spatial patterns of soil organic carbon in the contiguous united states
TRANSCRIPT
Spatial Patterns of Soil Organic Carbon in the Contiguous United StatesJeffrey S. Kern*
ABSTRACTSpatial patterns and total amounts of soil organic C (SOC) are
important data for studies of soil productivity, soil hydraulic proper-ties, and the cycling of C-based greenhouse gases. This study evaluatedseveral approaches for characterizing SOC to determine their relativemerits. The first approach entailed grouping data from a global pedonSOC database by type of ecosystem, resulting in a total of 78.0 Pg ofC (Pg = 1015 g) to 1-m depth for the contiguous USA. In a secondapproach, a pedon database was aggregated using soil taxonomy,resulting in a total for the contiguous USA of 80.7 ± 18.6 Pg of Cwhen the great group SOC was spatially distributed with Major LandResource Areas (MLRAs) using the 1982 National Resource Inventory(NRI) and the Soil Interpretation Record databases. The third ap-proach used pedon and spatial data from a global soil map groupedby soil unit that resulted in 84.5 Pg of C for the contiguous USA.Although the ecosystem and soil taxonomic approaches resulted insimilar totals, the taxonomic approaches are recommended becausethey gave more realistic results in areas of Histosols, shallow soils,and soils with high rock fragment content. The ecosystem approachdid not give reh'able spatial patterns and is only useful for verybroad-scale work where precisely georeferenced data are not needed.Grouping data by great group provided more information than group-ing by order or suborder. The approach based on soil taxonomy isvery useful because it is based on the NRI statistical framework andit allows stratification by other NRI items, such as land use andvegetation.
SOIL ORGANIC MATTER, which is approximately 56%SOC (Nelson and Sommers, 1982), is a major factor
in plant nutrition (Stevenson, 1982), soil structure, com-pactability (Soane, 1990), and water-holding capacity(De Jong et al., 1983). In addition, soil is the largestterrestrial pool of C (Post et al., 1990) and must beconsidered for evaluating the flux of greenhouse gasesManTech Environmental Technology Inc., USEPA Environmental Re-search Lab., 200 SW 35th St., Corvallis, OR 97333. The information in thisdocument has been funded wholly by the U.S. Environmental ProtectionAgency (EPA) under Contract 68-C8-0006 to ManTech EnvironmentalTechnology, Inc. It has been subjected to the agency's peer and administra-tive review, and it has been approved as an EPA document. Received 1May 1992. "Corresponding author.
Published in Soil Sci. Soc. Am. J. 58:439-455 (1994).
between the terrestrial biosphere and the atmosphere.Spatial databases of SOC are needed, especially thosethat model the impact of agricultural tillage on SOC(Kern and Johnson, 1993), assess the impact of erosionon SOC pools (Kern, 1992), and provide informationfor biogeochemical modeling (Running and Coughlan,1988; Running and Gower, 1991). The purpose of thisstudy was to evaluate methods to estimate the spatialdistribution and amount of SOC in the contiguous USAto assess the relative merits of the methods. This isimportant for selecting which method to use for a givenapplication.
The SOC content of soil in the USA has been studiedin many locations; however, there have been few regionalor U.S. national-scale assessments. Franzmeier et al.(1985) characterized the SOC in the north central USAbased on a regional soil map, with laboratory data frompublished reports, theses, and unpublished data fromsoil survey activities. The map unit composition wasdetermined from published soil surveys, soil surveys inprogress, other inventories, and the expert judgment ofthe Technical Committee on Soil Survey. The upper 0.2 mof mineral soil ranged from 1.0 to 10.7 kg C m~2; theupper 1 m ranged from 2.3 to 19.2 kg C m~2.
Parton et al. (1987, 1989) used the CENTURY modelto analyze controlling factors of SOC accumulation inGreat Plains grasslands and modeled the geographic dis-tribution of SOC in the upper 20 cm. They concludedthat SOC content can be predicted from soil temperature,soil moisture, soil texture, plant lignin content, and Ninputs and that cool, moist, fine-textured soils have thegreatest SOC content. Burke et al. (1989) found thathigh SOC content was associated with high precipitation,high clay content, and low air temperature in grasslandsAbbreviations: SOC, soil organic C; NRI, National Resource Inventory;CDIC, Carbon Dioxide Information Center; NSSL-PD, pedon databasefrom the National Soil Survey Laboratory; MLRAs, major land resourceareas; GRID, Global Resource Information Database; SCS, Soil Conserva-tion Service; UN, United Nations; FAO, Food and Agriculture Organiza-tion; UNESCO, United Nations Educational, Scientific, and Cultural Orga-nization; CV, coefficient of variation; EROS, Earth Resources ObservationSatellite.
440 SOIL SCI. SOC. AM. J., VOL. 58, MARCH-APRIL 1994
of the U.S. Central Plains. In southern Great PlainsMollisols, clay content and, to a lesser degree, annualprecipitation were found to be factors of SOC accumula-tion (Nichols, 1984). In frigid and cryic soils of Montana,texture was not significantly correlated with SOC butcorrelation was found with elevation and precipitation(Sims and Nielsen, 1986). In a literature review, Oades(1988) concluded that water regimes and temperaturecontrolled the turnover of C in soil and the mineralizationof SOC was retarded by high clay content and basesaturation.
A study of the SOC in a northern hardwood forestecosystem (consisting predominantly of Spodosols) foundan average 16 kg C m~2 to a depth of 1 m or lessincluding O horizons (Huntington et al., 1988). The SOCof Spodosols in Florida was found to be somewhat lowerat 4.9 to 12.6 kg C m"2 to 1-m depth (Stone et al., 1993).Armentano and Menges (1986) summarized informationabout Histosol SOC content, which ranged from 113 kgC m~2 to 1-m depth for most of the contiguous USA to145 kg C m~2 for the southeastern USA.
One widely cited approach (Adams et al., 1990; Pren-tice and Fung, 1990; Jenkinson et al., 1991) for estimat-ing global SOC is the study by Post et al. (1982), whichused a large ( — 2700 pedons) database of SOC fromthroughout the world. The SOC data were grouped byHoldridge life zone (Holdridge, 1947) to derive meanSOC content from the surface to 1-m depth, and globalSOC was calculated by multiplying the SOC estimates
by the land area of the life zone groups. The SOC contentof wetlands was assumed to be 72.3 kg C m~2, and bogsor Histosols were not differentiated. A similar approachwas used by Schlesinger (1984) in which data for 117pedons were aggregated by ecosystem type. Buringh(1984) estimated SOC using areal estimates of soil ordersfrom the USDA-SCS and data from 400 pedons. Bohn(1976; 1982) used preliminary data from the UN FAO/UNESCO soil map of the world (FAO, 1974-1978) toestimate global SOC. Kimble et al. (1990) used the SCSNational Soil Survey Laboratory pedon database groupedby soil order and global areal estimates of orders tocalculate SOC globally for mineral soil. Eswaran et al.(1993) estimated global SOC by using the same pedondatabase as Kimble et al. (1990) with some additionsincluding Histosols, grouping the pedons by suborder,and using revised areal estimates of suborders.
METHODS AND MATERIALSEcosystem Complex Approach
The first approach closely follows the methods used byPost et al. (1982) except that ecosystem complexes (Olson etal., 1985), rather than Holdridge life zones (Holdridge, 1947),were used as a geographic base (Fig. 1, Table 1). This approachwas used because, as discussed above, the Post et al. (1982)study had been widely cited and used. The mean SOC perunit area was calculated for each ecosystem complex from apedon database assuming that each pedon equally representedthe ecosystem that it occurred in. The area and location of
S o u r c e s O l s o n e t a l . , 1985, USDA S o i l C o n s e r v a t i o nZ i n k e e t a l . , 1984 S e r v i c e
U N F o o d a n dA g r i c u l t u r e O r g a n .
