elevation (β = 00023plusmn 00014) and whether thereach was unconfined (β = 091plusmn075) significantly controlsoil moisture (model adjusted R2

= 035 p lt 00001)

4 Discussion

41 Understanding spatial variability in OCconcentration in floodplain soils (H1 and H2)

Comparing the wet basin to the semiarid basin shows thatthe wetter higher primary productivity basin is capable ofstoring greater concentrations of OC in floodplain soils butthat both regions generally store similar OC concentrations infloodplain soils This result partially agrees with the exami-nation of subalpine lake deltas by Scott and Wohl (2017) Inthat study subalpine lake deltas in the wet basin were com-pared to deltas in the drier Colorado Front Range Subalpinelake deltas displayed similar OC concentrations likely dueto competing but complementary OC stabilization and lossmechanisms in each region Those deltas represent a subsetof the broader valley bottom soils studied here This moreexpansive study points to both geomorphic controls such asvalley bottom geometry and factors influenced by climatesuch as canopy cover as controls on OC storage in valley bot-toms These results also agree with the results of Lininger et

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1109

Figure 3 Box plots of sample OC concentration binned by meansample depth for the wet (a) and semiarid (b) basins Ends of dot-ted lines represent 15 times the interquartile range which is rep-resented by boxes Bold line represents median Black circles rep-resent outliers Transparent grey points show all data for each binSample size for each bin is denoted by n

al (2018) which indicate that geomorphic context and vege-tation dynamics control OC concentration on floodplain soilsalong large lowland rivers in Alaska USA

At the reach or site scale wetter soil profiles consistentlyyield higher OC concentrations in all model groups How-ever moisture does not differ significantly between modelgroups (Fig 2c) indicating that this alone cannot explaindifferences between basins Soils tend to be finer in the wetbasin but clay content was not an important predictor of OCconcentration in studied soils Clay content likely influencesOC concentration based on previous research (Hoffmann etal 2009) although the inclusion of coarse soil material (in-cluding particulate organic matter) in our samples may ex-plain the lack of an observed correlation here Variabilityin soil redox conditions (ie the presence of anaerobic mi-crosites) may also introduce variability in respiration rate be-tween basins and among individual soil samples (Keiluweit

et al 2017) That is variability in groundwater hydrologymay result in a heterogeneous mix of aerobic (favoring mi-crobial activity) and anaerobic (suppressing microbial activ-ity) conditions within floodplain soils We did not record re-doximorphic features of our soil samples so we are unableto determine whether this influence contributes to the errorin our models Although confinement plays a strong role indetermining OC concentration in wet basin sites stratified byslope it does not differ significantly between basins (52 of semiarid basin reaches are unconfined compared to 63 of wet basin reaches) The major differences between thesebasins are their hydroclimatic and disturbance regimes Thewet basin is at a lower elevation and has denser and higherbiomass forests (Smithwick et al 2002) compared to thesparser parkland forests of the semiarid basin which likelyalso experiences more frequent fires based on fire historiesof nearby regions (recurrence interval on the order of 101ndash102 years Houston 1973 Loope and Gruell 1973) In addi-tion the volcanic soils present in the wet basin may suppresssoil OC respiration (Matus et al 2014) leading to a highersoil OC storage capacity there

Between basins it is likely that the hydroclimatic regimeinfluencing primary production plays some role in the wetbasinrsquos higher maximum OC concentrations in floodplainsoils compared to those of the semiarid basin Howeversmaller-scale factors such as soil texture and moisture alsolikely play a role and are not related to drainage area (Ta-ble 1) indicating that neither OC concentration nor its con-trolling factors vary continuously along a river network thussupporting H2a and H2b This also indicates that local fac-tors set largely by geomorphic and hydrologic dynamicsplay a significant role in modulating the effect of climate onOC concentrations If the wet and semiarid basins displayedsignificantly different OC concentrations our first hypothesisregarding the interbasin controls on OC concentration wouldbe supported However we instead found that climate andprimary productivity only partially determine OC concentra-tions especially when viewed in the context of geomorphicand hydrologic variability Thus the results do not supportH1

Each basin (or model group) is slightly different in termsof the controls on soil OC concentration moisture andtexture In the wet basin sites stratified by slope higher-elevation sites display higher OC concentrations This is con-trary to the general trend in primary productivity which de-creases with increasing elevation However it is importantto note that the headwaters of the wet basin are dominatedby lakes deltas and other depositional features in relativelybroad glacially carved valleys Subalpine lake deltas havebeen shown to store high OC concentrations in this basin(Scott and Wohl 2017) and many of the highest OC concen-trations we measured were located in broad wet meadowssubalpine lake deltas or other unconfined high-elevationreaches Such unconfined sites also likely have significantlycooler temperatures and tend to have higher soil moisture

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1110 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

contents as shown by our modeling (Table 1) As such al-though high-elevation wet basin sites may receive less OCinput they likely have a low rate of OC respiration resultingin higher OC concentrations on the whole which agrees withthe result of Bao et al (2017) In the semiarid basin our mod-eling suggests that the lower temperatures and higher mois-ture (Table 1) at higher elevations do not compensate for thelower primary productivity as elevation does not correlatewith OC concentration

In both basins unconfined reaches contain wetter andfiner-textured soils which may result in a higher soil OCcapacity Although confinement only relates directly to OCcontent in wet basin sites it does play a strong role in deter-mining moisture which in turn plays a role in regulating OCconcentration in both basins likely via inhibiting microbialactivity (Howard and Howard 1993) The relevance of chan-nel slope in determining soil moisture in the wet basin but notsemiarid basin may reflect the prevalence of high-gradientdebris-flow-dominated channels in the wet basin that largelyexhibit only gravel to boulder substrate which we assumestores minimal fine sediment moisture or OC

In the semiarid basin higher soil depths correlate withmoister and finer-textured soils but less OC concentrationThis indicates the trend in OC with depth likely dominatesthe signal of OC concentration with deeper sites containinga higher proportion of OC-depleted deep samples

42 Inferring sources of OC to floodplain soils (H3)

OC can be input to floodplain soils by two primary mech-anisms First dissolved and particulate OC (including largewood) can be deposited on floodplain surfaces by overbankdeposition thus integrating fluvial sedimentary OC into thefloodplain soil profile or in the case of large wood deposit-ing discrete but potentially large concentrations of OC thatcan later be integrated into the soil profile Second litter anddecomposing vegetation on the floodplain surface can act asautochthonous inputs of OC to floodplain soil

Our modeling of OC concentration yielded results consis-tent with previous investigations of controls on soil OC stor-age capacity (Jobbaacutegy and Jackson 2000 Sutfin and Wohl2017) Sites in the heterogeneous floodplain of the wet basindisplay a direct correlation between canopy cover and OCconcentration indicating that increased litter inputs lead toincreased floodplain soil OC concentration Sediment inputslikely differ between floodplain depositional unit types (egcoarser sediment may deposit on point bars compared tofilled secondary channels) although floodplain type does notpredict OC concentration This indicates that vegetation in-puts may be more dominant than fluvial sediment inputs atthese sites

The finding that buried OC peaks in the wet basin do nothave abnormally high clay contents supports the interpre-tation that wood and litter inputs to soil are the dominantsource of OC in the floodplain soils we examined Buried

peaks can be layers created by overbank deposition and sub-sequent burial of fine OC-bearing sediments (Blazejewski etal 2009 Ricker et al 2013) buried pieces of wood (Wohl2013) or buried organic horizons that are now capped bysediments that prevent OC respiration If overbank deposi-tion of fine sediment caused OC peaks we would expect tosee the soil samples classified as peaks exhibiting high claycontents indicating finer sediment Instead our results sug-gest that in the wet basin buried peaks are likely the result ofeither buried organic horizons or buried wood We observedlarge pieces of decaying buried wood in floodplain cut banksin the wet basin supporting this inference Overbank depo-sition of wood on the floodplain was only observed rarely inthis basin indicating that the OC measured in these soils islikely dominantly autochthonous

In the semiarid basin the two cores that exhibit peaks werecollected from the same meadow just downstream of a now-filled former lake that is a potential source of fine sedimentThe channels draining this meadow exhibit an anabranchingplanform indicating the potential to deposit and bury packetsof potentially OC-rich fine sediments However the majorityof cores do not exhibit OC peaks indicating OC input mainlyfrom vegetation at the surface and continuing OC respirationat depth

OC variation within each core is dominantly a function ofdepth We observe a negative correlation between depth be-low ground surface and OC concentration which has beenobserved in other studies including mountain wetlands andfloodplains (Jobbaacutegy and Jackson 2000 Scott and Wohl2017 Sutfin and Wohl 2017 Zhao et al 2017) In generalthis indicates that at least in mountain river floodplains OCis enriched at the surface and decomposes with depth sim-ilar to upland soils This fits with our finding that the ma-jority of our cores do not exhibit significant OC peaks atdepth and supports the dominance of litter and wood OC in-puts to floodplain soils These results support our hypothesisthat decaying litter and wood not overbank sediment depo-sition dominate the input of OC to floodplain soils in ourstudy basins (H3) We note that floodplain wood may alsoact as a trapping site for overbank fine organic matter facil-itating the deposition and input of OC from decaying vege-tation While these study basins likely accumulate soil OCmainly autochthonously other basins that experience over-bank flows accompanying deposition of fine sediment andburial of organic layers exhibit OC storage that is likely dom-inated by fluvial sediment deposition (eg Blazejewski etal 2009 DrsquoElia et al 2017 Ricker et al 2013) Thus itis likely that flow regime lateral connectivity and sedimenttransport dynamics regulate whether floodplain soil OC isdominantly input by the overbank deposition of fine mate-rial or litter and wood decay

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1111

Figure 4 Conceptual model of physical processes that influenceOC concentration in floodplain soils based on this study and otherliterature Each box corresponds to a major factor that influencesOC concentration Colored text within each box denotes factors thatinfluence OC inputs sorption capacity OC export or respirationrate (note that some processes influence both OC export and respi-ration rate) As sorption capacity increases so does the OC capacityof the soil Conversely as the rate of respiration andor OC exportincreases the soil OC capacity decreases Floodplain soils can onlydevelop high concentrations of OC if there are high rates of OCinput However the capacity of the soils to store OC regulates thatinput and is determined by the competing influences of sorption ca-pacity and the combination of respiration rate and OC export Seethe text for further details

43 Conceptual model of soil OC concentration infloodplain soils

We present a conceptual model to summarize our results andplace them in the context of work examining the controlson OC storage in soils (Fig 4) OC is input to floodplainseither through the decay of vegetation or the deposition offine OC-rich sediment This input of OC only determinesOC concentrations insofar as floodplain soils are capable ofstoring OC That storage is effectively determined by a bal-ance between processes that remove OC from floodplainsnamely respiration or erosion followed by respiration (Berheet al 2007) and processes that regulate OC availability tomicrobes namely the capability of the mineral fraction ofthe soil to sorb OC

OC sorption capacity reflects a few specific processes Al-though soil texture generally relates to the ability of OCto sorb to mineral grains and resulting OC availability soilchemistry also plays a strong and potentially dominant rolein regulating OC sorption capacity (Rasmussen et al 2018)Soil texture is largely determined by valley morphology ac-cording to our modeling (Table 1) placing valley morphol-ogy and resulting sediment transport dynamics (Gran andCzuba 2017 Wohl et al 2017b) as indirect controls on sorp-tion capacity

Respiration rate is largely determined by microbial activityand the availability of OC to microbes Erosion can rapidlyexpose soil OC to microbial respiration (Berhe et al 2007)whereas soils that reside in largely anoxic conditions can ex-hibit low rates of microbial respiration (Boye et al 2017)In addition soil mineralogy chemistry and redox condi-tions (Keiluweit et al 2017) can regulate microbial activity(eg andic soils may limit microbial respiration Matus et al2014) Our results suggesting that moisture controls OC con-tent support the idea that drier soils likely have higher ratesof microbial respiration of OC Moisture is a function of tex-ture valley bottom morphology and elevation (a proxy fortemperature) in our modeling (Table 1) Comparing flood-plain types in the wet basin we find that types with standingwater exhibit significantly higher soil moisture contents thanthose without standing water This indicates spatial variabil-ity in moisture content and likely microbial activity (Howardand Howard 1993) In our modeling this effect translates tospatial variability in OC concentration within floodplains andacross entire basins

OC export from soils refers specifically to leaching of dis-solved OC mainly from shallow soils While we do not di-rectly consider OC export in this study it likely introducesvariability in soil OC concentration and is regulated dom-inantly by hydrology (Aringgren et al 2014 Mcdowell andLikens 1988) soil pore water chemistry (including pH andionic strength Brooks et al 1999 Evans et al 2006) andredox conditions (Knorr 2013) OC export as DOC can actin conjunction with respiration to remove OC from soils andlike respiration rate is countered by sorption capacity

To summarize OC inputs regulated by the capacity ofsoils to store OC and suppress microbial respiration andOC export determine OC concentrations in floodplain soils(Fig 4) OC inputs to floodplain soils come from either au-tochthonous litter accumulation on the floodplain surface al-lochthonous wood deposition or allochthonous deposition offine OC-bearing sediments Our results from the mountain-ous basins we studied suggest that deposition of fine mate-rial in overbank flows is rare leading us to infer that au-tochthonous litter and allochthonous wood inputs to flood-plains dominate OC input in mountain rivers Where soils aremore moist microbial respiration is inhibited and more OCis stored Although soil texture is likely not a limiting factoron OC concentration in these floodplains finer-textured soilslikely have a higher sorption capacity retaining more of theOC input from decaying plant material Our results indicatethat geomorphic and hydrologic characteristics act as bound-ary conditions that regulate soil texture and moisture in turnregulating sorption capacity respiration rate and resultingOC concentrations in floodplain soils

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1112 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

5 Conclusions

We present floodplain soil OC concentration data from twodisparate watersheds to compare how interbasin variabilitybetween the two watersheds compares with intra-basin vari-ability in geomorphic and hydrologic characteristics in de-termining OC concentration Our results indicate that OCconcentration in mountain floodplain soils does not vary pre-dictably along a longitudinal gradient nor does it vary sub-stantially between basins with differing climatic and veg-etation characteristics Instead geomorphic and hydrologiccharacteristics such as valley bottom morphology and soilmoisture dominantly determine floodplain OC concentra-tion

In our study basins decaying litter and wood and not over-bank deposition of fine OC-bearing sediment is the mainsource of OC to floodplain soils It is unclear whether thatdecaying vegetation is dominated by autochthonous litter in-puts or transported downed wood In comparing our basinto other studied floodplain soils it seems that vegetation dy-namics play a strong role in determining OC concentrationswhen fine sediment is not regularly deposited on floodplainsurfaces However we suggest that floodplain soil character-istics set by geomorphic and hydrologic conditions regu-late how OC inputs translate to the spatial distribution of OCalong a river network

