Stream Nitrogen and Phosphorus Loads Are Differentially Affectedby Storm Events and the Difference May Be Exacerbated byConservation TillagePatrick T Kellydagger William H RenwickDagger Lesley Knoll∥ and Michael J Vannisect

daggerDepartment of Biology Rhodes College Memphis Tennessee 38112 United StatesDaggerDepartment of Geography and sectDepartment of Biology Miami University Oxford Ohio 45056 United States∥Itasca Biological Station University of MinnesotaminusTwin Cities Lake Itasca Minnesota 56470 United States

S Supporting Information

ABSTRACT Storm events disproportionately mobilize dis-solved phosphorus (P) compared to nitrogen (N) contribu-ting to reduction in load NP In agricultural watershedsconservation tillage may lead to even further declines in loadNP due to dissolved P accumulation in the top soil layers Dueto an increase in this management activity we were interestedin the impacts of conservation tillage on N and P loads duringstorm events Using a 20 year data set of nutrient loads to ahypereutrophic reservoir we observed disproportionatelyincreasing P loads relative to base flow during storm eventswhereas N loads were proportional to discharge We alsoobserved a change in that relationship ie even greater P loadrelative to base flow with more conservation tillage in thewatershed This suggests conservation tillage may contribute to significantly reduced NP loads during storms with potentialimplications for the water quality of receiving water bodies

INTRODUCTION

Anthropogenic inputs of nitrogen (N) and phosphorus (P) toaquatic ecosystems can result in algal blooms hypoxia andother symptoms of eutrophication1 Furthermore nutrientloading to aquatic systems may be exacerbated by climatechange especially by increases in total precipitation as well asin the relative contribution of storm events with very heavyprecipitation Nutrient inputs during just a few storm eventsmay represent the bulk of annual nutrient load to manylakes2minus4 This may be especially true in agricultural watershedsas spring storms coincide with fertilizer application onto farmfields5 Because high nutrient runoff may lead to eutrophica-tion and the dominance of cyanobacteria and associatedcyanotoxins1 high nutrient loads during storm events may beof concern for lake managers interested in preserving waterquality and ecosystem services of the receiving water bodiesFurther understanding of how N and P loads respond to largeprecipitation events is important in developing more effectivemanagement plans especially in light of documented andpredicted further increases in large events with future climatechange6

While the absolute loads of N and P into lakes are ofconcern for eutrophication the ratio of N to P (NP) may alsobe important especially in structuring phytoplankton com-munity composition Low NP in lakes has been associatedwith greater cyanobacterial dominance of the phytoplankton

community as some cyanobacteria taxa can overcome N-limitation through N-fixation7minus11 Because low NP supplypotentially plays a role in creating harmful algal blooms and theassociated reduction in ecosystem services a better under-standing of what regulates load NP ratios is importantPrevious work pointed to the potential role of storms inmediating NP ratios1213 At present we know little about howNP loads from large storm events are changing as a result oflong-term climate shifts or land management practicesManagement techniques have been implemented at large

scale in agricultural watersheds in order to mitigate soilerosion and therefore the loads of sediment-associatednutrients to water bodies One such management techniqueis conservation tillage which includes reduced- and no-tillmethods and is defined as leaving at least 30 of the soilsurface covered by crop residue after tillage activities The useof this technique has increased in the Lake Erie watershed1415

as well as other agriculturally dominated landscapes16 Whiledemonstrated to be effective in reducing soil erosion andparticulate nutrient loads to water bodies after several yearsconservation tillage may actually increase dissolved nutrient

Received September 12 2018Revised January 4 2019Accepted March 12 2019Published March 12 2019

Article

pubsacsorgestCite This Environ Sci Technol 2019 53 5613minus5621

copy 2019 American Chemical Society 5613 DOI 101021acsest8b05152Environ Sci Technol 2019 53 5613minus5621