G e o g r a p h i cP a t t e r n s
O l s o n E c o s y s t e m C o m p l e x e s llajor L a n d R e s o u r c e A r e a s S o i l M o p o f t h e War I d
M a p U n i tC o m p o s i t i o n
M a p C o m p o n e n tP r o p e r t i e s
O n e E c o s y s t e m p e rp o l y g o n
1 9 8 2 M o t i o n a l R e s o u r c e sI n v e n t o r y
A r e a e x t e n t o f c o m p o n e n t s
D e s c r i p t i v e L e g e n dA r e o I e x t e n t o f c o m p o n e n t s
r o c k f r o g m e n t ss o i l d e p t h
W o r l d i i d e S o i l O r g a n i cC and N D a t a b a s e
Z i n k e e t a l . , 19843700 p e d o n s u i e d w i t h :
s o i l C by * o I u m e
1982 S o i l I n t e r p r e t a t i o nR e c o r d D a t a b a s e
s o i l c l a s s i f i c a t i o ns o i l d e p t h
E x o m p I e P e d o nD e s c r i p t i o n s a n d D a t a2 5 5 p e d o n s u s e d w i t h :
s o i l Cb u l k d e n s i t y
S o i l S u r v e y L o b P e d o n D a t a b a s e2784 t o 3625 p e d o n s u s e d w i t h :
s o i l C b y w e i g h tb u l k d e n s i t y
Fig 1. Sources of input data for geographic databases of soil organic C.
KERN: SOIL ORGANIC CARBON SPATIAL PATTERNS 441
Table 1. Components of the soil organic C estimation approaches.Component Ecosystem approach Soil Taxonomy approach Soil map of the world approachSoil organic C
Digital map dataSpatial resolutionMinimum delineationMap unit composition
Rock fragmentsSoil depth
Bulk density
Pedon data, Zinke et al. (1984)4118 pedons global, 3700 usedEcosystem complexes, Olson et al. (1985)0.5° latitude/longitude= 360 x lO'ha1 ecosystem complex per unit
Included in calculations of pedonsIncluded in pedon calculations
Some measured, mostly estimated from ecosystem
Pedon data, SCS database2465 to 3625 pedons usedMajor Land Resource Areas, SCS (1981)1:7.5 million227 x 103 ha1982 National Resources Inventory
Texture modifier from NRISoil Interpretation Record for soil identi-
fied in NRIOnly measured data used
Pedon data255 pedons globalSoil map of the world1:5 million101 x 103 haIdentified dominant, associated
and inclusion soilsMap unit phaseMap unit phase
Some measured, mostly esti-mated from texture
each ecosystem complex was determined from a digital mapof ecosystem complexes. Global pedon data of SOC content(Zinke et al., 1984) were obtained from the CDIC. Thisdatabase contained information for 4118 pedons from a varietyof sources (2392 from North America), and the majority ofsamples were from the USA and central Eurasia. The datawere the basis of the Post et al. (1982) study with pedonsadded since that study. Many pedons are from work of thesenior compiler of the database, and the remainder come fromjournal articles, technical reports, theses, SCS Soil SurveyInvestigation Reports, and proceedings. Of the 4118 pedonswith SOC data, 3256 pedons (1990 from North America)had Holdridge classification and 3700 had Olson classification(2373 from North America). No information about soil classi-fication is included in the database. No quantification of dataquality is possible because of the wide variety of data sources.
The SOC determinations for the pedon database (Zinke etal., 1984) were made using a variety of methods includingwet combustion and loss on ignition, but the method for eachsample was not specified in the pedon database. Soil bulkdensity was measured using cores for 1800 of the 4118 pedons,and the remainder were estimated from regressions based ondepth, SOC content, and Olson ecosystem complex. The SOCcontent for each pedon was calculated by Zinke et al. (1984)on a volume basis to 1-m depth by multiplying mass SOC bybulk density and correcting for rock fragment content.
The geographic distribution of major world ecosystem com-plexes, as they existed in 1980 (Olson et al., 1985), wasobtained from the CDIC. The spatial resolution of the databaseis 0.5 by 0.5 ° latitude/longitude. The categories of ecosystemcomplexes in the spatial database were edited to agree withthe categories used in the pedon database. An advantage ofthe ecosystem complexes described by Olson et al. (1985) isthat, unlike the Holdridge system used by Post et al. (1982),it differentiates among bogs and other types of wetlands. TheHoldridge system (Holdridge, 1947) predicts plant formationsfrom climatic data and, thus, does not predict the extent ofwetlands, which is largely a function of landscape position.A map of the Olson ecosystem complexes of the contiguousUSA is presented in Fig. 2. The total SOC was calculated bysumming the mean of the pedon SOC content per unit areaaggregated by ecosystem complex times the area occupied.Water bodies were assigned zero SOC for all three approachesusing the perennial lakes, marshes, and reservoirs shown at1:2 million scale by the updated national atlas of the USA(U.S. Geological Survey, 1990).
Taxonomic Approach I: Soil TaxonomyThe Soil Taxonomy approach used the order, suborder, and
great group levels of classification (Soil Survey Staff, 1975)to aggregate pedon data. A national resource inventory was
used with a national land resources map to characterize thespatial distribution of soil taxonomic units (Fig. 1, Table 1).The SOC content of soil taxonomic units was calculated fromthe SCS pedon database from the NSSL-PD assuming thateach pedon equally represented the taxonomic unit. The SCSmaintains the NSSL-PD of soil samples analyzed at their labora-tories in Lincoln, NE, and Riverside, CA, as well as samplesanalyzed by the Agricultural Research Service (ARS) in Belts-ville, MD. Soil organic C was reported as percentage by weightdetermined by wet combustion with C^O?"2, and bulk densitywas determined using the clod method (Soil Survey Staff,1984). Data quality was enhanced by uniform methods amongthe laboratories and standardized sampling procedures, but noquantitative statement can be made about the data quality.
The database contained 15 789 pedons or sites sampled inthe contiguous USA at the time it was obtained, of which 6294pedons had SOC and measured bulk density data with eithergreat group classification or a series name that was not ataxadjunct. Of these 6294 pedons, 2258 listed the great groupclassification. The data for pedons lacking great group identifi-cation were merged by soil series with both the 1982 andcurrent SCS Soil Interpretation Record database to obtain theclassification (SCS, 1983). This data merge added great groupidentification for 3014 pedons or sites to yield a total of 5272pedons with potentially useful data.
Bulk density data were necessary to convert SOC measure-ments made on a weight basis to a volume basis. Only SCSdata with accompanying measured bulk density were usedbecause of the difficulty of estimating bulk density (Manriqueand Jones, 1991). Bulk density measured at 33 kPa moisturecontent was used because it more nearly represents field-moistconditions than that measured oven dry. In cases where oven-dry bulk density was available, but 33 kPa bulk density wasnot, the oven-dry bulk density was adjusted using the regressionequation based on NSSL-PD data:
p33 = (POD 0.880)+ 0.046(r2 = 0.89, n = 30 035 horizons)
where pas = bulk density at 33 kPa moisture, POD = bulkdensity oven dry.
The pedon dataset was analyzed by depth increments becauseit was recognized that many pedons were not sampled to 1-mdepth for each taxonomic unit. These depth increments wereadded together to construct an average SOC content to 1 m.The five depth intervals chosen were 0 to 8, 8 to 15, 15 to30, 30 to 70, and 70 to 100 cm. Soil organic C by volumewas calculated by multiplying SOC by weight, bulk density,and soil depth.