This implies that OC storage in floodplains likely cannotbe predicted using consistent downstream trends and thatmanagement prioritization designed to facilitate floodplainOC storage should be based on local geomorphic and hy-drologic process variability within each basin For instancemanagement to increase OC sequestration in floodplain soilswill likely be more effective where floodplains are uncon-fined and soils already experience high moisture conditionsfor much of the year Along these lines our results showthat modeling the floodplain biospheric OC pool to predictits response to warming and subsequent effects on climatebased on regional factors such as climate and net primaryproductivity likely misses the substantial interbasin variabil-ity in OC concentration and storage resulting from variabilityin valley bottom geometry and both geomorphic and hydro-logic processes (eg Doetterl et al 2015)

Although our results provide some insights the questionof whether OC stored in floodplain soil comes dominantlyfrom allochthonous versus autochthonous sources remainsopen Our results imply that more productive spatially het-erogeneous floodplains likely input more OC to soils Flood-plain OC concentration while mediated largely by moisturedynamics likely depends mainly on OC inputs from produc-tive riparian forests This implies that management of OCstorage in mountain river floodplains should focus on therestoration of riparian zones to maintain OC input to soil(eg Bullinger-Weber et al 2014) More detailed studiesin regions with varying sediment transport and hydrologicregimes are needed to determine what conditions favor au-

tochthonous versus allochthonous OC inputs but our resultssuggest that autochthonous sources dominate floodplain OCstorage in basins with relatively low rates of vertical accre-tion and high channelndashfloodplain connectivity that promotesfloodplain wetlands

Data availability Data to support the analyses presented here canbe found in Table S1 and in the Colorado State University DigitalRepository (Scott and Wohl 2018a)

Supplement The supplement related to this article is availableonline at httpsdoiorg105194esurf-6-1101-2018-supplement

Author contributions Field data collection was conducted pri-marily by DNS with assistance from EEW at some sites in theMiddle Fork Snoqualmie basin DNS performed data analysis anddrafted the initial version of the paper which was then edited andrewritten by DNS and EEW

Competing interests The authors declare that they have no con-flict of interest

Acknowledgements This work was funded by NSF grantEAR-1562713 We thank Ellen Daugherty for extensive assistancein fieldwork and discussion that improved the conceptual modelWe thank Katherine Lininger for stimulating discussion thatimproved the paper Detailed and constructive comments from twoanonymous reviewers and Robert Hilton improved the paper Wethank Sara Lowe for assistance in processing and analyzing soilsamples

Edited by Robert HiltonReviewed by two anonymous referees

References

Aringgren A M Buffam I Cooper D M Tiwari T Evans C Dand Laudon H Can the heterogeneity in stream dissolved or-ganic carbon be explained by contributing landscape elementsBiogeosciences 11 1199ndash1213 httpsdoiorg105194bg-11-1199-2014 2014

Appling A P Bernhardt E S and Stanford J A Flood-plain biogeochemical mosaics A multi-dimensional view ofalluvial soils J Geophys Res-Biogeo 119 1538ndash1553httpsdoiorg1010022013JG002543 2014

Aufdenkampe A K Mayorga E Raymond P A MelackJ M Doney S C Alin S R Aalto R E and YooK Riverine coupling of biogeochemical cycles between landoceans and atmosphere Front Ecol Environ 9 53ndash60httpsdoiorg101890100014 2011

Ballantyne A P Alden C B Miller J B Tans P P and WhiteJ W C Increase in observed net carbon dioxide uptake by

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1107

Tabl

e1

Mat

rix

ofal

lmod

els

pres

ente

din

the

text

Eac

hm

odel

islis

ted

bym

odel

grou

pre

spon

seva

riab

lea

ndsc

ale

Scal

ere

fers

toth

esa

mpl

eun

itof

the

mod

ela

ndsi

tere

fers

toa

core

with

the

resp

onse

aver

aged

over

allt

hein

divi

dual

soil

sam

ples

inth

eco

reF

orea

chva

riab

lean

dm

odel

an

aste

risk

()i

ndic

ates

that

the

vari

able

was

incl

uded

inm

odel

sele

ctio

nbu

tnot

the

full

mix

ed-e

ffec

tsm

odel

Am

inus

(minus)

indi

cate

sth

atth

eva

riab

lew

asse

lect

edas

impo

rtan

tin

pred

ictin

gth

ere

spon

sean

dde

note

san

indi

rect

corr

elat

ion

whe

reas

apl

us(+

)ind

icat

esa

dire

ctco

rrel

atio

nIn

the

case

ofco

nfine

men

ta

plus

indi

cate

sth

atun

confi

ned

stre

ams

disp

lay

ahi

gher

-mag

nitu

dere

spon

seva

riab

leI

nth

eca

seof

bed

mat

eria

la

plus

indi

cate

sth

atsa

mpl

esw

ithsa

ndex

hibi

tahi

gher

valu

eof

the

resp

onse

NA

indi

cate

sth

atei

ther

the

vari

able

was

notm

easu

red

for

that

basi

nor

mod

elgr

oup

orth

atit

isth

em

odel

resp

onse

Mod

elgr

oup

Res

pons

eSc

ale

(sam

ple

unit)

Var

iabl

esConfinement

Bedform

Channelslope

Bedmaterial

Multithread

Valleywidth

Bankfullwidth

Bankfulldepth

Streampower

Floodplaintype

Standingwater

Depth1

Claycontent

Moisture

Loggingnearby

Grassespresent

Shrubspresent

Treespresent

Elevation

Basinslope

Canopycover

NLCD

Drainagearea

Wet

basi

nst

rat-

ified

bysl

ope

OC

()

Soil

sam

ple

NA

NAminus

+

O

C(

)Si

te+

NA

NA

+

Moi

stur

e(

)Si

te+

minus

NA

NA

N

A+

Te

xtur

e(

)So

ilsa

mpl

e+

+

N

AN

A

NA

Wet

basi

nst

rat-

ified

byflo

od-

plai

nty

pe

OC

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

minus

N

AO

C(

)Si

teN

AN

AN

AN

AN

AN

AN

AN

AN

A

+

+N

AM

oist

ure

()

Site

NA

NA

NA

NA

NA

NA

NA

NA

NA

+

+

NA

NA

Text

ure2

()

Soil

sam

ple

NA

NA

NA

NA

NA

NA

NA

NA

NA

N

AN

A

Sem

iari

dba

sin

OC

()

Soil

sam

ple

N

A

NA

NA

N

AN

A

NA

NAminus

NA

NA

NA

NA

OC

()

Site

N

A

NA

NA

NA

NA

N

AN

Aminus

+N

AN

AN

AN

A

M

oist

ure

()

Site

+N

A

NA

NA

NA

NA

N

AN

A+

NA

NA

NA

NA

NA+

Te

xtur

e3(

)So

ilsa

mpl

e+

NA

N

AN

A+

NA

NA

N

AN

Aminus

NA

NA

NA

NA

NA

1D

epth

refe

rsto

eith

erth

eso

ilsa

mpl

ede

pth

belo

wth

egr

ound

orth

eto

tald

epth

ofth

eco

red

epen

ding

onth

esa

mpl

eun

it2

No

sign

ifica

ntre

sults

wer

eob

serv

edfo

rthi

sm

odel

3Fo

rthi

sm

odel

bot

hva

lley

wid

than

dco

nfine

men

tpre

dict

text

ure

and

can

bein

terp

rete

din

terc

hang

eabl

yH

owev

eri

nclu

ding

both

inth

esa

me

mod

elw

ould

yiel

dpr

oble

ms

due

tom

ultic

ollin

eari

ty

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1108 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

Figure 2 Box plots showing comparisons between model groups of OC concentration at the reach scale (a) OC concentration at the scaleof individual soil samples (b) moisture at the reach scale (c) and estimated clay content at the scale of individual soil samples (d) Endsof dotted lines represent 15 times the interquartile range which is represented by boxes Bold line represents median Circles representoutliers Letters indicate probable differences between groups based on pairwise Wilcoxon (andashc) or t tests (d) with a Holm correctionRanges in parentheses below letters show the 95 confidence interval of the median value for the group (andashc) or the mean value for thegroup (d) in which median confidence intervals were overly constrained due to the categorical nature of our estimated clay content data

33 Soil moisture

Soil moisture is less variable between basins than either tex-ture or OC concentration (Fig 2c) All model groups exhibitsimilar soil moisture conditions although there was signifi-cant variability within each model group

Soil moisture at wet basin sites stratified by slope is domi-nantly controlled by channel slope (β =minus1315plusmn911) ele-vation (β = 00046plusmn 00038) and whether the stream is un-confined (β = 389plusmn 267 model adjusted R2

= 038 p lt00001) At wet basin sites stratified by floodplain type es-timated clay content (β = 0060plusmn 0042) and whether thefloodplain unit had standing water (β = 115plusmn 071) sig-nificantly control soil moisture (model adjusted R2

= 046p lt 00001) In the semiarid basin soil depth (β = 0012plusmn0012) elevation (β = 00023plusmn 00014) and whether thereach was unconfined (β = 091plusmn075) significantly controlsoil moisture (model adjusted R2

= 035 p lt 00001)

4 Discussion

41 Understanding spatial variability in OCconcentration in floodplain soils (H1 and H2)

Comparing the wet basin to the semiarid basin shows thatthe wetter higher primary productivity basin is capable ofstoring greater concentrations of OC in floodplain soils butthat both regions generally store similar OC concentrations infloodplain soils This result partially agrees with the exami-nation of subalpine lake deltas by Scott and Wohl (2017) Inthat study subalpine lake deltas in the wet basin were com-pared to deltas in the drier Colorado Front Range Subalpinelake deltas displayed similar OC concentrations likely dueto competing but complementary OC stabilization and lossmechanisms in each region Those deltas represent a subsetof the broader valley bottom soils studied here This moreexpansive study points to both geomorphic controls such asvalley bottom geometry and factors influenced by climatesuch as canopy cover as controls on OC storage in valley bot-toms These results also agree with the results of Lininger et

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1109

Figure 3 Box plots of sample OC concentration binned by meansample depth for the wet (a) and semiarid (b) basins Ends of dot-ted lines represent 15 times the interquartile range which is rep-resented by boxes Bold line represents median Black circles rep-resent outliers Transparent grey points show all data for each binSample size for each bin is denoted by n

al (2018) which indicate that geomorphic context and vege-tation dynamics control OC concentration on floodplain soilsalong large lowland rivers in Alaska USA

At the reach or site scale wetter soil profiles consistentlyyield higher OC concentrations in all model groups How-ever moisture does not differ significantly between modelgroups (Fig 2c) indicating that this alone cannot explaindifferences between basins Soils tend to be finer in the wetbasin but clay content was not an important predictor of OCconcentration in studied soils Clay content likely influencesOC concentration based on previous research (Hoffmann etal 2009) although the inclusion of coarse soil material (in-cluding particulate organic matter) in our samples may ex-plain the lack of an observed correlation here Variabilityin soil redox conditions (ie the presence of anaerobic mi-crosites) may also introduce variability in respiration rate be-tween basins and among individual soil samples (Keiluweit

et al 2017) That is variability in groundwater hydrologymay result in a heterogeneous mix of aerobic (favoring mi-crobial activity) and anaerobic (suppressing microbial activ-ity) conditions within floodplain soils We did not record re-doximorphic features of our soil samples so we are unableto determine whether this influence contributes to the errorin our models Although confinement plays a strong role indetermining OC concentration in wet basin sites stratified byslope it does not differ significantly between basins (52 of semiarid basin reaches are unconfined compared to 63 of wet basin reaches) The major differences between thesebasins are their hydroclimatic and disturbance regimes Thewet basin is at a lower elevation and has denser and higherbiomass forests (Smithwick et al 2002) compared to thesparser parkland forests of the semiarid basin which likelyalso experiences more frequent fires based on fire historiesof nearby regions (recurrence interval on the order of 101ndash102 years Houston 1973 Loope and Gruell 1973) In addi-tion the volcanic soils present in the wet basin may suppresssoil OC respiration (Matus et al 2014) leading to a highersoil OC storage capacity there

Between basins it is likely that the hydroclimatic regimeinfluencing primary production plays some role in the wetbasinrsquos higher maximum OC concentrations in floodplainsoils compared to those of the semiarid basin Howeversmaller-scale factors such as soil texture and moisture alsolikely play a role and are not related to drainage area (Ta-ble 1) indicating that neither OC concentration nor its con-trolling factors vary continuously along a river network thussupporting H2a and H2b This also indicates that local fac-tors set largely by geomorphic and hydrologic dynamicsplay a significant role in modulating the effect of climate onOC concentrations If the wet and semiarid basins displayedsignificantly different OC concentrations our first hypothesisregarding the interbasin controls on OC concentration wouldbe supported However we instead found that climate andprimary productivity only partially determine OC concentra-tions especially when viewed in the context of geomorphicand hydrologic variability Thus the results do not supportH1

Each basin (or model group) is slightly different in termsof the controls on soil OC concentration moisture andtexture In the wet basin sites stratified by slope higher-elevation sites display higher OC concentrations This is con-trary to the general trend in primary productivity which de-creases with increasing elevation However it is importantto note that the headwaters of the wet basin are dominatedby lakes deltas and other depositional features in relativelybroad glacially carved valleys Subalpine lake deltas havebeen shown to store high OC concentrations in this basin(Scott and Wohl 2017) and many of the highest OC concen-trations we measured were located in broad wet meadowssubalpine lake deltas or other unconfined high-elevationreaches Such unconfined sites also likely have significantlycooler temperatures and tend to have higher soil moisture

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1110 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

contents as shown by our modeling (Table 1) As such al-though high-elevation wet basin sites may receive less OCinput they likely have a low rate of OC respiration resultingin higher OC concentrations on the whole which agrees withthe result of Bao et al (2017) In the semiarid basin our mod-eling suggests that the lower temperatures and higher mois-ture (Table 1) at higher elevations do not compensate for thelower primary productivity as elevation does not correlatewith OC concentration

In both basins unconfined reaches contain wetter andfiner-textured soils which may result in a higher soil OCcapacity Although confinement only relates directly to OCcontent in wet basin sites it does play a strong role in deter-mining moisture which in turn plays a role in regulating OCconcentration in both basins likely via inhibiting microbialactivity (Howard and Howard 1993) The relevance of chan-nel slope in determining soil moisture in the wet basin but notsemiarid basin may reflect the prevalence of high-gradientdebris-flow-dominated channels in the wet basin that largelyexhibit only gravel to boulder substrate which we assumestores minimal fine sediment moisture or OC

In the semiarid basin higher soil depths correlate withmoister and finer-textured soils but less OC concentrationThis indicates the trend in OC with depth likely dominatesthe signal of OC concentration with deeper sites containinga higher proportion of OC-depleted deep samples