Dow

nloa

ded

by M

IAM

I UN

IV a

t 13

410

033

4 on

May

29

201

9fr

om h

ttps

pub

sacs

org

doi

10

1021

acs

est

8b0

5152

loads compared to conventionally tilled fields due to anaccumulation of soluble P in the top soil layers1718 Thisldquostratificationrdquo of P in soil may also mean that in addition toincreases in overall loads of dissolved P there may be greatermobilization of P during storm events and ultimately an evenmore disproportionate effect of high precipitation events ondelivery P to the lake In contrast dissolved N loads are moredriven by shallow subsurface flow3 In addition conservationtillage may affect N and P loads differently In particularstream nitrate concentrations seem to respond more slowlythan P concentrations to conservation tillage but after 10minus20years declines in stream nitrate concentrations can besubstantial1920 These differences in nutrient mobilizationdynamics especially in landscapes with widespread conserva-tion tillage may therefore significantly reduce NP load duringstorm eventsGiven the relative importance of storm events in

contributing to the annual nutrient budget of many lakes agreater understanding of how extreme events mobilizenutrients and how storm-mediated mobilization differs for Nversus P will be useful in developing management approachesfor highly impacted lakes under new precipitation regimes Inparticular we need to understand the implications for reducedeutrophication and cyanobacteria populations especially underscenarios of predicted increases in storms with climate changeAdditionally there has been a great deal of interest in theefficacy of conservation tillage as a management solution toreduce N and P loads However given that dissolved andparticulate nutrient fractions often respond differently toconservation tillage and that N and P also respond differ-entially we still know relatively little about how N and P andthe NP ratio may react to large storm events under increasedconservation tillage To achieve this understanding we used a20 year data set of daily N and P load estimates to Acton Lake

a hypereutrophic reservoir in southwest Ohio United Stateswhose watershed has undergone a large increase inconservation tillage over approximately the last 25 decadesThe goals of this study are to (1) compare how dissolved andtotal N and P loads are affected by storm events and (2)determine whether the relationships between storms andnutrient loads have changed over a time period (20 years)associated with increased conservation tillage

MATERIALS AND METHODS

Site Description Acton Lake is situated within the UpperFour Mile Creek (UFMC) watershed in southwestern Ohioand southeastern Indiana21 The UFMC watershed iscomprised mostly of agricultural land2223 (sim80 Figure 1)with poorly drained soils of high-lime glacial until capped withsilt loess As such tile draining is a common practice within thewatershed and it is estimated that at least 50 of cropland isartificially drained with that extent likely increasing but notrapidly24 The lake is a sim 250 ha eutrophic reservoir thattypically exhibits poor water quality including high nutrientconcentration turbidity and phytoplankton biomass25 Dy-namics of changes in constituent concentration with streamdischarge and season are variable and described in detail in ref21 The loads calculated in this study were from three inletstreams which collectively drain approximately 86 of thewatershed (Four Mile Creek Little Four Mile Creek andMarshallrsquos Branch) Detailed descriptions of the hydrology andnutrient concentrations in these streams have been describedelsewhere162026 The Acton Lake watershed was targeted forerosion control in the early 1990s and as such has undergoneincreases in the proportion of land in conservation tillagemainly between 1991 and 20011623 (Figure 1) Thesemanagement techniques led to substantial declines in bankerosion and concentrations of suspended solids and soluble

Figure 1Map of the Acton Lake watershed showing three inlet streams land use patterns and location of gauging stations Also shown is a stackedbarplot of trends in both tillage and crop type as a percentage of cropland in the watershed through time Solid line represents the total percentageof cropland in conservation tillage

Environmental Science amp Technology Article

DOI 101021acsest8b05152Environ Sci Technol 2019 53 5613minus5621

5614

reactive phosphorus (SRP) in the inlet streams standardizedfor discharge and season from 1994 to 2006162027 Howeverstreamwater quality trends have varied over time In particularalthough stream SRP concentrations declined over roughly thefirst decade of our study (1994minus2004) they actually increasedover approximately the second decade (2004minus2014) althoughthey are still lower than at the beginning of the 20 year periodIn contrast stream NO3minusN did not change much in the firstdecade but declined rapidly in the second decade20