The data were grouped by soil order, soil suborder, and greatgroup for separate data analyses. There were not a sufficientnumber of samples to go to a more detailed category than
442 SOIL SCI. SOC. AM. J., VOL. 58, MARCH-APRIL 1994
1=1 Bogs and bog foods^3 Fonts or gross/scrub, cool^a Forms or gross scrub, form
Forns, grass/scrub, woods, warnI I II Forns, gross/scrub, foods, coolnun Forest, torn coniferOn Forest, cool deciduous
k \l Forest, iorm brood-leaved %ZM Soid/scrub/herbs or bore desertt\\\] Forest/ form complex, cool m-i Savanna and woodland, tropicolix A Forest/ form complex, warm H+H Scrub/iood/sovonno. Mediterranean£3^3 Grassland, miscellaneous K^XI Scrubland, subdesert /desert , hotE^ grassland, cool Scrublond. subdesert/desert, coolV7A Harsh, siampioods, littoral ES3 Taiga, northern or moritime/subolpine^^ Paddylonds and foods lili Taiga, main
WS Thorn/succulent foods, tropicalITOH TundraHUH Woods, hordfoods-coniters, cool
\K3 Woodland or scrubland, sparse
Fig. 2. Ecosystem complexes for the contiguous USA.
great group. Grouping the pedons by soil order resulted in3541, 3682, 3452, 2822, and 2505 pedons being used forthe five depth increments, respectively. When pedons weregrouped by suborder, the usable number of pedons was 3541,3681, 3449, 2819, and 2504. The results of the great groupanalysis were screened manually to eliminate outliers and bringthe CV to 80 or below. The 80% CV level was chosen becausepreliminary data analyses showed only a small number ofextreme values exceeded this level. The number of samplesused for the increments in the final analysis by great groupwere 3478, 3625, 3401, 2784, and 2465. There were fewersamples in the first depth increment than the second incrementbecause of the difficulty of obtaining bulk density measurementsfrom horizons near the soil surface.
The SOC content by great group was then geographicallydistributed using MLRAs as a map base, which are areas withsimilar patterns of soils, climate, water resources, and landuse that was published at a map scale of 1:7.5 million (SCS,1981). The areal extent of great groups in each MLRA wasdetermined using the 1982 NRI and the total area of eachMLRA. The MLRA map and 1982 NRI are known collectivelyas the National Soil Geographic Database (Reybold and TeS-
elle, 1989; Bliss, 1990). The dominant subgroups determinedfrom the 1982 NRI and the MLRA map are shown in Fig. 3.
The 1982 NRI is the most extensive inventory made in theUSA of soil, water, and related resources of nonfederallyowned land. The SCS coordinated data collection from 352 786primary sampling units, with three or less sampling pointseach, for the 1982 NRI. The data collected included soilcharacteristics, soil interpretations, land cover, land use, ero-sion, land management, conservation needs, and potential forconversion to farmland. There was a total of 841 860 samplingpoints in all counties of the USA (except for Alaska) and U.S.possessions in the Caribbean. Sites were selected to representthe MLRAs with a confidence limit of plus or minus onestandard deviation for attributes that comprise 10% of theMLRA (SCS, 1987). Thus, if corn (Zea mays L.) is producedon 10% of the area of a MLRA, then the theoretical confidencelimit is plus or minus one standard deviation. The analysesfor this study had a higher confidence limit because every(nonwater and nonurban) point of the database has soil as anattribute. The 1982 NRI is of limited use for characterizingwetlands because it does not contain information for millionsof acres that the NRI classified as water, but that might be
KERN: SOIL ORGANIC CARBON SPATIAL PATTERNS 443
E3 A q u a l f s B X e r p l f sS Boro l f s HD Arg idsSUdolfs nmOrthidsSUsta l fs ^Fluvents
Or then tsPsammentsHemlstsSapr i s t s
A q u e p t sOchrep tsUmbreptsAquol ls
Bo ro l l sUdollsUsto l lsXero l l s
AquodsOrthodsAquu l t sUdul ts
U s t e r t s
Fig. 3. Dominant soils from Soil. Taxonomy approach.
wetlands (Goebel and Dorsh, 1986). Each NRI point that isnot water or urban land has data for the surface texture andcoarse fragment content. In these analyses, the nonfederal landwithin the MLRAs is assumed to represent the federal landas well. The expansion factor for each point indicates the areathat the point represents.
Every point in the NRI has a number that links it to theSoil Interpretation Record that includes taxonomic informationand estimated soil properties based on the expert judgment ofpersonnel involved in soil survey work. Organic matter con-tents from the Soil Interpretation Record are of limited usefor estimating SOC because they are estimates that may notbe based on laboratory data, they are expressed as a range,and only organic matter data for surface horizons were re-corded. The soil properties for this project obtained fromthe Soil Interpretation Record are depth to bedrock and thetaxonomic classification. A special version of the Soil Interpre-tation Record from 1982 has been developed by the SCS foruse with the 1982 NRI data.
Some classes in soil taxonomy have changed since 1982,and the 1982 version of soil taxonomy is used in this studyfor the NRI points. For example, many soils that formerlywere in the Andept suborder of Inceptisols are now consideredAndisols. There have also been changes in the Oxisol classifi-
cations, because currently the suborder Orthox is not definedand previously there were no Udox (Soil Survey Staff, 1975,1990).
Great groups identified in the 1982 NRI that were notrepresented in the NSSL-PD were assigned the SOC contentof similar great groups with data. Miscellaneous land areas,considered nonsoil, were identified in the 1982 NRI. Miscella-neous land areas that were assigned SOC contents of 0.2 kgC m~2 to the 1-m depth were alluvial land, badlands, gulliedland, gypsum land, lava flows, playas, rubble land, salt flats,and scoria. The remaining miscellaneous areas were assigneda SOC value of 0.
An approximation of minimum and maximum SOC wascalculated by adding or subtracting one-half of the incrementSOC standard deviation from the mean SOC before geographi-cal distribution. The reasoning was that estimates based onthe 1982 NRI have a theoretical confidence limit of plus orminus one standard deviation, thus estimates based on the typeof soil that has a very large sample size should have a somewhatmore narrow confidence limit. The standard deviation of theSOC content for each combination of great group and incrementcould not be calculated because in some cases there was onlyone sample. The average CV (standard deviation/mean X 100)for each increment was also calculated by soil order. All
444 SOIL SCI. SOC. AM. J., VOL. 58, MARCH-APRIL 1994
A c r i s o l s E3 FluvisoisAndoso ls EEB GleysolsCombisols CD GreyzemsChernozems M Histosols
K a s t o n o z e m sLi thosolsLuvisolsPhaeozems
Planoso lsPodzo lsPodzo luv iso lsRegoso ls
So lone tzVer t i so lsX e r o s o l sY e r m o s o l s
Fig. 4. Dominant soils from the soil map of the world.
increments with missing standard deviations were then calcu-lated as the CV for that order multiplied by the SOC for theincrement, then divided by 100.
The SOC content for each depth increment by MLRA wascalculated by using the SOC content for that layer and addingor subtracting one-half of the standard deviation of the SOCto derive a minimum and maximum content. Water bodieswere assigned zero SOC using the perennial lakes, marshes,and reservoirs shown at 1:2 million scale by the updatednational atlas of the USA (U.S. Geological Survey, 1990).The SOC for each MLRA was calculated by
E SOCiayer x expansion factorE expansion factor
Taxonomic Approach II: Soil Map of the World
[2]
The FAO/UNESCO soil map of the world was chosen asa geographic layer because consistent data about the spatialdistribution was needed, not only for the USA but also globally.The soil map of the world is currently the most comprehensiveglobal-scale soil map available (Sombroek, 1989). The soilmap of the world was based on soil information available in
the 1960s to 1970s and is primarily a compilation of availablenational soil maps with additional field work by FAO staff.The soil map of the world, published at a scale of 1:5 million,was compiled from approximately 600 soil maps of differentscales and legends with 11000 other maps such as physiogra-phy, vegetation, climate, geology, and land use system, whichconstituted a system for correlating various taxonomic systems(FAO, 1974-1978). The texts that accompany the map sheetscontain information about map unit composition and generalproperties of the soil units.