42 Inferring sources of OC to floodplain soils (H3)

OC can be input to floodplain soils by two primary mech-anisms First dissolved and particulate OC (including largewood) can be deposited on floodplain surfaces by overbankdeposition thus integrating fluvial sedimentary OC into thefloodplain soil profile or in the case of large wood deposit-ing discrete but potentially large concentrations of OC thatcan later be integrated into the soil profile Second litter anddecomposing vegetation on the floodplain surface can act asautochthonous inputs of OC to floodplain soil

Our modeling of OC concentration yielded results consis-tent with previous investigations of controls on soil OC stor-age capacity (Jobbaacutegy and Jackson 2000 Sutfin and Wohl2017) Sites in the heterogeneous floodplain of the wet basindisplay a direct correlation between canopy cover and OCconcentration indicating that increased litter inputs lead toincreased floodplain soil OC concentration Sediment inputslikely differ between floodplain depositional unit types (egcoarser sediment may deposit on point bars compared tofilled secondary channels) although floodplain type does notpredict OC concentration This indicates that vegetation in-puts may be more dominant than fluvial sediment inputs atthese sites

The finding that buried OC peaks in the wet basin do nothave abnormally high clay contents supports the interpre-tation that wood and litter inputs to soil are the dominantsource of OC in the floodplain soils we examined Buried

peaks can be layers created by overbank deposition and sub-sequent burial of fine OC-bearing sediments (Blazejewski etal 2009 Ricker et al 2013) buried pieces of wood (Wohl2013) or buried organic horizons that are now capped bysediments that prevent OC respiration If overbank deposi-tion of fine sediment caused OC peaks we would expect tosee the soil samples classified as peaks exhibiting high claycontents indicating finer sediment Instead our results sug-gest that in the wet basin buried peaks are likely the result ofeither buried organic horizons or buried wood We observedlarge pieces of decaying buried wood in floodplain cut banksin the wet basin supporting this inference Overbank depo-sition of wood on the floodplain was only observed rarely inthis basin indicating that the OC measured in these soils islikely dominantly autochthonous

In the semiarid basin the two cores that exhibit peaks werecollected from the same meadow just downstream of a now-filled former lake that is a potential source of fine sedimentThe channels draining this meadow exhibit an anabranchingplanform indicating the potential to deposit and bury packetsof potentially OC-rich fine sediments However the majorityof cores do not exhibit OC peaks indicating OC input mainlyfrom vegetation at the surface and continuing OC respirationat depth

OC variation within each core is dominantly a function ofdepth We observe a negative correlation between depth be-low ground surface and OC concentration which has beenobserved in other studies including mountain wetlands andfloodplains (Jobbaacutegy and Jackson 2000 Scott and Wohl2017 Sutfin and Wohl 2017 Zhao et al 2017) In generalthis indicates that at least in mountain river floodplains OCis enriched at the surface and decomposes with depth sim-ilar to upland soils This fits with our finding that the ma-jority of our cores do not exhibit significant OC peaks atdepth and supports the dominance of litter and wood OC in-puts to floodplain soils These results support our hypothesisthat decaying litter and wood not overbank sediment depo-sition dominate the input of OC to floodplain soils in ourstudy basins (H3) We note that floodplain wood may alsoact as a trapping site for overbank fine organic matter facil-itating the deposition and input of OC from decaying vege-tation While these study basins likely accumulate soil OCmainly autochthonously other basins that experience over-bank flows accompanying deposition of fine sediment andburial of organic layers exhibit OC storage that is likely dom-inated by fluvial sediment deposition (eg Blazejewski etal 2009 DrsquoElia et al 2017 Ricker et al 2013) Thus itis likely that flow regime lateral connectivity and sedimenttransport dynamics regulate whether floodplain soil OC isdominantly input by the overbank deposition of fine mate-rial or litter and wood decay

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1111

Figure 4 Conceptual model of physical processes that influenceOC concentration in floodplain soils based on this study and otherliterature Each box corresponds to a major factor that influencesOC concentration Colored text within each box denotes factors thatinfluence OC inputs sorption capacity OC export or respirationrate (note that some processes influence both OC export and respi-ration rate) As sorption capacity increases so does the OC capacityof the soil Conversely as the rate of respiration andor OC exportincreases the soil OC capacity decreases Floodplain soils can onlydevelop high concentrations of OC if there are high rates of OCinput However the capacity of the soils to store OC regulates thatinput and is determined by the competing influences of sorption ca-pacity and the combination of respiration rate and OC export Seethe text for further details

43 Conceptual model of soil OC concentration infloodplain soils

We present a conceptual model to summarize our results andplace them in the context of work examining the controlson OC storage in soils (Fig 4) OC is input to floodplainseither through the decay of vegetation or the deposition offine OC-rich sediment This input of OC only determinesOC concentrations insofar as floodplain soils are capable ofstoring OC That storage is effectively determined by a bal-ance between processes that remove OC from floodplainsnamely respiration or erosion followed by respiration (Berheet al 2007) and processes that regulate OC availability tomicrobes namely the capability of the mineral fraction ofthe soil to sorb OC

OC sorption capacity reflects a few specific processes Al-though soil texture generally relates to the ability of OCto sorb to mineral grains and resulting OC availability soilchemistry also plays a strong and potentially dominant rolein regulating OC sorption capacity (Rasmussen et al 2018)Soil texture is largely determined by valley morphology ac-cording to our modeling (Table 1) placing valley morphol-ogy and resulting sediment transport dynamics (Gran andCzuba 2017 Wohl et al 2017b) as indirect controls on sorp-tion capacity

Respiration rate is largely determined by microbial activityand the availability of OC to microbes Erosion can rapidlyexpose soil OC to microbial respiration (Berhe et al 2007)whereas soils that reside in largely anoxic conditions can ex-hibit low rates of microbial respiration (Boye et al 2017)In addition soil mineralogy chemistry and redox condi-tions (Keiluweit et al 2017) can regulate microbial activity(eg andic soils may limit microbial respiration Matus et al2014) Our results suggesting that moisture controls OC con-tent support the idea that drier soils likely have higher ratesof microbial respiration of OC Moisture is a function of tex-ture valley bottom morphology and elevation (a proxy fortemperature) in our modeling (Table 1) Comparing flood-plain types in the wet basin we find that types with standingwater exhibit significantly higher soil moisture contents thanthose without standing water This indicates spatial variabil-ity in moisture content and likely microbial activity (Howardand Howard 1993) In our modeling this effect translates tospatial variability in OC concentration within floodplains andacross entire basins

OC export from soils refers specifically to leaching of dis-solved OC mainly from shallow soils While we do not di-rectly consider OC export in this study it likely introducesvariability in soil OC concentration and is regulated dom-inantly by hydrology (Aringgren et al 2014 Mcdowell andLikens 1988) soil pore water chemistry (including pH andionic strength Brooks et al 1999 Evans et al 2006) andredox conditions (Knorr 2013) OC export as DOC can actin conjunction with respiration to remove OC from soils andlike respiration rate is countered by sorption capacity

To summarize OC inputs regulated by the capacity ofsoils to store OC and suppress microbial respiration andOC export determine OC concentrations in floodplain soils(Fig 4) OC inputs to floodplain soils come from either au-tochthonous litter accumulation on the floodplain surface al-lochthonous wood deposition or allochthonous deposition offine OC-bearing sediments Our results from the mountain-ous basins we studied suggest that deposition of fine mate-rial in overbank flows is rare leading us to infer that au-tochthonous litter and allochthonous wood inputs to flood-plains dominate OC input in mountain rivers Where soils aremore moist microbial respiration is inhibited and more OCis stored Although soil texture is likely not a limiting factoron OC concentration in these floodplains finer-textured soilslikely have a higher sorption capacity retaining more of theOC input from decaying plant material Our results indicatethat geomorphic and hydrologic characteristics act as bound-ary conditions that regulate soil texture and moisture in turnregulating sorption capacity respiration rate and resultingOC concentrations in floodplain soils

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1112 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

5 Conclusions

We present floodplain soil OC concentration data from twodisparate watersheds to compare how interbasin variabilitybetween the two watersheds compares with intra-basin vari-ability in geomorphic and hydrologic characteristics in de-termining OC concentration Our results indicate that OCconcentration in mountain floodplain soils does not vary pre-dictably along a longitudinal gradient nor does it vary sub-stantially between basins with differing climatic and veg-etation characteristics Instead geomorphic and hydrologiccharacteristics such as valley bottom morphology and soilmoisture dominantly determine floodplain OC concentra-tion

In our study basins decaying litter and wood and not over-bank deposition of fine OC-bearing sediment is the mainsource of OC to floodplain soils It is unclear whether thatdecaying vegetation is dominated by autochthonous litter in-puts or transported downed wood In comparing our basinto other studied floodplain soils it seems that vegetation dy-namics play a strong role in determining OC concentrationswhen fine sediment is not regularly deposited on floodplainsurfaces However we suggest that floodplain soil character-istics set by geomorphic and hydrologic conditions regu-late how OC inputs translate to the spatial distribution of OCalong a river network

This implies that OC storage in floodplains likely cannotbe predicted using consistent downstream trends and thatmanagement prioritization designed to facilitate floodplainOC storage should be based on local geomorphic and hy-drologic process variability within each basin For instancemanagement to increase OC sequestration in floodplain soilswill likely be more effective where floodplains are uncon-fined and soils already experience high moisture conditionsfor much of the year Along these lines our results showthat modeling the floodplain biospheric OC pool to predictits response to warming and subsequent effects on climatebased on regional factors such as climate and net primaryproductivity likely misses the substantial interbasin variabil-ity in OC concentration and storage resulting from variabilityin valley bottom geometry and both geomorphic and hydro-logic processes (eg Doetterl et al 2015)

Although our results provide some insights the questionof whether OC stored in floodplain soil comes dominantlyfrom allochthonous versus autochthonous sources remainsopen Our results imply that more productive spatially het-erogeneous floodplains likely input more OC to soils Flood-plain OC concentration while mediated largely by moisturedynamics likely depends mainly on OC inputs from produc-tive riparian forests This implies that management of OCstorage in mountain river floodplains should focus on therestoration of riparian zones to maintain OC input to soil(eg Bullinger-Weber et al 2014) More detailed studiesin regions with varying sediment transport and hydrologicregimes are needed to determine what conditions favor au-

tochthonous versus allochthonous OC inputs but our resultssuggest that autochthonous sources dominate floodplain OCstorage in basins with relatively low rates of vertical accre-tion and high channelndashfloodplain connectivity that promotesfloodplain wetlands

Data availability Data to support the analyses presented here canbe found in Table S1 and in the Colorado State University DigitalRepository (Scott and Wohl 2018a)

Supplement The supplement related to this article is availableonline at httpsdoiorg105194esurf-6-1101-2018-supplement

Author contributions Field data collection was conducted pri-marily by DNS with assistance from EEW at some sites in theMiddle Fork Snoqualmie basin DNS performed data analysis anddrafted the initial version of the paper which was then edited andrewritten by DNS and EEW

Competing interests The authors declare that they have no con-flict of interest

Acknowledgements This work was funded by NSF grantEAR-1562713 We thank Ellen Daugherty for extensive assistancein fieldwork and discussion that improved the conceptual modelWe thank Katherine Lininger for stimulating discussion thatimproved the paper Detailed and constructive comments from twoanonymous reviewers and Robert Hilton improved the paper Wethank Sara Lowe for assistance in processing and analyzing soilsamples

Edited by Robert HiltonReviewed by two anonymous referees

References

Aringgren A M Buffam I Cooper D M Tiwari T Evans C Dand Laudon H Can the heterogeneity in stream dissolved or-ganic carbon be explained by contributing landscape elementsBiogeosciences 11 1199ndash1213 httpsdoiorg105194bg-11-1199-2014 2014

Appling A P Bernhardt E S and Stanford J A Flood-plain biogeochemical mosaics A multi-dimensional view ofalluvial soils J Geophys Res-Biogeo 119 1538ndash1553httpsdoiorg1010022013JG002543 2014

Aufdenkampe A K Mayorga E Raymond P A MelackJ M Doney S C Alin S R Aalto R E and YooK Riverine coupling of biogeochemical cycles between landoceans and atmosphere Front Ecol Environ 9 53ndash60httpsdoiorg101890100014 2011

Ballantyne A P Alden C B Miller J B Tans P P and WhiteJ W C Increase in observed net carbon dioxide uptake by

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

1108 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

Figure 2 Box plots showing comparisons between model groups of OC concentration at the reach scale (a) OC concentration at the scaleof individual soil samples (b) moisture at the reach scale (c) and estimated clay content at the scale of individual soil samples (d) Endsof dotted lines represent 15 times the interquartile range which is represented by boxes Bold line represents median Circles representoutliers Letters indicate probable differences between groups based on pairwise Wilcoxon (andashc) or t tests (d) with a Holm correctionRanges in parentheses below letters show the 95 confidence interval of the median value for the group (andashc) or the mean value for thegroup (d) in which median confidence intervals were overly constrained due to the categorical nature of our estimated clay content data

33 Soil moisture

Soil moisture is less variable between basins than either tex-ture or OC concentration (Fig 2c) All model groups exhibitsimilar soil moisture conditions although there was signifi-cant variability within each model group

Soil moisture at wet basin sites stratified by slope is domi-nantly controlled by channel slope (β =minus1315plusmn911) ele-vation (β = 00046plusmn 00038) and whether the stream is un-confined (β = 389plusmn 267 model adjusted R2

= 038 p lt00001) At wet basin sites stratified by floodplain type es-timated clay content (β = 0060plusmn 0042) and whether thefloodplain unit had standing water (β = 115plusmn 071) sig-nificantly control soil moisture (model adjusted R2

= 046p lt 00001) In the semiarid basin soil depth (β = 0012plusmn0012) elevation (β = 00023plusmn 00014) and whether thereach was unconfined (β = 091plusmn075) significantly controlsoil moisture (model adjusted R2

= 035 p lt 00001)

4 Discussion

41 Understanding spatial variability in OCconcentration in floodplain soils (H1 and H2)

Comparing the wet basin to the semiarid basin shows thatthe wetter higher primary productivity basin is capable ofstoring greater concentrations of OC in floodplain soils butthat both regions generally store similar OC concentrations infloodplain soils This result partially agrees with the exami-nation of subalpine lake deltas by Scott and Wohl (2017) Inthat study subalpine lake deltas in the wet basin were com-pared to deltas in the drier Colorado Front Range Subalpinelake deltas displayed similar OC concentrations likely dueto competing but complementary OC stabilization and lossmechanisms in each region Those deltas represent a subsetof the broader valley bottom soils studied here This moreexpansive study points to both geomorphic controls such asvalley bottom geometry and factors influenced by climatesuch as canopy cover as controls on OC storage in valley bot-toms These results also agree with the results of Lininger et