Constituent concentrations in these streams vary by 2minus5orders of magnitude but averaged 63 μg Lminus1 for SRP and 5966μg Lminus1 for NO3minusN over our study period While there is someevidence that precipitation amount and intensity have changedover the last two decades in this region and will likely continueto change2829 there is has been no discernible trend in meanannual discharge that would suggest systematic increasingtrends in mean annual precipitation (Figure 2)Stream Sampling and Constituent Analysis Stream

sampling stations have been installed in all the inlet streamssince 1994 however sampling was initiated midseason in1994 therefore data included in our analyses begin in 1995Detailed sampling and analytical methods are presented in refs16 and 21 Stage is recorded in each of the three inlet streamsat 10 min intervals using pressure transducers installed instilling wells to monitor discharge and is converted todischarge (m secminus1) using rating curves Water samples fornutrient analyses were collected using ISCO pumping samplers(Teledyne ISCO Lincoln NE) at 6minus8 h intervals All sampleswere processed at high flow events while three or four samplesper week were processed during low flow periods (becauseconcentrations do not change rapidly during low flow)Samples were analyzed for nitrate plus nitrite-N (hereafterNO3) soluble reactive phosphorus (SRP) and suspendedsediments (SS) over all years of the study period (1995minus2014) From 1994 to 1998 nutrients were analyzed manuallyand from 1999 on they were measured on a Lachatautoanalyzer Throughout the study SRP was assayed withthe molybdenum blue method NO3minusN was measured with

second-derivative spectroscopy from 1994 to 1998 and withcadmium reduction method from 1999 and onward Tostandardize concentrations for methodological changes (in-cluding the shift from manual methods to the autoanalyzer)we analyzed numerous samples using both pre- and post-1999methods The methods gave similar results but neverthelessall concentrations were standardized to those based on theautoanalyzer methods as described in ref 16 Suspendedsediment concentrations were estimated by passing a knownvolume of streamwater through a preweighed glass fiber filter(Pall AE) drying to constant mass and reweighing21 Total Nand total P were determined as the sum of total dissolved N(or P) plus particulate N (or P) none of which were directlymeasured routinely Instead total dissolved N and P (TDNand TDP) were determined using stream-specific regressionwith NO3minusN versus TDN and SRP versus TDP (r2 gt 078 and081 respectively) on a subset of samples on which bothdissolved inorganic and total dissolved fractions weremeasured Similarly particulate N and P (PN and PP) weredetermined via stream-specific regression with suspendedsediment (SS) concentration on a subset of samples onwhich SS PN and PP were all measured (r2 gt 082 and 067for PN and PP respectively) For detailed description ofregression equations see Table 3 in ref 21 We used thisapproach because high concentrations of inorganic soilparticles in these streams (mainly in storm samples) interfereswith common methods for measurement of total P thereforewe opted to measure dissolved and particulate fractionsseparately and add them to obtain total N and P (TN andTP)30

Load Calculation We used the loadf lex package in the Rstatistical environment31 to estimate daily constituent loads toActon Lake We used the composite method to estimate loadsusing the loadComp() function to give average daily estimatesof load Total loads were then divided by 086 to account forfraction of the watershed that is ungauged More detaileddescription of the loadf lex package can be found in ref 32 andthat of Acton Lake load calculations is found in ref 26

Figure 2 Trend in average discharge and loads for NO3minusN total nitrogen (TN) SRP total phosphorus (TP) NO3SRP and TNTP Each pointrepresents a yearly average and the line represents linear regression

Environmental Science amp Technology Article

DOI 101021acsest8b05152Environ Sci Technol 2019 53 5613minus5621

5615

Data Analysis To determine how N and P loads areinfluenced by storms relative to base flow conditions we usedtwo techniques First we defined the 10 days from 1995 to2014 with highest daily discharge as the top 10 ldquoeventsrdquo Foreach event we found a corresponding amount of consecutivedays during lower flow conditions (what we define as ldquobaseflowrdquo usually JulyminusOctober) of the same year thatcumulatively approximated the discharge from the associatedevent Discharge during the ldquoeventrdquo day and the correspondingbase-flow period differed by lt1 We then compared loads ofNO3minusN SRP TN and TP during the event versus theassociated base flow period to get a perspective of thedifference between extreme events and base flow for eachconstituent (Figure 3) Our second technique was more