The example pedons with laboratory data that accompanyall volumes of the soil map of the world (255 pedons) wereused to make SOC calculations grouped by soil unit (Fig. 1,Table 1). These pedons were presented as typical profiles, butit was cautioned that "one profile will not show the range ofsoil characteristics and climatic conditions within such broadunits" (FAO, 1974-1978). Thus, it was assumed that thesepedons represented broad modal soil characteristics. Missingmineral bulk density values were estimated, based on texture,from guidelines from the SCS (SCS, 1983). Organic horizonbulk density was assumed to be 0.15, which was derived byanalyzing NSSL-PD data for horizons of Histosols with >30%SOC by weight (75 measurements). The pedons were assumed
KERN: SOIL ORGANIC CARBON SPATIAL PATTERNS 445
Table 2. Soil organic C content of ecosystem complexes that occur in North America.Soil organic C
Ecosystem complex
Bogs and bog woodsCool farms or grass/scrubFarm, grass, or scrub with woods, warmFarms, grass/scrub with woods, coolForest, tropical/sub broad-leaved humidForest, warm coniferForest, cool deciduousForest, cool coniferForest, warm broad-leavedForest/farm complex, coolForest/farm complex, warmGrassland, miscellaneousGrassland, coolHeath, moorlandMarsh, swampwoods, and littoralPaddylands and associated woodsRangelands, coldSand/scrub/herbs or bare desertSavanna and woodland, tropicalScrub, lowScrub/wood/savanna, MediterraneanScrubland, hot subdesert/desertScrubland, cool/cold semidesert/desertTaiga, midcontinental southernTaiga, northern or maritime/subalpineTaiga, mainThorn/succulent woods, tropicalTundra, woodedTundra, non-woodedWarm farms or grass/scrubWoodland, seasonally dry tropicalWoodland or scrubland, sparseWoodlands, warm semiaridWoods, warm broad-leaved conifer mixWoods, cool hardwoods-conifers mixed
Mean
113.210.410.713.310.713.615.015.815.95.98.48.7
12.414.923.414.624.73.16.03.77.52.56.2
12.312.917.02.1
16.618.19.6
11.27.8
10.210.312.9
Min.
—— kg C m-2 1-m depth"1 ——50.42.41.60.10.80.32.20.41.95.73.81.40.65.85.33.86.00.50.41.20.60.31.10.41.42.91.02.40.91.62.32.88.91.22.9
Max.
183.624.034.146.297.545.161.5
349.4101.8
6.011.537.492.345.8
124.046.033.310.331.27.7
50.05.8
10.3142.534.770.45.4
66.260.645.224.627.111.547.540.0
cvt%43607666935783
132119
448777076
136925085764888737598
11180888989954850138576
ntno.
4114253
3873471098038723
936531113112
156914
259155
1796
635
20416322633
7657
t Coefficient of variation.$ Number of samples.
to be free of coarse fragments and extend to 100-cm depthunless they were Lithosols or indicated by the phase correction.The method of SOC analysis is the Walkley-Black method(FAO, 1974-1978). Soil units with missing SOC were assignedSOC from similar soil units in 52 cases. Lithosols were assigneddepths of 10 cm, soil with bedrock within 100-cm depth wereassigned 85 cm, and stony soils were assumed to contain 40%rock.
The soil geography of the soil map of the world for theUSA (Fig. 4) was based on the SCS general soil map (FAO,1974-1978) at 1:7.5 million map scale that was apparentlyalso a source of soil data for the MLRA map (SCS, 1981).The soil map of the world at 1:5 million map scale has greaterdetail than the MLRA map and also has soil phases indicatedas map overprints. The map has been digitized and is availablefrom the Global Resource Information Database program ofthe United Nations Environment Programme. The digital soilmap of the world data used in this study were obtained fromthe EROS Data Center of the U.S. Geological Survey. Thelegend of the soil map of the world has >5000 map units,which consist of soil units or associations of soil units. Somemap units are composed of 100% of the dominant soil; how-ever, more commonly at this scale of mapping, there areassociated soils and inclusions. Associated soils cover at least20% of the map unit area, and inclusions cover <20%. Phasesof map units were used to indicate indurated layers, hardbedrock at shallow depth, stoniness, salinity, or alkalinity(FAO, 1974-1978). The FAO estimated the composition ofeach map unit using the methodology developed in the Agroeco-
logical Zones Project (FAO, 1978). Water bodies were as-signed zero SOC using the perennial lakes, marshes, andreservoirs shown at 1:2 million scale by the updated nationalatlas of the USA (U.S. Geological Survey, 1990).
RESULTS AND DISCUSSIONSoil Organic Carbon by Ecosystem Complexes
The results of the average SOC content of the ecosys-tem complexes that occur in North America are listedin Table 2, and the spatial distribution for the contiguousUSA is shown in Fig. 5. One sample for bogs had 349.4kg C m~2, which is much higher than any pedon, andwas removed from the database. Bogs and bog woodshad the greatest SOC content at 113.2 kg C m~2 andtropical thorn-succulent woods had the least at 2.1 kgC m"2. Forests, ranging from 10.9 to 15.9 kg C irr2,tended to have greater SOC than grasslands, with 8.4to 12.4 kg C nr2. The variability of the SOC contentwithin the ecosystem complexes, as indicated by theCV was large (Table 2). Marshes, swamps, and littoralregions had the greatest variability (136% CV) with arange of 5.3 to 124.0 kg C m~2. Cool conifer forestshad a range of 0.4 to 349.4 kg C m~2. The total SOCfor the contiguous USA based on this approach was 78.0Pg. The CV of many ecosystem complexes was >80,and some CV values were MOO, which indicates that
446 SOIL SCI. SOC. AM. J., VOL. 58, MARCH-APRIL 1994
N
ki lometersI
0 900Albers coiic-tquol or«o
E3 0.1 to 4.0 Cm 4.1 to 8.0 EE3 8.1 to 12.0 Em 12.1 to 16.0 E
kg C3 16.1 to 20.0 m% 20.1 to 24.0 m3 24.1 to 28.0 m0 28.1 to 32.0 •
m ~ 2
1 32.1 to 36.0 11 36.1 to 40.0 i
40.1 to 44.0 I1 44.1 to 48.0
1 48.1 to 52.0i 52.1 to 56.0^ 56.1 to 113.2
Fig. 5. Soil organic C using ecosystem complexes to 1-m depth.
there are broad ranges of types of soil within the com-plexes. The minimum and maximum values were quitedifferent, in most cases, which suggests that ecosystemcomplexes are poor predictors of the amount of SOCcontent because of the great soil heterogeneity.
Soil Organic Carbon by Soil Taxonomy
The results for total SOC content and SOC densityfor depth intervals by soil order are listed in Table 3.Aridisols were the soil order with the lowest SOC, with
Table 3. Soil organic C (SOC) content by soil order.
Order
AlflsolsAridisolsEntisolsHistosolsInceptisolsMoUisolsOxisolsSpodosolsUltisolsVertisols
Totalsoct
kgm-2
7.05.66.9
84.311.712.111.516.77.0
10.3
0-8 cm
pihgm-2
2.11.271.68
10.873.092.752.673.852.151.99
CV§
%55686341565038416946
nlno.806385329
18404
10561867
36494
8-15 cm
Phgm-2
1.710.961.40
10.462.582.472.433.411.901.67
CV
%53565542655038476843
nno.78842433621
4341057
1992
402109
15-30 cm
Phg m"2
0.970.720.939.931.791.761.662.651.051.24
CV
%62566639715446517946
nno.68842830620
4439552089
387116
30-70 cm
Phgm"2
0.460.470.527.850.840.970.851.430.430.95
CV
%53628228986049787646
nno.60926324817
3597812161
36796
70-100 cm
Phgm-2
0.290.340.407.340.490.550.610.510.260.65
CV%58628535
1046359967753
nno.568199228
132916822169
34688
11-m depth.I Soil organic C density to 1-cm thickness.§ Coefficient of variation.1 Number of samples.
KERN: SOE, ORGANIC CARBON SPATIAL PATTERNS 447
5.6 kg C m~2 to the 1-m depth, and the greatest wereHistosols with 84.3 kg C nT2. The ranking of the SOCcontent of the soil orders was Aridisols < Entisols ~Ultisols = Alfisols < Vertisols < Oxisols < Inceptisols <Mollisols < Spodosols « Histosols. Histosols containednearly six times the SOC of Spodosols, which is thenext lowest order. Greater SOC density heterogeneitywas observed below 30-cm depth for most soils, asevidenced by higher CVs, and is due, in part, to smallersample sizes. Soil organic C content decreased withdepth for all soil orders.