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1109

Figure 3 Box plots of sample OC concentration binned by meansample depth for the wet (a) and semiarid (b) basins Ends of dot-ted lines represent 15 times the interquartile range which is rep-resented by boxes Bold line represents median Black circles rep-resent outliers Transparent grey points show all data for each binSample size for each bin is denoted by n

al (2018) which indicate that geomorphic context and vege-tation dynamics control OC concentration on floodplain soilsalong large lowland rivers in Alaska USA

At the reach or site scale wetter soil profiles consistentlyyield higher OC concentrations in all model groups How-ever moisture does not differ significantly between modelgroups (Fig 2c) indicating that this alone cannot explaindifferences between basins Soils tend to be finer in the wetbasin but clay content was not an important predictor of OCconcentration in studied soils Clay content likely influencesOC concentration based on previous research (Hoffmann etal 2009) although the inclusion of coarse soil material (in-cluding particulate organic matter) in our samples may ex-plain the lack of an observed correlation here Variabilityin soil redox conditions (ie the presence of anaerobic mi-crosites) may also introduce variability in respiration rate be-tween basins and among individual soil samples (Keiluweit

et al 2017) That is variability in groundwater hydrologymay result in a heterogeneous mix of aerobic (favoring mi-crobial activity) and anaerobic (suppressing microbial activ-ity) conditions within floodplain soils We did not record re-doximorphic features of our soil samples so we are unableto determine whether this influence contributes to the errorin our models Although confinement plays a strong role indetermining OC concentration in wet basin sites stratified byslope it does not differ significantly between basins (52 of semiarid basin reaches are unconfined compared to 63 of wet basin reaches) The major differences between thesebasins are their hydroclimatic and disturbance regimes Thewet basin is at a lower elevation and has denser and higherbiomass forests (Smithwick et al 2002) compared to thesparser parkland forests of the semiarid basin which likelyalso experiences more frequent fires based on fire historiesof nearby regions (recurrence interval on the order of 101ndash102 years Houston 1973 Loope and Gruell 1973) In addi-tion the volcanic soils present in the wet basin may suppresssoil OC respiration (Matus et al 2014) leading to a highersoil OC storage capacity there

Between basins it is likely that the hydroclimatic regimeinfluencing primary production plays some role in the wetbasinrsquos higher maximum OC concentrations in floodplainsoils compared to those of the semiarid basin Howeversmaller-scale factors such as soil texture and moisture alsolikely play a role and are not related to drainage area (Ta-ble 1) indicating that neither OC concentration nor its con-trolling factors vary continuously along a river network thussupporting H2a and H2b This also indicates that local fac-tors set largely by geomorphic and hydrologic dynamicsplay a significant role in modulating the effect of climate onOC concentrations If the wet and semiarid basins displayedsignificantly different OC concentrations our first hypothesisregarding the interbasin controls on OC concentration wouldbe supported However we instead found that climate andprimary productivity only partially determine OC concentra-tions especially when viewed in the context of geomorphicand hydrologic variability Thus the results do not supportH1

Each basin (or model group) is slightly different in termsof the controls on soil OC concentration moisture andtexture In the wet basin sites stratified by slope higher-elevation sites display higher OC concentrations This is con-trary to the general trend in primary productivity which de-creases with increasing elevation However it is importantto note that the headwaters of the wet basin are dominatedby lakes deltas and other depositional features in relativelybroad glacially carved valleys Subalpine lake deltas havebeen shown to store high OC concentrations in this basin(Scott and Wohl 2017) and many of the highest OC concen-trations we measured were located in broad wet meadowssubalpine lake deltas or other unconfined high-elevationreaches Such unconfined sites also likely have significantlycooler temperatures and tend to have higher soil moisture

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1110 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

contents as shown by our modeling (Table 1) As such al-though high-elevation wet basin sites may receive less OCinput they likely have a low rate of OC respiration resultingin higher OC concentrations on the whole which agrees withthe result of Bao et al (2017) In the semiarid basin our mod-eling suggests that the lower temperatures and higher mois-ture (Table 1) at higher elevations do not compensate for thelower primary productivity as elevation does not correlatewith OC concentration

In both basins unconfined reaches contain wetter andfiner-textured soils which may result in a higher soil OCcapacity Although confinement only relates directly to OCcontent in wet basin sites it does play a strong role in deter-mining moisture which in turn plays a role in regulating OCconcentration in both basins likely via inhibiting microbialactivity (Howard and Howard 1993) The relevance of chan-nel slope in determining soil moisture in the wet basin but notsemiarid basin may reflect the prevalence of high-gradientdebris-flow-dominated channels in the wet basin that largelyexhibit only gravel to boulder substrate which we assumestores minimal fine sediment moisture or OC

In the semiarid basin higher soil depths correlate withmoister and finer-textured soils but less OC concentrationThis indicates the trend in OC with depth likely dominatesthe signal of OC concentration with deeper sites containinga higher proportion of OC-depleted deep samples

42 Inferring sources of OC to floodplain soils (H3)

OC can be input to floodplain soils by two primary mech-anisms First dissolved and particulate OC (including largewood) can be deposited on floodplain surfaces by overbankdeposition thus integrating fluvial sedimentary OC into thefloodplain soil profile or in the case of large wood deposit-ing discrete but potentially large concentrations of OC thatcan later be integrated into the soil profile Second litter anddecomposing vegetation on the floodplain surface can act asautochthonous inputs of OC to floodplain soil

Our modeling of OC concentration yielded results consis-tent with previous investigations of controls on soil OC stor-age capacity (Jobbaacutegy and Jackson 2000 Sutfin and Wohl2017) Sites in the heterogeneous floodplain of the wet basindisplay a direct correlation between canopy cover and OCconcentration indicating that increased litter inputs lead toincreased floodplain soil OC concentration Sediment inputslikely differ between floodplain depositional unit types (egcoarser sediment may deposit on point bars compared tofilled secondary channels) although floodplain type does notpredict OC concentration This indicates that vegetation in-puts may be more dominant than fluvial sediment inputs atthese sites

The finding that buried OC peaks in the wet basin do nothave abnormally high clay contents supports the interpre-tation that wood and litter inputs to soil are the dominantsource of OC in the floodplain soils we examined Buried

peaks can be layers created by overbank deposition and sub-sequent burial of fine OC-bearing sediments (Blazejewski etal 2009 Ricker et al 2013) buried pieces of wood (Wohl2013) or buried organic horizons that are now capped bysediments that prevent OC respiration If overbank deposi-tion of fine sediment caused OC peaks we would expect tosee the soil samples classified as peaks exhibiting high claycontents indicating finer sediment Instead our results sug-gest that in the wet basin buried peaks are likely the result ofeither buried organic horizons or buried wood We observedlarge pieces of decaying buried wood in floodplain cut banksin the wet basin supporting this inference Overbank depo-sition of wood on the floodplain was only observed rarely inthis basin indicating that the OC measured in these soils islikely dominantly autochthonous

In the semiarid basin the two cores that exhibit peaks werecollected from the same meadow just downstream of a now-filled former lake that is a potential source of fine sedimentThe channels draining this meadow exhibit an anabranchingplanform indicating the potential to deposit and bury packetsof potentially OC-rich fine sediments However the majorityof cores do not exhibit OC peaks indicating OC input mainlyfrom vegetation at the surface and continuing OC respirationat depth

OC variation within each core is dominantly a function ofdepth We observe a negative correlation between depth be-low ground surface and OC concentration which has beenobserved in other studies including mountain wetlands andfloodplains (Jobbaacutegy and Jackson 2000 Scott and Wohl2017 Sutfin and Wohl 2017 Zhao et al 2017) In generalthis indicates that at least in mountain river floodplains OCis enriched at the surface and decomposes with depth sim-ilar to upland soils This fits with our finding that the ma-jority of our cores do not exhibit significant OC peaks atdepth and supports the dominance of litter and wood OC in-puts to floodplain soils These results support our hypothesisthat decaying litter and wood not overbank sediment depo-sition dominate the input of OC to floodplain soils in ourstudy basins (H3) We note that floodplain wood may alsoact as a trapping site for overbank fine organic matter facil-itating the deposition and input of OC from decaying vege-tation While these study basins likely accumulate soil OCmainly autochthonously other basins that experience over-bank flows accompanying deposition of fine sediment andburial of organic layers exhibit OC storage that is likely dom-inated by fluvial sediment deposition (eg Blazejewski etal 2009 DrsquoElia et al 2017 Ricker et al 2013) Thus itis likely that flow regime lateral connectivity and sedimenttransport dynamics regulate whether floodplain soil OC isdominantly input by the overbank deposition of fine mate-rial or litter and wood decay

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1111

Figure 4 Conceptual model of physical processes that influenceOC concentration in floodplain soils based on this study and otherliterature Each box corresponds to a major factor that influencesOC concentration Colored text within each box denotes factors thatinfluence OC inputs sorption capacity OC export or respirationrate (note that some processes influence both OC export and respi-ration rate) As sorption capacity increases so does the OC capacityof the soil Conversely as the rate of respiration andor OC exportincreases the soil OC capacity decreases Floodplain soils can onlydevelop high concentrations of OC if there are high rates of OCinput However the capacity of the soils to store OC regulates thatinput and is determined by the competing influences of sorption ca-pacity and the combination of respiration rate and OC export Seethe text for further details

43 Conceptual model of soil OC concentration infloodplain soils

We present a conceptual model to summarize our results andplace them in the context of work examining the controlson OC storage in soils (Fig 4) OC is input to floodplainseither through the decay of vegetation or the deposition offine OC-rich sediment This input of OC only determinesOC concentrations insofar as floodplain soils are capable ofstoring OC That storage is effectively determined by a bal-ance between processes that remove OC from floodplainsnamely respiration or erosion followed by respiration (Berheet al 2007) and processes that regulate OC availability tomicrobes namely the capability of the mineral fraction ofthe soil to sorb OC

OC sorption capacity reflects a few specific processes Al-though soil texture generally relates to the ability of OCto sorb to mineral grains and resulting OC availability soilchemistry also plays a strong and potentially dominant rolein regulating OC sorption capacity (Rasmussen et al 2018)Soil texture is largely determined by valley morphology ac-cording to our modeling (Table 1) placing valley morphol-ogy and resulting sediment transport dynamics (Gran andCzuba 2017 Wohl et al 2017b) as indirect controls on sorp-tion capacity

Respiration rate is largely determined by microbial activityand the availability of OC to microbes Erosion can rapidlyexpose soil OC to microbial respiration (Berhe et al 2007)whereas soils that reside in largely anoxic conditions can ex-hibit low rates of microbial respiration (Boye et al 2017)In addition soil mineralogy chemistry and redox condi-tions (Keiluweit et al 2017) can regulate microbial activity(eg andic soils may limit microbial respiration Matus et al2014) Our results suggesting that moisture controls OC con-tent support the idea that drier soils likely have higher ratesof microbial respiration of OC Moisture is a function of tex-ture valley bottom morphology and elevation (a proxy fortemperature) in our modeling (Table 1) Comparing flood-plain types in the wet basin we find that types with standingwater exhibit significantly higher soil moisture contents thanthose without standing water This indicates spatial variabil-ity in moisture content and likely microbial activity (Howardand Howard 1993) In our modeling this effect translates tospatial variability in OC concentration within floodplains andacross entire basins

OC export from soils refers specifically to leaching of dis-solved OC mainly from shallow soils While we do not di-rectly consider OC export in this study it likely introducesvariability in soil OC concentration and is regulated dom-inantly by hydrology (Aringgren et al 2014 Mcdowell andLikens 1988) soil pore water chemistry (including pH andionic strength Brooks et al 1999 Evans et al 2006) andredox conditions (Knorr 2013) OC export as DOC can actin conjunction with respiration to remove OC from soils andlike respiration rate is countered by sorption capacity

To summarize OC inputs regulated by the capacity ofsoils to store OC and suppress microbial respiration andOC export determine OC concentrations in floodplain soils(Fig 4) OC inputs to floodplain soils come from either au-tochthonous litter accumulation on the floodplain surface al-lochthonous wood deposition or allochthonous deposition offine OC-bearing sediments Our results from the mountain-ous basins we studied suggest that deposition of fine mate-rial in overbank flows is rare leading us to infer that au-tochthonous litter and allochthonous wood inputs to flood-plains dominate OC input in mountain rivers Where soils aremore moist microbial respiration is inhibited and more OCis stored Although soil texture is likely not a limiting factoron OC concentration in these floodplains finer-textured soilslikely have a higher sorption capacity retaining more of theOC input from decaying plant material Our results indicatethat geomorphic and hydrologic characteristics act as bound-ary conditions that regulate soil texture and moisture in turnregulating sorption capacity respiration rate and resultingOC concentrations in floodplain soils

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1112 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

5 Conclusions

We present floodplain soil OC concentration data from twodisparate watersheds to compare how interbasin variabilitybetween the two watersheds compares with intra-basin vari-ability in geomorphic and hydrologic characteristics in de-termining OC concentration Our results indicate that OCconcentration in mountain floodplain soils does not vary pre-dictably along a longitudinal gradient nor does it vary sub-stantially between basins with differing climatic and veg-etation characteristics Instead geomorphic and hydrologiccharacteristics such as valley bottom morphology and soilmoisture dominantly determine floodplain OC concentra-tion

In our study basins decaying litter and wood and not over-bank deposition of fine OC-bearing sediment is the mainsource of OC to floodplain soils It is unclear whether thatdecaying vegetation is dominated by autochthonous litter in-puts or transported downed wood In comparing our basinto other studied floodplain soils it seems that vegetation dy-namics play a strong role in determining OC concentrationswhen fine sediment is not regularly deposited on floodplainsurfaces However we suggest that floodplain soil character-istics set by geomorphic and hydrologic conditions regu-late how OC inputs translate to the spatial distribution of OCalong a river network

This implies that OC storage in floodplains likely cannotbe predicted using consistent downstream trends and thatmanagement prioritization designed to facilitate floodplainOC storage should be based on local geomorphic and hy-drologic process variability within each basin For instancemanagement to increase OC sequestration in floodplain soilswill likely be more effective where floodplains are uncon-fined and soils already experience high moisture conditionsfor much of the year Along these lines our results showthat modeling the floodplain biospheric OC pool to predictits response to warming and subsequent effects on climatebased on regional factors such as climate and net primaryproductivity likely misses the substantial interbasin variabil-ity in OC concentration and storage resulting from variabilityin valley bottom geometry and both geomorphic and hydro-logic processes (eg Doetterl et al 2015)