comprehensive and objective for each of the 20 years wecalculated the proportion of annual discharge (Qprop) and theproportion of annual load (loadprop) occurring on each dayThen we quantified the relationship between the Qprop versusthe loadprop for each year (n = 365 per year 366 in leap years)Using proportions of annual discharge and load allows for acomparison among years that may have differed in dischargeand among constituents that vary by orders of magnitude inload The slope of the Qprop versus loadprop relationship wasdetermined using a beta generalized additive model (gam)using a logit link function and including a smoothed predictorfor day of year to account for temporal autocorrelation Slope= 1 would indicate that load is simply proportional todischarge at the daily scale whereas slope gt 1 indicates adisproportionate increase in load during high-discharge eventsWe compared constituents by comparing the means of the

slopes for all constituents across all years of the time seriesThe years 1995 and 1997 were removed for TN and TP asmodel fits were poor defined here as having an R2 lt 085 Wealso looked at trends in these slopes over the 20 years to relateany changes that correspond to changes in the conservationtillage As the effects of conservation tillage are generally notimmediate151820 we tested the trends in slopes from the startof the data set as well as the years between 2000minus2014 as pastwork has observed changes in stream nutrient concentrationsaround that time Furthermore the increase in conservationtillage largely occurred in the 1990s the proportion ofcropland in conservation tillage varied from year to year after2000 but there is no net increasing or decreasing trend20 Toquantify these trends we used a MannminusKendall test afterchecking for significant temporal autocorrelation All analyseswere performed in R31 using the ldquomgcvrdquo package33 to run gammodels and the ldquotrendrdquo package34 for MannminusKendall tests

RESULTSPatterns in Discharge and Loads through Time

Annual discharge and constituent loads showed varyingpatterns throughout the time series with trends in differentdirections depending on forms of N and P (Figure 2) Annualmean discharge showed an oscillating pattern of dry and wetand increased slightly but not significantly from 1995 to 2014(P = 058 z = 055) Annual mean SRP and TP loads showeddifferent patterns as SRP load increased throughout the timeseries while TP declined subtly Nitrate-N and TN loads bothtended to decline throughout the time series Load ratios forNO3SRP and TNTP were also variable throughout the timeseries with a nonsignificant decline in mean NO3SRP from1995minus2014 (P = 014 z = minus146) and no trend in TNTPNone of the constituent loads displayed a statisticallysignificant trend from 1995minus2014 (P gt 005 in all cases)1996 was an outlier year in both discharge and all constituentloads as it was a very wet year with very high loads2123

Relationship between Extreme Events and Loads Ploads increased relatively more during storm events than Nloads (Figure 4) When comparing the 10 largest daily eventsthroughout the 20 year time series to comparable cumulativebase flow during the same year the mean ratios of event tobaseflow NO3 and TN loads were 11 and 13 respectively(Figure 4ac) In contrast the mean ratios of event SRP andTP to baseflow loads were 84 and 55 (Figure 4bd) Thedifferential responses of N versus P led to declines in loadNO3minusNSRP and TNTP during storm events MedianNO3minusNSRP by mass was sim13 for events compared to 51for base flow conditions (Figure 4e) Similarly median TNTPby mass was 9 during events compared to 25 during base flow(Figure 4f) Mean event daily discharge was slightly greaterthan cumulative base flow discharge (mean ratio of eventdischarge (Q) to baseflow Q = 1003 Figure 4g)Our second approach also showed that storms affect P loads

more than N loads The slopes of the relationship betweenloadprop versus Qprop were greater for P than for N for bothdissolved and total fractions (Figures 5 and 6) The averageslope of this relationship across all years (n = 20) was 151and 135 for SRP and TP (SE = 003 and 003) respectively Incontrast slopes for NO3minusN and TN were 095 and 100 (SE =001 and 000) respectively (Figures 5 and 6) The slopes near1 for N fractions show that loads are proportional to dischargewhereas for P loads increase more than proportionally withdischarge