Table 4 lists the SOC results for the suborders repre-sented in the database. Alfisol suborders had the SOCcontent trend of Udalfs « Ustalfs < Xeralfs < Aqualfs< Boralfs. The Argid and Orthid suborders of Aridisolshad nearly equal SOC content. The suborders of Entisolshad considerable variation, with a low of 4.9 kg C m~2
for the sandy Psamments and 10.5 kg C m~2 for wetAquents. Arents had consistently high SOC densitythroughout the profile, resulting in high total SOC content(but there was only one pedon represented). Sapristswere the only suborder of Histosols represented in the
data, with SOC of 86.9 kg C nr2. The Inceptisol subor-ders showed the trend Ochrepts < Tropepts < Aquepts~ Andepts. Aquepts and Andepts had a great deal ofvariability in SOC density for the lower horizons. Allsuborders of Mollisols had SOC content >10 kg C m~2
with the trend Ustolls < Xerolls < Borolls < Albolls <Udolls < Aquolls. Oxisols were not well representedin the database because they are not extensive in thecontiguous USA. Oxisols were not identified in the 1982NRI for the contiguous USA but are included here forcomparison. The trend for Oxisol suborders were Or-thoxs < Torroxs < Ustoxs < Udoxs. Spodosol subordersdisplayed a different trend than many others because theaquic suborder had less SOC than the orthic. The Aquodshad relatively high SOC density in the upper solum, butthe amount decreased more sharply with depth than inthe Orthods. The majority of the Ultisol samples wereUdults, which had considerably lower SOC content thanthe other suborders. The trend for Ultisol suborders wasUdults < Aquults < Ustults « Xerults < Humults. Thetrend for Vertisol suborders was Xererts < Uderts <Torrerts < Usterts.
Table 4. Soil organic C (SOC) content by soil suborder.
Suborder
AqualfsBoralfsUdalfsUstalfsXeralfsArgidsOrthidsAquentsArentsFluventsOrthentsPsammentsSapristsAndeptsAqueptsOchreptsTropeptsUmbreptsAlbollsAquollsBorollsUdollsUstollsXerollsOrthoxsTorroxsUdoxsUstoxsAquodsOrthodsAquultsHumultsUdultsUstultsXerultsTorrertsUdertsUstertsXererts
Totalsoct
kgm-2
7.98.26.36.57.35.55.9
10.514.47.46.14.9
86.918.513.59.0
13.318.613.715.713.414110.010.58.8
10.612.911.510.917.210.412.66.1
11.211.410.09.3
11.98.5
0-8 cm
Pthgm-2
2.172.891.971.362.371.21.362.231.891.461.71.65
10.873.573.692.573.454.573.443.593.272.842.082.622.221.852.892.864.123.823.313.611.832.664.592.262.352.001.55
CV§
%4945507056627346
NA#51686941634852434155464432465874
NA34213542604862544724324566
nlno.16676
34485
135230155371
9914151184849
244184516
123185198301231
41767
603025
29559
12213724
8-15 cm
Phgnr2
2.051.991.661.141.610.901.042.022.151.391.321.17
10.453.103.262.082.724.283.363.492.802.781.872.101.241.852.782.593.103.443.053.451.602.393.051.651.731.811.39
CV
%4754475756555646NA50574843785954484158474433425282
NA31214247804056464130333961
n
no.16285
32489
128249175381
9715347205254
269184115
125197192288239
31969
833328
32669
15224428
15-30 cm
Phgm-2
1.181.190.860.801.010.710.741.491.650.950.850.68
10.212.512.021.391.973.152.142.661.922.181.421.451.281.852.031.421.142.741.481.880.931.691.501.281.021.451.04
CV
%5955595166536063
NA66555539757353694735525235415638
NA40558548
1354760584828514157
nno.11876
27995
120256172341
10112446185658
27318381498
203119288232
41875
832923
32168
19214729
30-70 cm
Phgnr2
0.520.480.390.570.470.460.500.831.320.580.430.338.091.620.920.591.031.320.941.171.061.200.860.820.740.950.970.780.801.510.560.820.360.960.640.920.781.110.82
CV
%4754515252616399
NA6355552580
12458
1146222557146436717
NA39667575
1315450532835594045
nno.10570
24877
10916697311
819441164250
22115311281
164105239180
317
106
532726
30158
13204221
70-100 cm
Phgm-2
0.330.290.250.380.300.340.320.601.230.500.300.157.771.080.550.350.540.670.550.600.590.620.510.520.450.430.570.720.190.540.400.420.220.350.270.500.540.830.51
CV
%4653595366655683
NA7159612781
12468
1057050607057547055
NA35676093994665404932644064
nno.11068
2297388
12178301
878129123040
18312261272
129100217151
41796
633022
28077
14193718
11-m depth.t Soil organic C density to 1-cm thickness.§ Coefficient of variation.5 Number of samples.it Not applicable.
448 SOIL SCI. SOC. AM. J.( VOL. 58, MARCH-APRIL 1994
Table 5. Soil organic C (SOC) content of selected great groups.
Great group
OchraqualfsNatraqualfsEutroborallsHapludalfeNstrudfllfsHaplustalfsHaploxeralfsNatrixeralfsDurixeralfs
HaplargidsCamborthidsDurorthidsGypsiorthidsPaleorthids
HydraquentsHaplaquentsCryaquentsFluvaquentsXerofluventsUstifluventsUdifluventsCryofluventsTorrifluventsUdorthentsXerorthentsCryorthentsUstorthentsTorriorthentsCryopsammentsXeropsammentsUdipsammentsUstipsammentsTorripsamments
BorosapristsMedisaprists
Totalsoct
kgm-2
8.55.88.76.96.06.57.87.14.4
5.26.55.24.43.2
28.89.99.89.19.18.78.26.96.38.06.96.65.75.66.45.45.34.32.8
97.280.1
0-8 cm
Pihgm"2
2.371.693.142.081.451.342.582.291.30
1.171.641.000.821.04
2.171.74.052.211.951.571.541.931.232.411.983.321.471.272.591.322.021.171.13
10.7510.91
CV§
%
453242491771524863
6376526768
NA57
NA46495143NA5536306446626454615546
3344
"1no.
1061437
2282
49848
15
12251165
19
121
33129
321
45201013217610118
145
414
8-15 cm
phgm'2
2.201.522.031.791.421.211.791.410.89
0.891.080.790.811.16
2.171.632.352.031.681.341.591.521.182.151.461.271.291.091.211.061.451.060.49
9.4310.89
CV
%
423461452058502834
5455576158
NA53
NA48535740NA5137473755593243505123
4144
n
no.Alfisols
1051137
2162
57788
15Aridisols
14160161017
Entisols121
34139
291
45231013258299
10133
Histosols6
14
15-30 cm
Phgm"2
1.280.711.180.970.610.841.090.920.61
0.720.780.620.710.63
3.541.241.621.361.140.931.120.680.831.341.050.780.740.790.930.930.870.620.30
10.939.85
CV
%
55456053
NA#52673941
5556506089
3654
NA59546254NA7628394145632637436027
3841
n
no.
711034
1841
56748
16
14356151119
221
291210261
52128
122072
569
148
612
30-70 cm
Phgm"2
0.550.420.500.430.410.550.500.500.40
0.420.540.470.380.40
3.500.950.640.630.760.790.620.680.500.490.470.270.380.460.370.420.260.350.19
9.237.40
CV
%
46394947NA59524356
58536644NA
3842NA55498153NA6428644349565048206341
2125
n
no.
659
33165
14271
513
8336960
221
26116
17NA47779
1556886
105
610
70-100 cm
Phgm"2
0.350.240.340.290.450.370.290.320.10
0.290.360.360.210.23
2.090.600.600.480.570.670.561.100.420.230.290.270.350.290.210.190.120.110.14
9.596.48
CV
%
44535055NA57666328
555247399
2863NA56718378NA5947405856628350722952
1820
n
no.