Although our results provide some insights the questionof whether OC stored in floodplain soil comes dominantlyfrom allochthonous versus autochthonous sources remainsopen Our results imply that more productive spatially het-erogeneous floodplains likely input more OC to soils Flood-plain OC concentration while mediated largely by moisturedynamics likely depends mainly on OC inputs from produc-tive riparian forests This implies that management of OCstorage in mountain river floodplains should focus on therestoration of riparian zones to maintain OC input to soil(eg Bullinger-Weber et al 2014) More detailed studiesin regions with varying sediment transport and hydrologicregimes are needed to determine what conditions favor au-

tochthonous versus allochthonous OC inputs but our resultssuggest that autochthonous sources dominate floodplain OCstorage in basins with relatively low rates of vertical accre-tion and high channelndashfloodplain connectivity that promotesfloodplain wetlands

Data availability Data to support the analyses presented here canbe found in Table S1 and in the Colorado State University DigitalRepository (Scott and Wohl 2018a)

Supplement The supplement related to this article is availableonline at httpsdoiorg105194esurf-6-1101-2018-supplement

Author contributions Field data collection was conducted pri-marily by DNS with assistance from EEW at some sites in theMiddle Fork Snoqualmie basin DNS performed data analysis anddrafted the initial version of the paper which was then edited andrewritten by DNS and EEW

Competing interests The authors declare that they have no con-flict of interest

Acknowledgements This work was funded by NSF grantEAR-1562713 We thank Ellen Daugherty for extensive assistancein fieldwork and discussion that improved the conceptual modelWe thank Katherine Lininger for stimulating discussion thatimproved the paper Detailed and constructive comments from twoanonymous reviewers and Robert Hilton improved the paper Wethank Sara Lowe for assistance in processing and analyzing soilsamples

Edited by Robert HiltonReviewed by two anonymous referees

References

Aringgren A M Buffam I Cooper D M Tiwari T Evans C Dand Laudon H Can the heterogeneity in stream dissolved or-ganic carbon be explained by contributing landscape elementsBiogeosciences 11 1199ndash1213 httpsdoiorg105194bg-11-1199-2014 2014

Appling A P Bernhardt E S and Stanford J A Flood-plain biogeochemical mosaics A multi-dimensional view ofalluvial soils J Geophys Res-Biogeo 119 1538ndash1553httpsdoiorg1010022013JG002543 2014

Aufdenkampe A K Mayorga E Raymond P A MelackJ M Doney S C Alin S R Aalto R E and YooK Riverine coupling of biogeochemical cycles between landoceans and atmosphere Front Ecol Environ 9 53ndash60httpsdoiorg101890100014 2011

Ballantyne A P Alden C B Miller J B Tans P P and WhiteJ W C Increase in observed net carbon dioxide uptake by

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1109

Figure 3 Box plots of sample OC concentration binned by meansample depth for the wet (a) and semiarid (b) basins Ends of dot-ted lines represent 15 times the interquartile range which is rep-resented by boxes Bold line represents median Black circles rep-resent outliers Transparent grey points show all data for each binSample size for each bin is denoted by n

al (2018) which indicate that geomorphic context and vege-tation dynamics control OC concentration on floodplain soilsalong large lowland rivers in Alaska USA

At the reach or site scale wetter soil profiles consistentlyyield higher OC concentrations in all model groups How-ever moisture does not differ significantly between modelgroups (Fig 2c) indicating that this alone cannot explaindifferences between basins Soils tend to be finer in the wetbasin but clay content was not an important predictor of OCconcentration in studied soils Clay content likely influencesOC concentration based on previous research (Hoffmann etal 2009) although the inclusion of coarse soil material (in-cluding particulate organic matter) in our samples may ex-plain the lack of an observed correlation here Variabilityin soil redox conditions (ie the presence of anaerobic mi-crosites) may also introduce variability in respiration rate be-tween basins and among individual soil samples (Keiluweit

et al 2017) That is variability in groundwater hydrologymay result in a heterogeneous mix of aerobic (favoring mi-crobial activity) and anaerobic (suppressing microbial activ-ity) conditions within floodplain soils We did not record re-doximorphic features of our soil samples so we are unableto determine whether this influence contributes to the errorin our models Although confinement plays a strong role indetermining OC concentration in wet basin sites stratified byslope it does not differ significantly between basins (52 of semiarid basin reaches are unconfined compared to 63 of wet basin reaches) The major differences between thesebasins are their hydroclimatic and disturbance regimes Thewet basin is at a lower elevation and has denser and higherbiomass forests (Smithwick et al 2002) compared to thesparser parkland forests of the semiarid basin which likelyalso experiences more frequent fires based on fire historiesof nearby regions (recurrence interval on the order of 101ndash102 years Houston 1973 Loope and Gruell 1973) In addi-tion the volcanic soils present in the wet basin may suppresssoil OC respiration (Matus et al 2014) leading to a highersoil OC storage capacity there

Between basins it is likely that the hydroclimatic regimeinfluencing primary production plays some role in the wetbasinrsquos higher maximum OC concentrations in floodplainsoils compared to those of the semiarid basin Howeversmaller-scale factors such as soil texture and moisture alsolikely play a role and are not related to drainage area (Ta-ble 1) indicating that neither OC concentration nor its con-trolling factors vary continuously along a river network thussupporting H2a and H2b This also indicates that local fac-tors set largely by geomorphic and hydrologic dynamicsplay a significant role in modulating the effect of climate onOC concentrations If the wet and semiarid basins displayedsignificantly different OC concentrations our first hypothesisregarding the interbasin controls on OC concentration wouldbe supported However we instead found that climate andprimary productivity only partially determine OC concentra-tions especially when viewed in the context of geomorphicand hydrologic variability Thus the results do not supportH1

Each basin (or model group) is slightly different in termsof the controls on soil OC concentration moisture andtexture In the wet basin sites stratified by slope higher-elevation sites display higher OC concentrations This is con-trary to the general trend in primary productivity which de-creases with increasing elevation However it is importantto note that the headwaters of the wet basin are dominatedby lakes deltas and other depositional features in relativelybroad glacially carved valleys Subalpine lake deltas havebeen shown to store high OC concentrations in this basin(Scott and Wohl 2017) and many of the highest OC concen-trations we measured were located in broad wet meadowssubalpine lake deltas or other unconfined high-elevationreaches Such unconfined sites also likely have significantlycooler temperatures and tend to have higher soil moisture

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1110 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

contents as shown by our modeling (Table 1) As such al-though high-elevation wet basin sites may receive less OCinput they likely have a low rate of OC respiration resultingin higher OC concentrations on the whole which agrees withthe result of Bao et al (2017) In the semiarid basin our mod-eling suggests that the lower temperatures and higher mois-ture (Table 1) at higher elevations do not compensate for thelower primary productivity as elevation does not correlatewith OC concentration

In both basins unconfined reaches contain wetter andfiner-textured soils which may result in a higher soil OCcapacity Although confinement only relates directly to OCcontent in wet basin sites it does play a strong role in deter-mining moisture which in turn plays a role in regulating OCconcentration in both basins likely via inhibiting microbialactivity (Howard and Howard 1993) The relevance of chan-nel slope in determining soil moisture in the wet basin but notsemiarid basin may reflect the prevalence of high-gradientdebris-flow-dominated channels in the wet basin that largelyexhibit only gravel to boulder substrate which we assumestores minimal fine sediment moisture or OC

In the semiarid basin higher soil depths correlate withmoister and finer-textured soils but less OC concentrationThis indicates the trend in OC with depth likely dominatesthe signal of OC concentration with deeper sites containinga higher proportion of OC-depleted deep samples

42 Inferring sources of OC to floodplain soils (H3)

OC can be input to floodplain soils by two primary mech-anisms First dissolved and particulate OC (including largewood) can be deposited on floodplain surfaces by overbankdeposition thus integrating fluvial sedimentary OC into thefloodplain soil profile or in the case of large wood deposit-ing discrete but potentially large concentrations of OC thatcan later be integrated into the soil profile Second litter anddecomposing vegetation on the floodplain surface can act asautochthonous inputs of OC to floodplain soil

Our modeling of OC concentration yielded results consis-tent with previous investigations of controls on soil OC stor-age capacity (Jobbaacutegy and Jackson 2000 Sutfin and Wohl2017) Sites in the heterogeneous floodplain of the wet basindisplay a direct correlation between canopy cover and OCconcentration indicating that increased litter inputs lead toincreased floodplain soil OC concentration Sediment inputslikely differ between floodplain depositional unit types (egcoarser sediment may deposit on point bars compared tofilled secondary channels) although floodplain type does notpredict OC concentration This indicates that vegetation in-puts may be more dominant than fluvial sediment inputs atthese sites

The finding that buried OC peaks in the wet basin do nothave abnormally high clay contents supports the interpre-tation that wood and litter inputs to soil are the dominantsource of OC in the floodplain soils we examined Buried

peaks can be layers created by overbank deposition and sub-sequent burial of fine OC-bearing sediments (Blazejewski etal 2009 Ricker et al 2013) buried pieces of wood (Wohl2013) or buried organic horizons that are now capped bysediments that prevent OC respiration If overbank deposi-tion of fine sediment caused OC peaks we would expect tosee the soil samples classified as peaks exhibiting high claycontents indicating finer sediment Instead our results sug-gest that in the wet basin buried peaks are likely the result ofeither buried organic horizons or buried wood We observedlarge pieces of decaying buried wood in floodplain cut banksin the wet basin supporting this inference Overbank depo-sition of wood on the floodplain was only observed rarely inthis basin indicating that the OC measured in these soils islikely dominantly autochthonous

In the semiarid basin the two cores that exhibit peaks werecollected from the same meadow just downstream of a now-filled former lake that is a potential source of fine sedimentThe channels draining this meadow exhibit an anabranchingplanform indicating the potential to deposit and bury packetsof potentially OC-rich fine sediments However the majorityof cores do not exhibit OC peaks indicating OC input mainlyfrom vegetation at the surface and continuing OC respirationat depth

OC variation within each core is dominantly a function ofdepth We observe a negative correlation between depth be-low ground surface and OC concentration which has beenobserved in other studies including mountain wetlands andfloodplains (Jobbaacutegy and Jackson 2000 Scott and Wohl2017 Sutfin and Wohl 2017 Zhao et al 2017) In generalthis indicates that at least in mountain river floodplains OCis enriched at the surface and decomposes with depth sim-ilar to upland soils This fits with our finding that the ma-jority of our cores do not exhibit significant OC peaks atdepth and supports the dominance of litter and wood OC in-puts to floodplain soils These results support our hypothesisthat decaying litter and wood not overbank sediment depo-sition dominate the input of OC to floodplain soils in ourstudy basins (H3) We note that floodplain wood may alsoact as a trapping site for overbank fine organic matter facil-itating the deposition and input of OC from decaying vege-tation While these study basins likely accumulate soil OCmainly autochthonously other basins that experience over-bank flows accompanying deposition of fine sediment andburial of organic layers exhibit OC storage that is likely dom-inated by fluvial sediment deposition (eg Blazejewski etal 2009 DrsquoElia et al 2017 Ricker et al 2013) Thus itis likely that flow regime lateral connectivity and sedimenttransport dynamics regulate whether floodplain soil OC isdominantly input by the overbank deposition of fine mate-rial or litter and wood decay

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1111

Figure 4 Conceptual model of physical processes that influenceOC concentration in floodplain soils based on this study and otherliterature Each box corresponds to a major factor that influencesOC concentration Colored text within each box denotes factors thatinfluence OC inputs sorption capacity OC export or respirationrate (note that some processes influence both OC export and respi-ration rate) As sorption capacity increases so does the OC capacityof the soil Conversely as the rate of respiration andor OC exportincreases the soil OC capacity decreases Floodplain soils can onlydevelop high concentrations of OC if there are high rates of OCinput However the capacity of the soils to store OC regulates thatinput and is determined by the competing influences of sorption ca-pacity and the combination of respiration rate and OC export Seethe text for further details

43 Conceptual model of soil OC concentration infloodplain soils

We present a conceptual model to summarize our results andplace them in the context of work examining the controlson OC storage in soils (Fig 4) OC is input to floodplainseither through the decay of vegetation or the deposition offine OC-rich sediment This input of OC only determinesOC concentrations insofar as floodplain soils are capable ofstoring OC That storage is effectively determined by a bal-ance between processes that remove OC from floodplainsnamely respiration or erosion followed by respiration (Berheet al 2007) and processes that regulate OC availability tomicrobes namely the capability of the mineral fraction ofthe soil to sorb OC

OC sorption capacity reflects a few specific processes Al-though soil texture generally relates to the ability of OCto sorb to mineral grains and resulting OC availability soilchemistry also plays a strong and potentially dominant rolein regulating OC sorption capacity (Rasmussen et al 2018)Soil texture is largely determined by valley morphology ac-cording to our modeling (Table 1) placing valley morphol-ogy and resulting sediment transport dynamics (Gran andCzuba 2017 Wohl et al 2017b) as indirect controls on sorp-tion capacity

Respiration rate is largely determined by microbial activityand the availability of OC to microbes Erosion can rapidlyexpose soil OC to microbial respiration (Berhe et al 2007)whereas soils that reside in largely anoxic conditions can ex-hibit low rates of microbial respiration (Boye et al 2017)In addition soil mineralogy chemistry and redox condi-tions (Keiluweit et al 2017) can regulate microbial activity(eg andic soils may limit microbial respiration Matus et al2014) Our results suggesting that moisture controls OC con-tent support the idea that drier soils likely have higher ratesof microbial respiration of OC Moisture is a function of tex-ture valley bottom morphology and elevation (a proxy fortemperature) in our modeling (Table 1) Comparing flood-plain types in the wet basin we find that types with standingwater exhibit significantly higher soil moisture contents thanthose without standing water This indicates spatial variabil-ity in moisture content and likely microbial activity (Howardand Howard 1993) In our modeling this effect translates tospatial variability in OC concentration within floodplains andacross entire basins

OC export from soils refers specifically to leaching of dis-solved OC mainly from shallow soils While we do not di-rectly consider OC export in this study it likely introducesvariability in soil OC concentration and is regulated dom-inantly by hydrology (Aringgren et al 2014 Mcdowell andLikens 1988) soil pore water chemistry (including pH andionic strength Brooks et al 1999 Evans et al 2006) andredox conditions (Knorr 2013) OC export as DOC can actin conjunction with respiration to remove OC from soils andlike respiration rate is countered by sorption capacity

To summarize OC inputs regulated by the capacity ofsoils to store OC and suppress microbial respiration andOC export determine OC concentrations in floodplain soils(Fig 4) OC inputs to floodplain soils come from either au-tochthonous litter accumulation on the floodplain surface al-lochthonous wood deposition or allochthonous deposition offine OC-bearing sediments Our results from the mountain-ous basins we studied suggest that deposition of fine mate-rial in overbank flows is rare leading us to infer that au-tochthonous litter and allochthonous wood inputs to flood-plains dominate OC input in mountain rivers Where soils aremore moist microbial respiration is inhibited and more OCis stored Although soil texture is likely not a limiting factoron OC concentration in these floodplains finer-textured soilslikely have a higher sorption capacity retaining more of theOC input from decaying plant material Our results indicatethat geomorphic and hydrologic characteristics act as bound-ary conditions that regulate soil texture and moisture in turnregulating sorption capacity respiration rate and resultingOC concentrations in floodplain soils