Figure 3 Example of a comparison of SRP and NO3minusN loads for oneof the top 10 largest event days during the time series Red pointsrepresent the event and the blue bars represent the period of the yeargiving approximately equal cumulative discharge as the event

Environmental Science amp Technology Article

DOI 101021acsest8b05152Environ Sci Technol 2019 53 5613minus5621

5616

The slope of the relationship of loadprop versus Qpropincreased over time for both P forms (SRP z = 214 P =003 TP z = 268 P lt 001 Figure 7) Slopes for N declinedbut not significantly (NO3 z = minus009 P = 092 TN z =minus027 P = 079) These trends were different when looking at2000minus2014 with stronger trends in SRP compared to thosefor TP (SRP z = 277 P lt 001 TP z = 188 P = 006)Trends in N were similar (NO3 z = minus039 P = 069 TN z =0 P = 1)

DISCUSSIONLoads of N versus P may be an important influence on lakephytoplankton community composition as a low NP loadmay contribute to N-limitation and ultimately an increase inprevalence of N-fixing cyanobacteria Storm events tend togenerate large fractions of both N and P loads for a given yearbecause of their significant contribution to the annualdischarge21 (and for P because of increased concentrationsat high stream discharge) Therefore storms may be importantin regulating NP of the lake particularly during and after theevents We observed differences in the way N and P react tostorm events versus base flow conditions with larger stormsbringing in proportionally more P than N leading to sharply

declining NP during events This pattern held for bothdissolved and total fractions of N and P We found that therelationship between proportional discharge versus propor-tional load also changed over time during the 20 years butonly for P while N stayed relatively static Our data suggest alarger proportion of the annual P load may be currentlycontributed by large storm events than in the past presumablyas a response to increased conservation tillage in combinationwith or interacting with extensive tile drainage in thewatershed This indicates management efforts aimed atreducing nutrient load have the potential to lead to changesin the mobilization of P during storm events therebyincreasing SRP loads during storms and reducing load NP

Relationship between Storm Events and N and PLoads Large storm events bring significant pulses of bothdissolved and total P to Acton Lake This is characteristic ofmany agricultural watersheds as the primary flow path fordissolved P is overland flow leading to mobilization andtransport downstream in runoff35minus37 Particulate P is alsodriven exclusively by overland flow as it is coupled to sedimentfluxes36 As a consequence the proportion of annual P inputbudgets are almost entirely dominated by relatively few days ofhigh discharge events4 For example in Lake Mendota WI amajority of the P load for the year (sim74) was delivered overa span of 29 days with predicted annual P loads increasing asthe number of high discharge days increase438 While thesedynamics may in part be a function of timing of events andapplication of fertilizer on the landscape we observe significantdelivery of dissolved P during events throughout the calendaryear (Figure 3)In contrast to the relationship between discharge and P

loads daily N loads were proportional to daily discharge (iewe observed an approximate 11 relationship between Qpropand loadprop for N) Nitrate is more readily mobilized bysubsurface flow especially in areas with significant tiledrainage leading to a greater loadprop (compared to that forP) transported during base flow conditions31339 Additionallylarge storm events may lead to an initial decline in in-streamconcentrations of NO3 due to dilution with surface runoffbefore increasing later in the event2126 Therefore large pulsesof water do not increase N load to the same degree as PAdditionally our comparison of loads during events versusbaseflows for the top 10 events throughout the time seriesincludes a seasonal component for NO3 but no seasonal effectof SRP Nitrate concentrations in the in-flowing streams tendto be higher in the spring and decline in the middle of thesummer while SRP concentrations remain more seasonallyconstant2021 As such the relative difference between theresponses of NO3 compared to SRP is likely conservative asNO3 loading should be higher during events partially becauseNO3 concentrations are higher during the time of year whenlarge events typically occur (winter and spring) Ammonium(NH4minusN) responds in a manner similar to that for dissolved Pas it is mobilized by large storm events as opposed tosubsurface flow37 However NH4minusN is a minor contributor tothe total dissolved N load to Acton Lake (median sim1) and inmost agricultural watersheds in which fertilizers are heavilyapplied521