668
35142
1415758
6732572
220
26126
211
478S3
145144684
57
InceptisolsDystrandeptsHydrandeptsEutrandeptsCryandeptsVitrandeptsAndaqueptsHalaqueptsHumitropeptsHaplumbreptsXerumbreptsFragiumbreptsCryumbrepts
ArgialbollsNatralbollsCalciaquollsHaplaquollsArgiaquollsNatraquollsHaploborollsArgiborollsCalciborollsNatriborollsHapludollsArgiudollsNatrustollsCalciustollsHaplustollsArgiustollsHaploxerollsArgixerollsNatrixerolls
27.727.618.013.28.7
32.56.2
34.920.118.813.411.0
14.09.2
20.415.514.78.3
12.711.510.99.9
14.613.812.210.610.39.7
10.610.46.9
4.735.725.242.712.364.002.544.984.575.213.683.67
3.532.134.663.453.373.493.033.002.232.912.822.853.092.062.102.012.602.692.27
52446447527978183652
537
54NA3948396853401938362934564545595940
1732
1114624
281025
151
1774273
424687
741221231
10513489986
5.025.595.102.341.414.151.014.594.314.793.683.40
3.461.914.583.383.362.212.642.452.342.822.782.782.531.971.871.832.052.142.14
54426245536663243652
545
57NA3749395947452852363036494439564943
1632
1218643
25925
Mollisols141
1776274
465596
74117
93194
13691996
3.824.503.632.011.163.560.674.013.512.762.552.34
2.201.373.172.702.431.23.72.59.76.29.17.21.91.58.42.38
1.471.390.59
552652545360813134716
79
34NA4352494753412653383333394538594958
1932
1318644
23825
131
1757194
545996
51676
3590
13091933
2.452.310.8010.732.920.373.741.491.320.890.52
0.960.701.761.141.070.351.020.890.860.671.241.180.970.950.880.840.830.810.51
576
NA777690593060484468
21NA455439456851
'4759553346395238627663
18318
11342
19723
111
1150163
4047106
51537
2573
1118060
5
1.641.050.540.810.413.160.342.240.740.640.190.05
0.560.430.590.620.600.250.590.500.490.350.730.540.430.480.570.460.550.490.24
5932
NA7074NA35
NA6653NANA
51NA8460362272665256624535396447696481
133158121
20411
1117
49132
333874
39596
1973
10168514
KERN: SOIL ORGANIC CARBON SPATIAL PATTERNS 449
Table 5. (cont.)
Great group
HaplorthoxsAcrorthoxsTorroxsHapludoxsKandiudoxsAcrudoxsEutrustoxs
HaplaquodsCryorthodsHaplorthodsFragiorthods
UmbraquultsPalehumultsHaplohumultsTropohumultsPaleudultsKandhapludultsKandiudultsHaplustultsHaploxerults
TorrertsChromudertsPelludertsPellustertsChromustertsPelloxerertsChromoxererts
Totalsoct
kgm-2
10.08.4
10.614.112.010.211.5
7.922.615.515.3
34.014.812.311.45.54.94.3
11.211.4
10.010.39.0
12.311.711.56.8
0-8 cm
Pthg m~2
3.152.121.852.713.183.022.86
3.523.903.982.44
8.523.843.435.841.621.681.232.664.59
2.262.192.392.081.961.831.39
CV§
%
58NANA4626NA21
31562969
453253
NA5156485447
24223234514482
nlno.
2114216
319365
25
191
78323159
n6
1312259
15
8-15 cm
Phgm"2
1.722.001.852.912.523.022.59
2.733.783.302.58
9.543.153.591.931.421.361.162.393.05
1.652.131.471.861.791.801.16
CV
%
50NANA404
NA21
51524329
343041NA5647424641
30232640403775
nno.
Oxisols2015216
Spodosols6
29513
Ultisols35
221
80332969
Vertisols157
1314301018
15-30 cm
Phg m"2
1.081.901.852.331.831.621.42
0.543.102.543.19
6.162.291.761.990.860.690.661.691.50
1.281.171.001.491.431.460.84
CV
%
31NANA487
NA55
19623427
625246NA5848475848
28336049362667
n
no.
3114217
228532
35
171
79312668
197
1217309
20
30-70 cm
Phg m~2
0.740.730.951.100.920.550.78
0.462.381.241.46
2.021.210.800.280.330.300.270.960.64
0.920.870.761.141.091.100.65
CV
%
25NANA4112
NA66
59647426
836140954753485328
35317040402251
nno.
211421
10
312374
34
202
73252758
136
1315278
13
70-100 cm
Phgm"2
0.580.320.430.690.440.350.72
0.190.900.450.34
1.080.430.410.460.230.130.160.350.27
0.500.630.520.860.800.750.36
CV
%
4766
NA2512
NA67
56857853
874449NA5455644049
32367535434362
nno.
2214219
315426
32
191
70212877
145
1314237
1111-m depth.$ Soil organic C density to 1-cm thickness.§ Coefficient of variation.1 Number of samples (0 for estimated values).# Not applicable.
The SOC for selected great groups arranged by soilorder is shown in Table 5. In a few cases, the depthincrement SOC density was estimated based on the valuesof surrounding increments, hi which case zero was en-tered for a number of samples. When the most centraltype of each Alfisol suborder was chosen, the trend wasHaploxeralfs < Haplustalfs < Hapludalfs < Ochraqualfs< Eutroboralfs. Alfisol great groups with indications ofan arid environment, such as Natraqualfs, Natrudalfs,Natrixeralfs, and Durixeralfs, had relatively low totalSOC content. The range of total SOC content for mostAridisol great groups (Table 5) was 5.2 (Durorthids andHaplargids) to 6.5 kg C m~2 (Camborthids). Paleorthidand Gypsiorthid SOC contents were particularly low, at3.2 and 4.4 kg C nr2.
The SOC content of Entisols is listed hi Table 5.Hydraquents, clayey soils of tidal marshes (Soil SurveyStaff, 1975), had the greatest SOC of all the Entisolgreat groups. The other aquic Entisols (Haplaquents,Cryaquents, and Fluvaquents) had moderately high SOCcontent. The arid Entisol great groups (Torrifluvents,Torriorthents, and Torripsamments) had lower SOC con-tent than did great groups with the same suborders. Thecold great groups, Cryofluvents and Cryorthents, hadsomewhat low SOC content, but the SOC was relatively
high for the cold, coarse-textured Cryopsamments. Greatgroups with xeric moisture regimes (Xerofluvents, Xer-orthents, and Xeropsamments) had moderately high SOCcontents. Great groups with ustic moisture regimes (Ust-orthents and Ustipsamments) had relatively low SOCexcept for the alluvial Ustifluvents. Udorthents, withudic moisture regimes had high SOC content, whereasother udic great groups (Udifluvents and Udipsamments)had moderate SOC compared with the same suborders.
There were only 18 samples to characterize Histosolgreat groups (Table 5), and they were all in the samesuborder. Borosaprists had lower SOC content than Med-isaprists, although they had similar SOC densities hi theupper 30 cm. Inceptisols showed a wide variation inSOC content by great group (Table 5). Inceptisol greatgroups influenced by volcanic parent material (Dystran-depts, Hydrandepts, Eutrandepts, and Cryandepts) hadhigh SOC content, except for the coarse-textured Vitran-depts. Andaquepts, which have both andic and hydricproperties, had particularly high SOC content. The salt-affected Halaquepts had the lowest SOC content of theaquic great groups. Of the Tropepts, the tropical Incepti-sols, only the Humitropepts great group had exceptionallyhigh SOC content. Great groups with umbric epipedons,which by definition have organic matter accumulations,
450 SOIL SCI. SOC. AM. J., VOL. 58, MARCH-APRIL 1994
had high SOC content (Haplumbrepts and Xerumbrepts)except for soils with fragipans and cryic temperatures(Fragiumbrepts and Cryumbrepts).
The haplic great groups of Mollisols (Table 5) showedthe following trend: Haplustolls « Haploxerolls <Haploborolls < Hapludolls < Haplaquolls. Mollisol greatgroups with argillic horizon development had the follow-ing trend: Argiustolls < Argixerolls < Argiborolls <Argiudolls « Argialbolls < Argiaquolls. Sodium-affected Mollisols ranked Natrixerolls < Natraquolls <Natrialbolls < Natriborolls < Natrustolls. The trend forcalcareous Mollisols was Calciustolls < Calciborolls <Calciaquolls < Calcixerolls.