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1112 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

5 Conclusions

We present floodplain soil OC concentration data from twodisparate watersheds to compare how interbasin variabilitybetween the two watersheds compares with intra-basin vari-ability in geomorphic and hydrologic characteristics in de-termining OC concentration Our results indicate that OCconcentration in mountain floodplain soils does not vary pre-dictably along a longitudinal gradient nor does it vary sub-stantially between basins with differing climatic and veg-etation characteristics Instead geomorphic and hydrologiccharacteristics such as valley bottom morphology and soilmoisture dominantly determine floodplain OC concentra-tion

In our study basins decaying litter and wood and not over-bank deposition of fine OC-bearing sediment is the mainsource of OC to floodplain soils It is unclear whether thatdecaying vegetation is dominated by autochthonous litter in-puts or transported downed wood In comparing our basinto other studied floodplain soils it seems that vegetation dy-namics play a strong role in determining OC concentrationswhen fine sediment is not regularly deposited on floodplainsurfaces However we suggest that floodplain soil character-istics set by geomorphic and hydrologic conditions regu-late how OC inputs translate to the spatial distribution of OCalong a river network

This implies that OC storage in floodplains likely cannotbe predicted using consistent downstream trends and thatmanagement prioritization designed to facilitate floodplainOC storage should be based on local geomorphic and hy-drologic process variability within each basin For instancemanagement to increase OC sequestration in floodplain soilswill likely be more effective where floodplains are uncon-fined and soils already experience high moisture conditionsfor much of the year Along these lines our results showthat modeling the floodplain biospheric OC pool to predictits response to warming and subsequent effects on climatebased on regional factors such as climate and net primaryproductivity likely misses the substantial interbasin variabil-ity in OC concentration and storage resulting from variabilityin valley bottom geometry and both geomorphic and hydro-logic processes (eg Doetterl et al 2015)

Although our results provide some insights the questionof whether OC stored in floodplain soil comes dominantlyfrom allochthonous versus autochthonous sources remainsopen Our results imply that more productive spatially het-erogeneous floodplains likely input more OC to soils Flood-plain OC concentration while mediated largely by moisturedynamics likely depends mainly on OC inputs from produc-tive riparian forests This implies that management of OCstorage in mountain river floodplains should focus on therestoration of riparian zones to maintain OC input to soil(eg Bullinger-Weber et al 2014) More detailed studiesin regions with varying sediment transport and hydrologicregimes are needed to determine what conditions favor au-

tochthonous versus allochthonous OC inputs but our resultssuggest that autochthonous sources dominate floodplain OCstorage in basins with relatively low rates of vertical accre-tion and high channelndashfloodplain connectivity that promotesfloodplain wetlands

Data availability Data to support the analyses presented here canbe found in Table S1 and in the Colorado State University DigitalRepository (Scott and Wohl 2018a)

Supplement The supplement related to this article is availableonline at httpsdoiorg105194esurf-6-1101-2018-supplement

Author contributions Field data collection was conducted pri-marily by DNS with assistance from EEW at some sites in theMiddle Fork Snoqualmie basin DNS performed data analysis anddrafted the initial version of the paper which was then edited andrewritten by DNS and EEW

Competing interests The authors declare that they have no con-flict of interest

Acknowledgements This work was funded by NSF grantEAR-1562713 We thank Ellen Daugherty for extensive assistancein fieldwork and discussion that improved the conceptual modelWe thank Katherine Lininger for stimulating discussion thatimproved the paper Detailed and constructive comments from twoanonymous reviewers and Robert Hilton improved the paper Wethank Sara Lowe for assistance in processing and analyzing soilsamples

Edited by Robert HiltonReviewed by two anonymous referees

References

Aringgren A M Buffam I Cooper D M Tiwari T Evans C Dand Laudon H Can the heterogeneity in stream dissolved or-ganic carbon be explained by contributing landscape elementsBiogeosciences 11 1199ndash1213 httpsdoiorg105194bg-11-1199-2014 2014

Appling A P Bernhardt E S and Stanford J A Flood-plain biogeochemical mosaics A multi-dimensional view ofalluvial soils J Geophys Res-Biogeo 119 1538ndash1553httpsdoiorg1010022013JG002543 2014

Aufdenkampe A K Mayorga E Raymond P A MelackJ M Doney S C Alin S R Aalto R E and YooK Riverine coupling of biogeochemical cycles between landoceans and atmosphere Front Ecol Environ 9 53ndash60httpsdoiorg101890100014 2011

Ballantyne A P Alden C B Miller J B Tans P P and WhiteJ W C Increase in observed net carbon dioxide uptake by

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

1110 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

contents as shown by our modeling (Table 1) As such al-though high-elevation wet basin sites may receive less OCinput they likely have a low rate of OC respiration resultingin higher OC concentrations on the whole which agrees withthe result of Bao et al (2017) In the semiarid basin our mod-eling suggests that the lower temperatures and higher mois-ture (Table 1) at higher elevations do not compensate for thelower primary productivity as elevation does not correlatewith OC concentration

In both basins unconfined reaches contain wetter andfiner-textured soils which may result in a higher soil OCcapacity Although confinement only relates directly to OCcontent in wet basin sites it does play a strong role in deter-mining moisture which in turn plays a role in regulating OCconcentration in both basins likely via inhibiting microbialactivity (Howard and Howard 1993) The relevance of chan-nel slope in determining soil moisture in the wet basin but notsemiarid basin may reflect the prevalence of high-gradientdebris-flow-dominated channels in the wet basin that largelyexhibit only gravel to boulder substrate which we assumestores minimal fine sediment moisture or OC

In the semiarid basin higher soil depths correlate withmoister and finer-textured soils but less OC concentrationThis indicates the trend in OC with depth likely dominatesthe signal of OC concentration with deeper sites containinga higher proportion of OC-depleted deep samples

42 Inferring sources of OC to floodplain soils (H3)

OC can be input to floodplain soils by two primary mech-anisms First dissolved and particulate OC (including largewood) can be deposited on floodplain surfaces by overbankdeposition thus integrating fluvial sedimentary OC into thefloodplain soil profile or in the case of large wood deposit-ing discrete but potentially large concentrations of OC thatcan later be integrated into the soil profile Second litter anddecomposing vegetation on the floodplain surface can act asautochthonous inputs of OC to floodplain soil

Our modeling of OC concentration yielded results consis-tent with previous investigations of controls on soil OC stor-age capacity (Jobbaacutegy and Jackson 2000 Sutfin and Wohl2017) Sites in the heterogeneous floodplain of the wet basindisplay a direct correlation between canopy cover and OCconcentration indicating that increased litter inputs lead toincreased floodplain soil OC concentration Sediment inputslikely differ between floodplain depositional unit types (egcoarser sediment may deposit on point bars compared tofilled secondary channels) although floodplain type does notpredict OC concentration This indicates that vegetation in-puts may be more dominant than fluvial sediment inputs atthese sites

The finding that buried OC peaks in the wet basin do nothave abnormally high clay contents supports the interpre-tation that wood and litter inputs to soil are the dominantsource of OC in the floodplain soils we examined Buried

peaks can be layers created by overbank deposition and sub-sequent burial of fine OC-bearing sediments (Blazejewski etal 2009 Ricker et al 2013) buried pieces of wood (Wohl2013) or buried organic horizons that are now capped bysediments that prevent OC respiration If overbank deposi-tion of fine sediment caused OC peaks we would expect tosee the soil samples classified as peaks exhibiting high claycontents indicating finer sediment Instead our results sug-gest that in the wet basin buried peaks are likely the result ofeither buried organic horizons or buried wood We observedlarge pieces of decaying buried wood in floodplain cut banksin the wet basin supporting this inference Overbank depo-sition of wood on the floodplain was only observed rarely inthis basin indicating that the OC measured in these soils islikely dominantly autochthonous

In the semiarid basin the two cores that exhibit peaks werecollected from the same meadow just downstream of a now-filled former lake that is a potential source of fine sedimentThe channels draining this meadow exhibit an anabranchingplanform indicating the potential to deposit and bury packetsof potentially OC-rich fine sediments However the majorityof cores do not exhibit OC peaks indicating OC input mainlyfrom vegetation at the surface and continuing OC respirationat depth

OC variation within each core is dominantly a function ofdepth We observe a negative correlation between depth be-low ground surface and OC concentration which has beenobserved in other studies including mountain wetlands andfloodplains (Jobbaacutegy and Jackson 2000 Scott and Wohl2017 Sutfin and Wohl 2017 Zhao et al 2017) In generalthis indicates that at least in mountain river floodplains OCis enriched at the surface and decomposes with depth sim-ilar to upland soils This fits with our finding that the ma-jority of our cores do not exhibit significant OC peaks atdepth and supports the dominance of litter and wood OC in-puts to floodplain soils These results support our hypothesisthat decaying litter and wood not overbank sediment depo-sition dominate the input of OC to floodplain soils in ourstudy basins (H3) We note that floodplain wood may alsoact as a trapping site for overbank fine organic matter facil-itating the deposition and input of OC from decaying vege-tation While these study basins likely accumulate soil OCmainly autochthonously other basins that experience over-bank flows accompanying deposition of fine sediment andburial of organic layers exhibit OC storage that is likely dom-inated by fluvial sediment deposition (eg Blazejewski etal 2009 DrsquoElia et al 2017 Ricker et al 2013) Thus itis likely that flow regime lateral connectivity and sedimenttransport dynamics regulate whether floodplain soil OC isdominantly input by the overbank deposition of fine mate-rial or litter and wood decay

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1111

Figure 4 Conceptual model of physical processes that influenceOC concentration in floodplain soils based on this study and otherliterature Each box corresponds to a major factor that influencesOC concentration Colored text within each box denotes factors thatinfluence OC inputs sorption capacity OC export or respirationrate (note that some processes influence both OC export and respi-ration rate) As sorption capacity increases so does the OC capacityof the soil Conversely as the rate of respiration andor OC exportincreases the soil OC capacity decreases Floodplain soils can onlydevelop high concentrations of OC if there are high rates of OCinput However the capacity of the soils to store OC regulates thatinput and is determined by the competing influences of sorption ca-pacity and the combination of respiration rate and OC export Seethe text for further details

43 Conceptual model of soil OC concentration infloodplain soils

We present a conceptual model to summarize our results andplace them in the context of work examining the controlson OC storage in soils (Fig 4) OC is input to floodplainseither through the decay of vegetation or the deposition offine OC-rich sediment This input of OC only determinesOC concentrations insofar as floodplain soils are capable ofstoring OC That storage is effectively determined by a bal-ance between processes that remove OC from floodplainsnamely respiration or erosion followed by respiration (Berheet al 2007) and processes that regulate OC availability tomicrobes namely the capability of the mineral fraction ofthe soil to sorb OC

OC sorption capacity reflects a few specific processes Al-though soil texture generally relates to the ability of OCto sorb to mineral grains and resulting OC availability soilchemistry also plays a strong and potentially dominant rolein regulating OC sorption capacity (Rasmussen et al 2018)Soil texture is largely determined by valley morphology ac-cording to our modeling (Table 1) placing valley morphol-ogy and resulting sediment transport dynamics (Gran andCzuba 2017 Wohl et al 2017b) as indirect controls on sorp-tion capacity

Respiration rate is largely determined by microbial activityand the availability of OC to microbes Erosion can rapidlyexpose soil OC to microbial respiration (Berhe et al 2007)whereas soils that reside in largely anoxic conditions can ex-hibit low rates of microbial respiration (Boye et al 2017)In addition soil mineralogy chemistry and redox condi-tions (Keiluweit et al 2017) can regulate microbial activity(eg andic soils may limit microbial respiration Matus et al2014) Our results suggesting that moisture controls OC con-tent support the idea that drier soils likely have higher ratesof microbial respiration of OC Moisture is a function of tex-ture valley bottom morphology and elevation (a proxy fortemperature) in our modeling (Table 1) Comparing flood-plain types in the wet basin we find that types with standingwater exhibit significantly higher soil moisture contents thanthose without standing water This indicates spatial variabil-ity in moisture content and likely microbial activity (Howardand Howard 1993) In our modeling this effect translates tospatial variability in OC concentration within floodplains andacross entire basins

OC export from soils refers specifically to leaching of dis-solved OC mainly from shallow soils While we do not di-rectly consider OC export in this study it likely introducesvariability in soil OC concentration and is regulated dom-inantly by hydrology (Aringgren et al 2014 Mcdowell andLikens 1988) soil pore water chemistry (including pH andionic strength Brooks et al 1999 Evans et al 2006) andredox conditions (Knorr 2013) OC export as DOC can actin conjunction with respiration to remove OC from soils andlike respiration rate is countered by sorption capacity

To summarize OC inputs regulated by the capacity ofsoils to store OC and suppress microbial respiration andOC export determine OC concentrations in floodplain soils(Fig 4) OC inputs to floodplain soils come from either au-tochthonous litter accumulation on the floodplain surface al-lochthonous wood deposition or allochthonous deposition offine OC-bearing sediments Our results from the mountain-ous basins we studied suggest that deposition of fine mate-rial in overbank flows is rare leading us to infer that au-tochthonous litter and allochthonous wood inputs to flood-plains dominate OC input in mountain rivers Where soils aremore moist microbial respiration is inhibited and more OCis stored Although soil texture is likely not a limiting factoron OC concentration in these floodplains finer-textured soilslikely have a higher sorption capacity retaining more of theOC input from decaying plant material Our results indicatethat geomorphic and hydrologic characteristics act as bound-ary conditions that regulate soil texture and moisture in turnregulating sorption capacity respiration rate and resultingOC concentrations in floodplain soils

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1112 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

5 Conclusions

We present floodplain soil OC concentration data from twodisparate watersheds to compare how interbasin variabilitybetween the two watersheds compares with intra-basin vari-ability in geomorphic and hydrologic characteristics in de-termining OC concentration Our results indicate that OCconcentration in mountain floodplain soils does not vary pre-dictably along a longitudinal gradient nor does it vary sub-stantially between basins with differing climatic and veg-etation characteristics Instead geomorphic and hydrologiccharacteristics such as valley bottom morphology and soilmoisture dominantly determine floodplain OC concentra-tion

In our study basins decaying litter and wood and not over-bank deposition of fine OC-bearing sediment is the mainsource of OC to floodplain soils It is unclear whether thatdecaying vegetation is dominated by autochthonous litter in-puts or transported downed wood In comparing our basinto other studied floodplain soils it seems that vegetation dy-namics play a strong role in determining OC concentrationswhen fine sediment is not regularly deposited on floodplainsurfaces However we suggest that floodplain soil character-istics set by geomorphic and hydrologic conditions regu-late how OC inputs translate to the spatial distribution of OCalong a river network