Particulate fractions of P and N may respond differently thandissolved fractions because in streams draining agriculturallandscapes particulate nutrient concentrations are drivenprimarily by concentrations of sediment However suspendedsediment loads to lakes are also heavily influenced by large

Figure 4 Boxplots for constituent loads (aminusd) load ratios (e and f)and discharge (g) for the top 10 events throughout the time seriesBoxes represent quantiles 5 25 50 75 and 95 of the 10 data pointsfor the base flow period and the event

Environmental Science amp Technology Article

DOI 101021acsest8b05152Environ Sci Technol 2019 53 5613minus5621

5617

storm events similar to dissolved P40minus42 As an exampleSharpley et al observed a majority of particulate P export (56minus93) in an agricultural watershed occurred during storm

flow43 The mean slopes of loadprop to Qprop for both particulateN and P loads to Acton Lake were 116 (SE = 006 and 004respectively) over the time series suggesting a stronger

Figure 5 Scatter plots of proportion of annual load vs the proportion of annual discharge for each year of the time series Black points representSRP load and blue points represent NO3minusN load (N = 365 or 366) Red dashed line is the 11 line Figure S1 shows only linear relationship ofthese data

Figure 6 Scatter plots of proportion of annual load vs the proportion of annual discharge for each year of the time series Black points represent TPload and blue points represent TN load (N = 365 or 366) Red dashed lines are the 11 line Figure S2 shows only linear relationship of these data

Environmental Science amp Technology Article

DOI 101021acsest8b05152Environ Sci Technol 2019 53 5613minus5621

5618

influence of storm events on particulate nutrient mobilizationcompared to NO3 but not SRP In general the composition ofTN is dominated by NO3 compared to TP which has morecontribution from particulate P21 Because of the combinationof these factors the relationship between loadprop and Qprop(slope) for TN is similar to but slightly above that for NO3minusNwhile the mean slope for TP is somewhat lower than that forSRP Additionally slopes between loadprop and Qprop for SSwere highly variable as sediment mobilization is related totiming and size of storm eventThe dynamics of both elements during base flow versus

storm flow contribute to rapidly declining load NP duringstorms Median load NO3SRP (mass) during the top 10events throughout the time series was approximately 13compared to 51 for base flow conditions Median load TNTP(mass) during those events was 9 compared 25 for base flowconditions This shift in NP is important as these eventsrepresent a significant contribution to the annual water andnutrient budgets of the lake On average the top 10 event daysthroughout the time series represent lt1 of the days 6 ofthe annual water load and 19 of the total annual SRP loadfor each of the given years in which the events took placeCumulative discharge over a storm event (which may occurover multiple days) may contribute up to 20 of the annualwater load As such these events while short in durationrepresent significant pulses of nutrients that may influence theultimate nutrient concentrations and ratios of NP in thelake444

Impact of Conservation Tillage Conservation tillage is amanagement technique designed to reduce soil erosion and byextension soil-bound N and P loading into lakes Howeverthere is also evidence that over long periods (10minus20 years)conservation tillage contributes to accumulation of P in the topsoil layers1845 especially in landscapes with significant Papplication from fertilizers184647 Tilling reduces this accumu-lation of P in the top soil layers distributes P into the soilmatrix and may disrupt preferential flow paths leading todeclines in P being leached from the soil compared toconservation tillage47 No-till approaches may also increase soilpore size allowing less water retention compared to conven-tionally tilled soil4849 Thus fields under conservation tillage

for long periods of time contribute more dissolved P duringrunoff events compared to conventionally tilled soils475152