The SOC contents of the Oxisol great group are pre-sented in Table 5. The greatest SOC content was forHapludoxs, but the lowest SOC content was for Haplor-thox, which has nearly the same definition in the olderversion of Soil Taxonomy. The SOC trend for the Oxisolswas Acrorthoxs < Haplorthoxs < Acrudoxs < Torroxs< Eutrustoxs < Kandiudoxs < Hapludoxs.
The Spodosol great groups (Table 5) all had relativelyhigh SOC content with the exception of the Haplaquods.The Cryorthods had considerably greater SOC than didthe other great groups. The Spodosols, in general, didnot have a marked decrease in SOC density hi the 30-to 70-cm increment, as was seen in many other greatgroups, apparently because of illuviation of organicmatter.
The great groups of Ultisols had quite a bit of variationin SOC content (Table 5). The poorly drained Um-braquults, which by definition have organic matter accu-mulations, had nearly three times the SOC content ofother great groups. The heavily weathered great groups(Paleudults, Kandhaphludults, and Kandiudults) all hadlow SOC contents of 5.5 kg C m~2 or less. The greatgroups that are defined by high SOC accumulationsshowed the trend Palehumults > Haplohumults > Tropo-humults. Haplustults and Haploxerults had nearly thesame SOC content as the humult great groups.
The SOC of the great groups of Vertisols is listed inTable 5. Chromusterts and Pellusterts, with ustic mois-ture regimes, had the highest SOC content. The Torrertsfrom arid climates, as well as the Chromuderts andPelluderts from humid regions, had similar SOC content.For the Xererts, Pelloxererts had nearly as much SOCcontent as Usterts but the Chromoxererts had half.
The CV of the SOC density estimates by soil order wasanalyzed to assign theoretical SOC standard deviations toincrements with only one sample. The CVs tended tobe <50 with an average of 45, 42, 46, 46, and 51 fordepth intervals of 0 to 8, 8 to 15, 15 to 30, 30 to 70,and 70 to 100 cm, respectively.
There was too much variation hi the kinds of soilsgrouped into soil orders, with the possible exception ofHistosols, to make them very useful for predicting SOCcontent. Grouping data by suborders provided a betterestimate than by orders because suborders give moreindications of climate, drainage, and coarse textures,which are important factors for SOC accumulation. Soilorganic C by great group provided even more detailabout soil-forming factors that affect SOC accumulation.
Great group classification provided only a limited indica-tion of soil texture (very sandy and some very clayeysoils) and temperature, which is included in family levelclassification. There was an insufficient number of pedonsto do these analyses at a family level of classification.
The tendency for soils from arid climates to have lowSOC content (<6 kg C m~2) is illustrated by the resultsfor great groups such as Torripsamments, Gypsiorthids,Haplargids, Naturargids, Torriorthents, and Paleargids.Coarse-textured soils had low SOC as shown by theTorrispamments, Ustipsamments, Quartzipsamments,Xeropsamments, and Udipsamments. Some highlyweathered soils have low SOC contents, such as thekandic great groups with their low-activity clays.
The influence of the soil moisture regimes varied fromorder to order. The general tendency for wet (aquic)and cold (boric and cryic) groupings to have greaterSOC was fairly consistent. Aquods, an exception, didnot have high amounts of SOC in their subsoils, whichmay be a function of the impedance to illuviation ofhumus by the poor drainage. Within similar great groups,udic groups tended to be greater than xeric and usticbut the relation of these two groups varied. In haplicAlfisols, xeric was less than ustic, but ustic was lessthan xeric hi haplic, argic, and calcic great groups ofMollisols. Great groups with the formative elementsumbr, hum, and umb tend to have higher SOC than othersimilar great groups. Soils with volcanic material tendto have high SOC content.
The Histosols stand out because they are, by definition,composed almost entirely of organic material. There arenot enough data for Histosol great groups to indicatewith certainty if Borosaprists consistently have moreSOC than Medisaprists.
Spatially distributing the great group SOC for thecontiguous USA results in a mean of 80.7 Pg of C; thespatial patterns are shown hi Fig. 6. The maximumSOC, as calculated by adding one-half of the standarddeviations of the increment SOC densities, is 99.3 Pg,and the minimum, by subtracting one-half of the SOCstandard deviations, resulted hi 62.1 Pg of C. The spatialpatterns of SOC as characterized by the Soil Taxonomyapproach (Fig. 6) showed the greatest SOC content hiareas of extensive Histosols and poorly drained soilssuch as in the northern Midwest, coastal Southeast andLouisiana, and southern Florida. The northcentral Mid-west, with its extensive Mollisols, had relatively highSOC content. The northern Northeast had high SOCcontent because of the extensive distribution of Spodo-sols. The western Pacific Northwest also had relativelyhigh amounts of SOC, probably because of high amountsof precipitation and cool temperatures.
Soil Organic Carbon from the Soil Map of the WorldThe SOC content of the soil units from the soil map
of the world are presented hi Table 6. The groupingswith the lowest amount of SOC were the arid Xerosolsand Yermosols. The Yermosols, which are the most aridgrouping, were extremely low hi SOC content, but therewas only one representative pedon of each. The Histo-
KERN: SOIL ORGANIC CARBON SPATIAL PATTERNS 451
kilometer!
0 900Alters coiic-iquol or««
E3 0.1 to 6.0E2 6.1 to 7.5ES! 7.6 to 9.0m 9.1 to 10.5
kgM 10.6 to 12.0^ 12.1 to 13.5EJ 13.6 to 15.0S3 15.1 to 16.5
C m" 2
ra 16.6 to 18.0 Wm 18.1 to 19.5 i• 19.6 to 21.0 IHDD 21.1 to 22.5
M 22.5 to 25.01 25.1 to 35.0I 35.1 to 57.6
Fig. 6. Soil organic C using Soil Taxonomy to 1-m depth.
sols, with up to 99.2 kg C m~2, had considerably moreSOC than any other grouping.
All groupings showed considerable variation, but gen-erally the tendency was for higher values for Humic,Gelic, and Gleyic soils. Acrisols tended to have lowSOC content. The Andosols, with the exception of Vitric,had high SOC content. The coarse-textured Arenosolsall tended to have low SOC content. The soils withdark surface horizons (Chernozems, Kastanozems, andPhaoezems) tended to have high SOC content. The poorlydrained Gleysols had moderate to high SOC content.Spatial application of the soil unit SOC contents resultedin 84.5 Pg of C for the contiguous USA (Fig. 7).
The soil units with the lowest SOC content were thearid Yermosols, which were much less than comparableAridisol great groups. The sandy Arenosols had lowSOC content, which is comparable to similar psammentgreat groups. Soil units from dry climates (Xerosols,Solonchaks, and Solonetz) tended to have low SOC con-tent except for soil units with mollic prefixes. ChromicVertisols had low SOC content, which is similar to the
great group Chromoxerts, but other chromic Vertisolgreat groups (Chromouderts and Chromousterts) hadtwice the SOC content of the soil map of the worldvalues.
The Luvisol SOC contents were comparable to Alfisolgreat groups with the exception of the much higherAlbic Luvisols. The great group Eutroboralf is similarin morphology to Albic Luvisols but had 8.7 kg C m"2
compared with 21.7 kg C m""2 for Albic Luvisols. Therewas not much difference among Kastanozems andPhaoezems (12-15 kg C m~2), which were comparableto the similar great groups of Ustoll and Udoll suborders.The Chernozem SOC content was slightly higher thancomparable Boroll great groups. The Podzols had asimilar range to the Spodosol great groups with theexception of the Gleyic Podzol, which was relativelyhigh.
The spatial distribution of SOC for the contiguousUSA from the soil map of the world data (Fig. 7) wassimilar to the mean Soil Taxonomy approach (Fig. 6)except that the soil map of the world data indicated
452 SOIL SCI. SOC. AM. J., VOL. 58, MARCH-APRIL 1994
Table 6. Soil organic C to 1-m depth calculated from the soil mapof the world.