This implies that OC storage in floodplains likely cannotbe predicted using consistent downstream trends and thatmanagement prioritization designed to facilitate floodplainOC storage should be based on local geomorphic and hy-drologic process variability within each basin For instancemanagement to increase OC sequestration in floodplain soilswill likely be more effective where floodplains are uncon-fined and soils already experience high moisture conditionsfor much of the year Along these lines our results showthat modeling the floodplain biospheric OC pool to predictits response to warming and subsequent effects on climatebased on regional factors such as climate and net primaryproductivity likely misses the substantial interbasin variabil-ity in OC concentration and storage resulting from variabilityin valley bottom geometry and both geomorphic and hydro-logic processes (eg Doetterl et al 2015)

Although our results provide some insights the questionof whether OC stored in floodplain soil comes dominantlyfrom allochthonous versus autochthonous sources remainsopen Our results imply that more productive spatially het-erogeneous floodplains likely input more OC to soils Flood-plain OC concentration while mediated largely by moisturedynamics likely depends mainly on OC inputs from produc-tive riparian forests This implies that management of OCstorage in mountain river floodplains should focus on therestoration of riparian zones to maintain OC input to soil(eg Bullinger-Weber et al 2014) More detailed studiesin regions with varying sediment transport and hydrologicregimes are needed to determine what conditions favor au-

tochthonous versus allochthonous OC inputs but our resultssuggest that autochthonous sources dominate floodplain OCstorage in basins with relatively low rates of vertical accre-tion and high channelndashfloodplain connectivity that promotesfloodplain wetlands

Data availability Data to support the analyses presented here canbe found in Table S1 and in the Colorado State University DigitalRepository (Scott and Wohl 2018a)

Supplement The supplement related to this article is availableonline at httpsdoiorg105194esurf-6-1101-2018-supplement

Author contributions Field data collection was conducted pri-marily by DNS with assistance from EEW at some sites in theMiddle Fork Snoqualmie basin DNS performed data analysis anddrafted the initial version of the paper which was then edited andrewritten by DNS and EEW

Competing interests The authors declare that they have no con-flict of interest

Acknowledgements This work was funded by NSF grantEAR-1562713 We thank Ellen Daugherty for extensive assistancein fieldwork and discussion that improved the conceptual modelWe thank Katherine Lininger for stimulating discussion thatimproved the paper Detailed and constructive comments from twoanonymous reviewers and Robert Hilton improved the paper Wethank Sara Lowe for assistance in processing and analyzing soilsamples

Edited by Robert HiltonReviewed by two anonymous referees

References

Aringgren A M Buffam I Cooper D M Tiwari T Evans C Dand Laudon H Can the heterogeneity in stream dissolved or-ganic carbon be explained by contributing landscape elementsBiogeosciences 11 1199ndash1213 httpsdoiorg105194bg-11-1199-2014 2014

Appling A P Bernhardt E S and Stanford J A Flood-plain biogeochemical mosaics A multi-dimensional view ofalluvial soils J Geophys Res-Biogeo 119 1538ndash1553httpsdoiorg1010022013JG002543 2014

Aufdenkampe A K Mayorga E Raymond P A MelackJ M Doney S C Alin S R Aalto R E and YooK Riverine coupling of biogeochemical cycles between landoceans and atmosphere Front Ecol Environ 9 53ndash60httpsdoiorg101890100014 2011

Ballantyne A P Alden C B Miller J B Tans P P and WhiteJ W C Increase in observed net carbon dioxide uptake by

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1111

Figure 4 Conceptual model of physical processes that influenceOC concentration in floodplain soils based on this study and otherliterature Each box corresponds to a major factor that influencesOC concentration Colored text within each box denotes factors thatinfluence OC inputs sorption capacity OC export or respirationrate (note that some processes influence both OC export and respi-ration rate) As sorption capacity increases so does the OC capacityof the soil Conversely as the rate of respiration andor OC exportincreases the soil OC capacity decreases Floodplain soils can onlydevelop high concentrations of OC if there are high rates of OCinput However the capacity of the soils to store OC regulates thatinput and is determined by the competing influences of sorption ca-pacity and the combination of respiration rate and OC export Seethe text for further details

43 Conceptual model of soil OC concentration infloodplain soils

We present a conceptual model to summarize our results andplace them in the context of work examining the controlson OC storage in soils (Fig 4) OC is input to floodplainseither through the decay of vegetation or the deposition offine OC-rich sediment This input of OC only determinesOC concentrations insofar as floodplain soils are capable ofstoring OC That storage is effectively determined by a bal-ance between processes that remove OC from floodplainsnamely respiration or erosion followed by respiration (Berheet al 2007) and processes that regulate OC availability tomicrobes namely the capability of the mineral fraction ofthe soil to sorb OC

OC sorption capacity reflects a few specific processes Al-though soil texture generally relates to the ability of OCto sorb to mineral grains and resulting OC availability soilchemistry also plays a strong and potentially dominant rolein regulating OC sorption capacity (Rasmussen et al 2018)Soil texture is largely determined by valley morphology ac-cording to our modeling (Table 1) placing valley morphol-ogy and resulting sediment transport dynamics (Gran andCzuba 2017 Wohl et al 2017b) as indirect controls on sorp-tion capacity

Respiration rate is largely determined by microbial activityand the availability of OC to microbes Erosion can rapidlyexpose soil OC to microbial respiration (Berhe et al 2007)whereas soils that reside in largely anoxic conditions can ex-hibit low rates of microbial respiration (Boye et al 2017)In addition soil mineralogy chemistry and redox condi-tions (Keiluweit et al 2017) can regulate microbial activity(eg andic soils may limit microbial respiration Matus et al2014) Our results suggesting that moisture controls OC con-tent support the idea that drier soils likely have higher ratesof microbial respiration of OC Moisture is a function of tex-ture valley bottom morphology and elevation (a proxy fortemperature) in our modeling (Table 1) Comparing flood-plain types in the wet basin we find that types with standingwater exhibit significantly higher soil moisture contents thanthose without standing water This indicates spatial variabil-ity in moisture content and likely microbial activity (Howardand Howard 1993) In our modeling this effect translates tospatial variability in OC concentration within floodplains andacross entire basins

OC export from soils refers specifically to leaching of dis-solved OC mainly from shallow soils While we do not di-rectly consider OC export in this study it likely introducesvariability in soil OC concentration and is regulated dom-inantly by hydrology (Aringgren et al 2014 Mcdowell andLikens 1988) soil pore water chemistry (including pH andionic strength Brooks et al 1999 Evans et al 2006) andredox conditions (Knorr 2013) OC export as DOC can actin conjunction with respiration to remove OC from soils andlike respiration rate is countered by sorption capacity

To summarize OC inputs regulated by the capacity ofsoils to store OC and suppress microbial respiration andOC export determine OC concentrations in floodplain soils(Fig 4) OC inputs to floodplain soils come from either au-tochthonous litter accumulation on the floodplain surface al-lochthonous wood deposition or allochthonous deposition offine OC-bearing sediments Our results from the mountain-ous basins we studied suggest that deposition of fine mate-rial in overbank flows is rare leading us to infer that au-tochthonous litter and allochthonous wood inputs to flood-plains dominate OC input in mountain rivers Where soils aremore moist microbial respiration is inhibited and more OCis stored Although soil texture is likely not a limiting factoron OC concentration in these floodplains finer-textured soilslikely have a higher sorption capacity retaining more of theOC input from decaying plant material Our results indicatethat geomorphic and hydrologic characteristics act as bound-ary conditions that regulate soil texture and moisture in turnregulating sorption capacity respiration rate and resultingOC concentrations in floodplain soils

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1112 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

5 Conclusions

We present floodplain soil OC concentration data from twodisparate watersheds to compare how interbasin variabilitybetween the two watersheds compares with intra-basin vari-ability in geomorphic and hydrologic characteristics in de-termining OC concentration Our results indicate that OCconcentration in mountain floodplain soils does not vary pre-dictably along a longitudinal gradient nor does it vary sub-stantially between basins with differing climatic and veg-etation characteristics Instead geomorphic and hydrologiccharacteristics such as valley bottom morphology and soilmoisture dominantly determine floodplain OC concentra-tion

In our study basins decaying litter and wood and not over-bank deposition of fine OC-bearing sediment is the mainsource of OC to floodplain soils It is unclear whether thatdecaying vegetation is dominated by autochthonous litter in-puts or transported downed wood In comparing our basinto other studied floodplain soils it seems that vegetation dy-namics play a strong role in determining OC concentrationswhen fine sediment is not regularly deposited on floodplainsurfaces However we suggest that floodplain soil character-istics set by geomorphic and hydrologic conditions regu-late how OC inputs translate to the spatial distribution of OCalong a river network

This implies that OC storage in floodplains likely cannotbe predicted using consistent downstream trends and thatmanagement prioritization designed to facilitate floodplainOC storage should be based on local geomorphic and hy-drologic process variability within each basin For instancemanagement to increase OC sequestration in floodplain soilswill likely be more effective where floodplains are uncon-fined and soils already experience high moisture conditionsfor much of the year Along these lines our results showthat modeling the floodplain biospheric OC pool to predictits response to warming and subsequent effects on climatebased on regional factors such as climate and net primaryproductivity likely misses the substantial interbasin variabil-ity in OC concentration and storage resulting from variabilityin valley bottom geometry and both geomorphic and hydro-logic processes (eg Doetterl et al 2015)

Although our results provide some insights the questionof whether OC stored in floodplain soil comes dominantlyfrom allochthonous versus autochthonous sources remainsopen Our results imply that more productive spatially het-erogeneous floodplains likely input more OC to soils Flood-plain OC concentration while mediated largely by moisturedynamics likely depends mainly on OC inputs from produc-tive riparian forests This implies that management of OCstorage in mountain river floodplains should focus on therestoration of riparian zones to maintain OC input to soil(eg Bullinger-Weber et al 2014) More detailed studiesin regions with varying sediment transport and hydrologicregimes are needed to determine what conditions favor au-

tochthonous versus allochthonous OC inputs but our resultssuggest that autochthonous sources dominate floodplain OCstorage in basins with relatively low rates of vertical accre-tion and high channelndashfloodplain connectivity that promotesfloodplain wetlands

Data availability Data to support the analyses presented here canbe found in Table S1 and in the Colorado State University DigitalRepository (Scott and Wohl 2018a)

Supplement The supplement related to this article is availableonline at httpsdoiorg105194esurf-6-1101-2018-supplement

Author contributions Field data collection was conducted pri-marily by DNS with assistance from EEW at some sites in theMiddle Fork Snoqualmie basin DNS performed data analysis anddrafted the initial version of the paper which was then edited andrewritten by DNS and EEW

Competing interests The authors declare that they have no con-flict of interest

Acknowledgements This work was funded by NSF grantEAR-1562713 We thank Ellen Daugherty for extensive assistancein fieldwork and discussion that improved the conceptual modelWe thank Katherine Lininger for stimulating discussion thatimproved the paper Detailed and constructive comments from twoanonymous reviewers and Robert Hilton improved the paper Wethank Sara Lowe for assistance in processing and analyzing soilsamples

Edited by Robert HiltonReviewed by two anonymous referees

References

Aringgren A M Buffam I Cooper D M Tiwari T Evans C Dand Laudon H Can the heterogeneity in stream dissolved or-ganic carbon be explained by contributing landscape elementsBiogeosciences 11 1199ndash1213 httpsdoiorg105194bg-11-1199-2014 2014

Appling A P Bernhardt E S and Stanford J A Flood-plain biogeochemical mosaics A multi-dimensional view ofalluvial soils J Geophys Res-Biogeo 119 1538ndash1553httpsdoiorg1010022013JG002543 2014

Aufdenkampe A K Mayorga E Raymond P A MelackJ M Doney S C Alin S R Aalto R E and YooK Riverine coupling of biogeochemical cycles between landoceans and atmosphere Front Ecol Environ 9 53ndash60httpsdoiorg101890100014 2011

Ballantyne A P Alden C B Miller J B Tans P P and WhiteJ W C Increase in observed net carbon dioxide uptake by

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

1112 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

5 Conclusions

We present floodplain soil OC concentration data from twodisparate watersheds to compare how interbasin variabilitybetween the two watersheds compares with intra-basin vari-ability in geomorphic and hydrologic characteristics in de-termining OC concentration Our results indicate that OCconcentration in mountain floodplain soils does not vary pre-dictably along a longitudinal gradient nor does it vary sub-stantially between basins with differing climatic and veg-etation characteristics Instead geomorphic and hydrologiccharacteristics such as valley bottom morphology and soilmoisture dominantly determine floodplain OC concentra-tion

In our study basins decaying litter and wood and not over-bank deposition of fine OC-bearing sediment is the mainsource of OC to floodplain soils It is unclear whether thatdecaying vegetation is dominated by autochthonous litter in-puts or transported downed wood In comparing our basinto other studied floodplain soils it seems that vegetation dy-namics play a strong role in determining OC concentrationswhen fine sediment is not regularly deposited on floodplainsurfaces However we suggest that floodplain soil character-istics set by geomorphic and hydrologic conditions regu-late how OC inputs translate to the spatial distribution of OCalong a river network

This implies that OC storage in floodplains likely cannotbe predicted using consistent downstream trends and thatmanagement prioritization designed to facilitate floodplainOC storage should be based on local geomorphic and hy-drologic process variability within each basin For instancemanagement to increase OC sequestration in floodplain soilswill likely be more effective where floodplains are uncon-fined and soils already experience high moisture conditionsfor much of the year Along these lines our results showthat modeling the floodplain biospheric OC pool to predictits response to warming and subsequent effects on climatebased on regional factors such as climate and net primaryproductivity likely misses the substantial interbasin variabil-ity in OC concentration and storage resulting from variabilityin valley bottom geometry and both geomorphic and hydro-logic processes (eg Doetterl et al 2015)

Although our results provide some insights the questionof whether OC stored in floodplain soil comes dominantlyfrom allochthonous versus autochthonous sources remainsopen Our results imply that more productive spatially het-erogeneous floodplains likely input more OC to soils Flood-plain OC concentration while mediated largely by moisturedynamics likely depends mainly on OC inputs from produc-tive riparian forests This implies that management of OCstorage in mountain river floodplains should focus on therestoration of riparian zones to maintain OC input to soil(eg Bullinger-Weber et al 2014) More detailed studiesin regions with varying sediment transport and hydrologicregimes are needed to determine what conditions favor au-

tochthonous versus allochthonous OC inputs but our resultssuggest that autochthonous sources dominate floodplain OCstorage in basins with relatively low rates of vertical accre-tion and high channelndashfloodplain connectivity that promotesfloodplain wetlands

Data availability Data to support the analyses presented here canbe found in Table S1 and in the Colorado State University DigitalRepository (Scott and Wohl 2018a)

Supplement The supplement related to this article is availableonline at httpsdoiorg105194esurf-6-1101-2018-supplement

Author contributions Field data collection was conducted pri-marily by DNS with assistance from EEW at some sites in theMiddle Fork Snoqualmie basin DNS performed data analysis anddrafted the initial version of the paper which was then edited andrewritten by DNS and EEW

Competing interests The authors declare that they have no con-flict of interest

Acknowledgements This work was funded by NSF grantEAR-1562713 We thank Ellen Daugherty for extensive assistancein fieldwork and discussion that improved the conceptual modelWe thank Katherine Lininger for stimulating discussion thatimproved the paper Detailed and constructive comments from twoanonymous reviewers and Robert Hilton improved the paper Wethank Sara Lowe for assistance in processing and analyzing soilsamples

Edited by Robert HiltonReviewed by two anonymous referees

References

Aringgren A M Buffam I Cooper D M Tiwari T Evans C Dand Laudon H Can the heterogeneity in stream dissolved or-ganic carbon be explained by contributing landscape elementsBiogeosciences 11 1199ndash1213 httpsdoiorg105194bg-11-1199-2014 2014

Appling A P Bernhardt E S and Stanford J A Flood-plain biogeochemical mosaics A multi-dimensional view ofalluvial soils J Geophys Res-Biogeo 119 1538ndash1553httpsdoiorg1010022013JG002543 2014

Aufdenkampe A K Mayorga E Raymond P A MelackJ M Doney S C Alin S R Aalto R E and YooK Riverine coupling of biogeochemical cycles between landoceans and atmosphere Front Ecol Environ 9 53ndash60httpsdoiorg101890100014 2011

Ballantyne A P Alden C B Miller J B Tans P P and WhiteJ W C Increase in observed net carbon dioxide uptake by

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018

D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration 1113

land and oceans during the past 50 years Nature 488 70ndash73httpsdoiorg101038nature11299 2012

Bao H Kao S Lee T Zehetner F Huang J Chang Y LuJ and Lee J Distribution of organic carbon and lignin in soilsin a subtropical small mountainous river basin Geoderma 30681ndash88 httpsdoiorg101016jgeoderma201707011 2017

Battin T J Luyssaert S Kaplan L A Aufdenkampe A KRichter A and Tranvik L J The boundless carbon cycle NatGeosci 2 598ndash600 httpsdoiorg101038ngeo618 2009

Benfield E F Comparison of Litterfall Input to Streams J N AmBenthol Soc 16 104ndash108 httpsdoiorg10230714682421997

Berhe A A Harte J Harden J W and Torn M S The signifi-cance of the erosion-induced terrestrial carbon sink Bioscience57 337ndash346 httpsdoiorg101641B570408 2007

Blazejewski G A Stolt M H Gold A J Gurwick N andGroffman P M Spatial Distribution of Carbon in the Sub-surface of Riparian Zones Soil Sci Soc Am J 73 1733httpsdoiorg102136sssaj20070386 2009

Boye K Noeumll V Tfaily M M Bone S E Williams K HBargar J R and Fendorf S Thermodynamically controlledpreservation of organic carbon in floodplains Nat Geosci 10415ndash419 httpsdoiorg101038NGEO2940 2017

Brooks P D Mcknight D M and Bencala K E The rela-tionship between soil heterotrophic activity soil dissolved or-ganic carbon (DOC) leachate and catchment-scale DOC exportin headwater catchments Water Resour Res 35 1895ndash19021999

Bullinger-Weber G Le Bayon R-C C Theacutebault ASchlaepfer R and Guenat C Carbon storage and soilorganic matter stabilisation in near-natural restored and em-banked Swiss floodplains Geoderma 228ndash229 122ndash131httpsdoiorg101016jgeoderma201312029 2014

DrsquoElia A H Liles G C Viers J H and Smart D R Deepcarbon storage potential of buried floodplain soils Sci Rep-UK7 8181 httpsdoiorg101038s41598-017-06494-4 2017

Doetterl S Stevens A Six J Merckx R Van Oost K PintoM C Casanova-katny A Muntildeoz C Boudin M Venegas EZ and Boeckx P Soil carbon storage controlled by interac-tions between geochemistry and climate Nat Geosci 8 780ndash783 httpsdoiorg101038NGEO2516 2015

Evans C D Chapman P J Clark J M Monteith D T andCresser M S Alternative explanations for rising dissolved or-ganic carbon export from organic soils Glob Change Biol 122044ndash2053 httpsdoiorg101111j1365-2486200601241x2006

Fall P L Modern Pollen Spectra and Vegetation in the Wind RiverRange Wyoming USA Arctic Alpine Res 26 383ndash392 1994

Falloon P Jones C D Ades M and Paul K Direct soil mois-ture controls of future global soil carbon changes An impor-tant source of uncertainty Global Biogeochem Cy 25 1ndash14httpsdoiorg1010292010GB003938 2011

Gran K B and Czuba J A Sediment pulse evolution andthe role of network structure Geomorphology 277 17ndash30httpsdoiorg101016jgeomorph201512015 2017

Hanberry B B Kabrick J M and He H S Potential tree andsoil carbon storage in a major historical floodplain forest withdisrupted ecological function Perspect Plant Ecol 17 17ndash23httpsdoiorg101016jppees201412002 2015

Harvey J W and Gooseff M River corridor science Hy-drologic exchange and ecological consequences frombed-forms to basins Water Resour Res 51 6893ndash6922httpsdoiorg1010022015WR017617 2015

Hilton R G Galy A Hovius N Horng M J and Chen HEfficient transport of fossil organic carbon to the ocean by steepmountain rivers An orogenic carbon sequestration mechanismGeology 39 71ndash74 httpsdoiorg101130G313521 2011

Hoffmann T Glatzel S and Dikau R A carbonstorage perspective on alluvial sediment storage inthe Rhine catchment Geomorphology 108 127ndash137httpsdoiorg101016jgeomorph200711015 2009

Holm S A Simple Sequentially Rejective Multiple Test ProcedureScand J Stat 6 65ndash70 1979

Homer C G Dewitz J A Yang L Jin S Danielson P XianG Coulston J Herold N D Wickham J D and MegownK Completion of the 2011 National Land Cover Database forthe conterminous United States-Representing a decade of landcover change information Photogramm Eng Rem S 81 345ndash354 2015

Hoogsteen M J J Lantinga E A Bakker E J GrootJ C J and Tittonell P A Estimating soil organic car-bon through loss on ignition Effects of ignition conditionsand structural water loss Eur J Soil Sci 66 320ndash328httpsdoiorg101111ejss12224 2015

Houston D B Wildfires in Northern Yellowstone National ParkEcology 54 1111ndash1117 httpsdoiorg1023071935577 1973

Howard D M and Howard P J A Relationships be-tween CO2 evolution moisture content and temperature fora range of soil types Soil Biol Biochem 25 1537ndash1546httpsdoiorg1010160038-0717(93)90008-Y 1993

Jobbaacutegy E G and Jackson R B The vertical distribution ofsoil organic carbon and its relation to climate and vegeta-tion Ecol Appl 10 423ndash436 httpsdoiorg1018901051-0761(2000)010[0423TVDOSO]20CO2 2000

Kaiser K and Guggenberger G The role of DOM sorption tomineral surfaces in the preservation of organic matter in soilsOrg Geochem 31 711ndash725 httpsdoiorg101016S0146-6380(00)00046-2 2000

Keiluweit M Wanzek T Kleber M Nico P andFendorf S Anaerobic microsites have an unaccountedrole in soil carbon stabilization Nat Commun 8 1ndash8httpsdoiorg101038s41467-017-01406-6 2017

Knorr K-H DOC-dynamics in a small headwater catchmentas driven by redox fluctuations and hydrological flow pathsndash are DOC exports mediated by iron reductionoxidation cy-cles Biogeosciences 10 891ndash904 httpsdoiorg105194bg-10-891-2013 2013

Leithold E L Blair N E and Wegmann K W Source-to-sink sedimentary systems and global carbon burialA river runs through it Earth-Sci Rev 153 30ndash42httpsdoiorg101016jearscirev201510011 2016

Lininger K B Wohl E and Rose J R Geomorphic Controlson Floodplain Soil Organic Carbon in the Yukon Flats InteriorAlaska From Reach to River Basin Scales Water Resour Res54 1934ndash1951 httpsdoiorg1010022017WR022042 2018

Loope L L and Gruell G E The ecological role of fire in theJackson Hole area northwestern Wyoming Quatenary Res 3425ndash443 httpsdoiorg1010160033-5894(73)90007-0 1973

wwwearth-surf-dynamnet611012018 Earth Surf Dynam 6 1101ndash1114 2018

1114 D N Scott and E E Wohl Geomorphic regulation of floodplain soil organic carbon concentration

Matus F Rumpel C Neculman R Panichini Mand Mora M L Soil carbon storage and stabilisa-tion in andic soils A review Catena 120 102ndash110httpsdoiorg101016jcatena201404008 2014

Mcdowell W H and Likens G E Origin Composition and Fluxof Dissolved Organic Carbon in the Hubbard Brook Valley EcolMonogr 58 177ndash195 1988

Montgomery D R and Buffington J M Channel-reach morphol-ogy in mountain drainage basins Bull Geol Soc Am 109596ndash611 1997

Omengo F O Geeraert N Bouillon S and GoversG Deposition and fate of organic carbon in flood-plains along a tropical semiarid lowland river (TanaRiver Kenya) J Geophys Res-Biogeo 121 1131ndash1143httpsdoiorg1010022015JG003288 2016

Oregon State University PRISM Climate Group 2004Rasmussen C Heckman K Wieder W R Keiluweit M

Lawrence C R Berhe A A Blankinship J C Crow S EDruhan J L Hicks Pries C E Marin-Spiotta E Plante AF Schaumldel C Schimel J P Sierra C A Thompson A andWagai R Beyond clay towards an improved set of variablesfor predicting soil organic matter content Biogeochemistry 137297ndash306 httpsdoiorg101007s10533-018-0424-3 2018

Rathburn S Bennett G L Wohl E Briles C McElroy Band Sutfin N The fate of sediment wood and organic carboneroded during an extreme flood Colorado Front Range USAGeology 45 1ndash14 httpsdoiorg101130G389351 2017

R Core Team R A Language and Environment for Statistical Com-puting 2017

Ricker M C Stolt M H Donohue S W Blazejewski G Aand Zavada M S Soil Organic Carbon Pools in Riparian Land-scapes of Southern New England Soil Sci Soc Am J 771070ndash1079 httpsdoiorg102136sssaj20120297 2013

Schimel D S and Braswell B H The role of mid-latitude moun-tains in the carbon cycle Global perspective and a Western UScase study in Global Change and Mountain Regions editedby Huber U M Bugmann H K M and Reasoner M ASpringer 449ndash456 2005

Schmidt M W I I Torn M S Abiven S Dittmar T Guggen-berger G Janssens I A Kleber M Koumlgel-Knabner ILehmann J Manning D A C C Nannipieri P Rasse DP Weiner S and Trumbore S E Persistence of soil or-ganic matter as an ecosystem property Nature 478 49ndash56httpsdoiorg101038nature10386 2011

Scott D N and Wohl E Evaluating Carbon Storage on Sub-alpine Lake Deltas Earth Surf Proc Land 42 1472ndash1481httpsdoiorg101002esp4110 2017

Scott D and Wohl E Dataset for Geomorphic regula-tion of floodplain soil organic carbon concentration inwatersheds of the Rocky and Cascade Mountains USAhttpsdoiorg102567510217187762 2018a

Scott D N and Wohl E Natural and Anthropogenic Con-trols on Wood Loads in River Corridors of the Rocky Cas-cade and Olympic Mountains USA Water Resour Reshttpsdoiorgdoiorg1010292018WR022754 2018b

Smithwick E Harmon M E Remillard S M Acker SA and Franklin J F Potential upper bounds of car-bon stores in forests of the Pacific Northwest EcolAppl 12 1303ndash1317 httpsdoiorg1018901051-0761(2002)012[1303PUBOCS]20CO2 2002

Sparks D L Methods of Soil Analysis Part 3 Chemical Methodsedited by Sparks D L Page A L Helmke P A Loeppert RH Soltanpour P N Tabatabai M A Johnston C T SumberM E Bartels J M and Bingham J M Soil Science Societyof America Inc Madison Wisconsin 1996

Sun O J Campbell J Law B E and Wolf V Dynamicsof carbon stocks in soils and detritus across chronosequencesof different forest types in the Pacific Northwest USA GlobChange Biol 10 1470ndash1481 httpsdoiorg101111j1365-2486200400829x 2004

Sutfin N A and Wohl E Substantial soil organic carbon retentionalong floodplains of mountain streams J Geophys Res-Earth122 1325ndash1338 httpsdoiorg1010022016JF004004 2017

Sutfin N A Wohl E and Dwire K A Banking carbon A reviewof organic carbon storage and physical factors influencing re-tention in floodplains and riparian ecosystems Earth Surf ProcLand 60 38ndash60 httpsdoiorg101002esp3857 2016

Thien S J A flow diagram for teaching texture-by-feel analysisJ Agron Educ 8 54ndash55 1979

Wagenmakers E-J and Farrell S AIC model selectionusing Akaike weights Psychon B Rev 11 192ndash196httpsdoiorg103758BF03206482 2004

Wilcoxon F Individual Comparisons by Ranking Methods Bio-metrics Bull 1 80ndash83 httpsdoiorg1023073001946 1945

Wohl E Floodplains and wood Earth-Sci Rev 123 194ndash212httpsdoiorg101016jearscirev201304009 2013

Wohl E and Pfeiffer A Organic carbon storage in floodplainsoils of the US prairies River Res Appl 34 406ndash416httpsdoiorg101002rra3269 2018

Wohl E Dwire K Sutfin N Polvi L and Bazan R Mecha-nisms of carbon storage in mountainous headwater rivers NatCommun 3 1263 httpsdoiorg101038ncomms2274 2012

Wohl E Hall R O Lininger K B Sutfin N A andWalters D M Carbon dynamics of river corridors and theeffects of human alterations Ecol Monogr 87 379ndash409httpsdoiorg101002ecm1261 2017a

Wohl E Lininger K B and Scott D N River beads as a con-ceptual framework for building carbon storage and resilience toextreme climate events into river management Biogeochemistry1ndash19 httpsdoiorg101007s10533-017-0397-7 2017b

Zhao B Li Z Li P Xu G Gao H Cheng Y ChangE Yuan S Zhang Y and Feng Z Spatial distributionof soil organic carbon and its influencing factors under thecondition of ecological construction in a hilly-gully water-shed of the Loess Plateau China Geoderma 296 10ndash17httpsdoiorg101016jgeoderma201702010 2017

Earth Surf Dynam 6 1101ndash1114 2018 wwwearth-surf-dynamnet611012018