Soil unit
AcrisolsGleyicFerricOrthicHumic
AndosolsVitricOchricHumicMollic
ArenosolsCambicFerralicLuvicAlbic
CambisolsCalcicEutricVerticFerralicChromicDystricHumicGleyicGelic
ChernozemsLuvicCalcicHaplic
FerralsolsXanthicAcricOrthicHumicRhodic
FluvisolsCalcaricDystricThiomicEutric
GleysolsDystricEutricCalcaricHumicGelicMollicOrthic
HistosolsEutricDystricGelic
KastanozemsHaplicCalcicLuvic
LuvisolsFerricCalcicPlinthicOrthicChromicGleyicAlbic
NitosolsDystricEutric
SoilCmean
kg C m-2
5.57.09.6
16.5
18.024.126.427.4
3.05.16.17.5
8.310.311.913.313.814.114.318.823.1
17.218.422.8
7.79.3
12.916.016.8
6.817.218.020.2
12.112.112.313.722.527.624.6
82.086.799.2
12.412.715.2
5.57.57.8Q 3O. J
10.212.921.7
9.19.2
cvt%
13382733
404
3823
321981
171911551111304330
13634
2328392530
30
4131
4121
3463
3
1716
372823
3316-214142
435
"tno
3366
4252
3221
452234322
233
22433
5124
4713212
361
543
4214741
33
SoilCSoil unit mean CVt nj
(cont.)Phaeozems
Haplic 12.4 16 5Luvic 13.8 35 3Gleyic 15.0 11 2Calcaric 15.2 - 1
PodzolsOrthic 13.7 82 3Humic 17.1 46 4Gleyic 19.0 46 3Placic 19.0 27 2Leptic 21.9 27 2
PodzoluvisolsGleyic 6.5 43 2Dystric 7.9 - 1Eutric 9.4 18 3Rankers 5.9 - 1
RegosolsDystric 5.2 - 1Eutric 6.3 11 3Calcaric 6.8 - 1Gelic 19.9 19 2
Rendizinas 15.7 32 3oOlODCnftKS
Gleyic 4.9 72 2Takyric 5.6 66 2Orthic 6.3 42 2Mollic 10.5 - 1
SolonetzOrthic 5.8 34 5Mollic 10.4 30 3Gleyic 12.9 - 1
V rfi 1vertisoisChromic 5.3 49 6PeUic 11.5 38 4
XerosolsHaplic 6.4 19 4Luvic 6.5 12 2Calcic 6.6 26 3
YermosolsGypsic 2.1 - 1Calcic 2.2 53 4Luvic 2.2 - 1Takyric 3.5 - i
Total 255
t Coefficient of variation.$ Number of samples.
greater SOC content in the northern Rockies, northernAppalachia, and mountainous parts of the West, andlower SOC was indicated in the upper Great Lakes andmuch of Southwest. The northern Rockies and Utah haveconsiderable federal land, and the 1982 NRI data maynot adequately describe these regions.
CONCLUSIONSSoil organic C can be sufficiently characterized on
very broad scales using ecosystem zones for aggregation;however, there are limitations to this approach for moredetailed work because this approach does not accountfor local variations hi parent materials (organic materials,coarse fragment content, and mineralogy) and soil depth.There was a great deal of heterogeneity of soil within
KERN: SOIL ORGANIC CARBON SPATIAL PATTERNS 453
ki lometers
0 900Albers conic-tquol or«o
E3 0.1E2 6.1E3 7.6S 9.1
totototo
6.07.59.010.5
m 10^ 12H 13m 15
.6 to
.1 to
.6 to
.1 to
kg (12.0 i13.5 E15.0 i16.5 i
: m'21 16.6^ 18.11 19.6I 21.1
totototo
18.0 &19.5 I21.0 B22.5
% 22.51 25.1i 35.1
tototo
25.035.086.7
Fig. 7. Soil organic C using the soil map of the world to 1-m depth.
ecosystem complexes, which makes this method of dataaggregation of limited use. The SOC content based onecosystems of the contiguous USA (78.0 Pg of C) waswithin one-half of the SOC standard deviation. Thus, atthat scale, this method was useful, but the spatial patternsmight not be useful for more detailed studies. The ecosys-tem approach might be useful for global studies whereit is not combined with other georeferenced data andwhere only a large region total is needed. The unreliabil-ity of the spatial patterns of SOC make this approachof limited use hi studies where georeferenced data, suchas climate or land use, are used. The pedon data onwhich the ecosystem complex approach is based containonly one value for SOC from 0- to 1-m depth, whichlimits its use when some other depth increment is needed.The pedon data for the other approaches can be subsetat any depth to be used for applications such as SOCloss from erosion (Kern, 1992) or changes in surfaceSOC from tillage (Kern and Johnson, 1993).
Soil classification is a much better framework fordata aggregation than is ecosystem complex. The soil
classification systems do not have narrow limits of SOCin their criteria, but soils with similar classification oftenhave similar factors that affect SOC accumulation. Soilorders and suborders were not very effective for aggregat-ing pedon data. Great groups were a better frameworkand were a good compromise for level of detail becauseaggregation at a more specific level, such as subgroup orfamily, would require extremely large pedon databases.
The SOC content of the contiguous USA using thesoil map of the world (84.5 Pg of C) came very close tothe mean SOC content as calculated by the Soil Taxonomyapproach (80.7 Pg of C). This similarity may be due,in part, to the common origin of the map data. Thetwo approaches yielded very similar SOC per unit areaestimates for similar types of soils. The spatial detail ofthe soil map of the world was slightly greater than thatof the Soil Taxonomy approach (MLRAs), which is anadvantage for some applications. An advantage of theSoil Taxonomy approach is that it is possible to estimateupper and lower limits using assumptions about the NRIdata. The Soil Taxonomy approach may be less reliable
454 SOIL SCI. SOC. AM. J., VOL. 58, MARCH-APRIL 1994
in areas of extensive federal land because the 1982 NRIexcludes federal land. This limitation may be minimizedif nonfederal land studied in the MLRAs was similar tofederal land. The Soil Taxonomy approach has manyadvantages because of additional data contained in the1982 NRI, such as land use. This is illustrated by astudy of the impact of conservation tillage on SOC (Kernand Johnson, 1993) where cropland SOC was estimatedby using points from the NRI that were cropland andoverlaying these points with a map of the areal extentof planted cropland. The 1982 NRI provides a greaternumber of map unit components than does the soil mapof the world, which is very useful for applications wherethe maximum or minimum values, rather than meanvalues, are critical. The SCS data approach providesa wealth of auxiliary information, such as land use,vegetation, and erosion from the 1982 NRI, which canbe used for data aggregation.
Improved statistical confidence will require largernumbers of sampled pedons to be included in the analy-ses. It would l>e very useful if existing databases, suchas the Zinke et al. (1984) dataset, could be updated byincluding the soil classification for samples that havealready been analyzed for SOC content. There are aconsiderable number of pedons or sites in the NSSL-PDwith missing series classification. The land use of pedonssampled should also be identified. There is not a lackof SOC measurements by weight, but there is a shortageof accompany big bulk density measurements. Measure-ment of bulk density in Histosols can be problematic,but it makes an enormous difference in the SOC estimatesby volume.
Better information about how the sampled pedons rep-resent the taxonomic unit used to aggregate the data isneeded. In this study, each pedon was assumed to equallyrepresent ecosystem complexes or taxonomic units, butthis was probably not the case. Information about whetherthe pedon is typical or atypical would enable differentweighting. Similarly, better locational data would permitfurther aggregating data by river basin, county, or othersubarea.
ACKNOWLEDGMENTSThis research was aided by the generous help of many
agencies. The USDA-SCS was very helpful in providing dataand advice. Henry Mount, of the Quality Assurance Staff,Lincoln, NE, helped with the acquisition of the Soil Interpreta-tion database. Benny R. Brasher and Steven L. Baird, of theNational Soil Survey Laboratory, Lincoln, NE, were veryhelpful in the acquisition and use of laboratory characterizationdatabase. Harvey Terpstra, of the Statistical Laboratory, IowaState University, provided the Soil Interpretation database andhelp with its use. Norm Bliss at the U.S. EROS Data Centerprovided the digital data for the United Nations soil map ofthe world, as well as helpful advice about linking together theNational Soil Geographic Database and the 1982 NationalResources Inventory.
KERN: SOIL ORGANIC CARBON SPATIAL PATTERNS 455