While we do not have soil P data for the Acton Lakewatershed the trends in P concentrations in the in-flowingstreams and conservation tillage are consistent with those ofnearby Lake Erie watersheds20 where they have observed soilP stratification after years of conservation tillage compared toconventionally tilled fields18 The consistency of this patternacross study systems suggests the accumulation andstratification of soil P becomes more readily mobilized andtransported during storm events and highlights this as thelikely mechanism in the change in the relationship betweenQprop and loadprop (increase in slope) for SRP throughout ourtime series Potentially as a consequence of this changingrelationship between events and P concentration in runofftime series data suggest initial declines in dissolved Pconcentrations in in-flowing streams tend to reverse leadingto increases in SRP concentrations and loads after a period ofabout 5minus10 years151820 N loads particularly those of NO3seem to show patterns opposite to those of P loads in responseto conservation tillage For example in the Maumee River (amajor Lake Erie inflow) NO3 loads increased while TP loadsdecreased over about a 20 year period (1980minus2000) whenconservation tillage increased greatly However over thefollowing sim15 years NO3 load decreased rapidly as loads ofboth SRP and TP increased19 Similarly in the Acton Lakewatershed NO3 concentrations declined sharply in the seconddecade of our study period as SRP concentrations increased20

In addition to conservation tillage as being a potential driverof P mobilization tile drains in the watershed may move asignificant amount of P off the landscape Smith et al observedapproximately 50 of P losses in the St Joseph River innortheastern Indiana another agriculturally dominated water-shed occurred through tile drains50 While we do not have anydata on the exact extent of tile drainage in the Acton Lakewatershed (estimated to be about 50 of cropland) thisartificial drainage represents another potential avenue forsimilarly significant losses of P from the landscape and may inpart contribute to the patterns observed in our data setSubsequently the increase in conservation tillage in thewatershed may play a role in P export via tile drains asincreased soil pore size due to conservation tillage mayincrease P export via the macropores and ultimately tile drainsIn general subsurface and tile drain transport of NO3minusN ishigh compared to dissolved P41 so while tile drains may be animportant source of P transport from the landscape to the in-flowing streams they alone may not be a likely source of thechange in P mobilization observed in the time seriesThe increase in SRP concentration and load observed with

conversion to no-till techniques has been documented but inaddition we demonstrate here that (1) P loads increaseproportionally much more than N loads during storm eventsand (2) increases in conservation tillage and potentialinteraction with tile drainage systems may lead to a greaterstorm-derived mobilization of dissolved P than that ofdissolved N (based on the slopes of the relationship betweenQprop versus loadprop) We observe a change in the relationshipbetween Qprop and loadprop throughout the time seriessuggesting that over time an increasing fraction of the annualSRP load comes from events versus base flow This is especiallytrue when considering the trends from 2000 to 2014 whereprevious work has observed significant shifts in stream SRPconcentration20 The sharp increase in the slope of Qprop versus

Figure 7 Plots of the GAM slope for the relationship betweenproportion of annual Q and proportion of annual load for SRP andNO3 and TP and TN Each point represents the slope for the yearBlack points represent P and blue points represent N Years with poormodel fits for the relationship between proportion of annual Q andproportion of annual load not shown in the time series

Environmental Science amp Technology Article

DOI 101021acsest8b05152Environ Sci Technol 2019 53 5613minus5621

5619

loadprop for SRP after 2000 especially when compared to themodest increase in the slope for TP (which includes sedimentderived P) suggests this accumulation of dissolved P in theupper sediment layers is potentially contributing to an increasein mobilization during large storm events We do not see thissame pattern in NO3minusN or TN and instead see a subtledecline in the slope between loadprop and Qprop This hasimplications for load and ultimately NP in the lake with apossible shift of agricultural lakes to more N-limitationespecially during and after large storm events Indeed N-limitation of Acton Lake phytoplankton increased relative to Plimitation after sim201053 suggesting that shifts in phytoplank-ton nutrient status have already started to occur

ASSOCIATED CONTENTS Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI 101021acsest8b05152

linear model fits of loadprop versus Qprop for dissolved andtotal fractions of N and P (PDF)

AUTHOR INFORMATIONCorresponding AuthorE-mail kellyprhodeseduORCIDPatrick T Kelly 0000-0002-7279-5792NotesThe authors declare no competing financial interest

ACKNOWLEDGMENTSThis research was supported mainly by National ScienceFoundation awards 9318452 9726877 0235755 0743192and 1255159 Additional support was provided by the MiamiValley Resource Conservation and Development District

REFERENCES(1) Smith V H Tilman G D Nekola J C EutrophicationImpacts of excess nutrient inputs on freshwater marine and terrestrialecosystems Environ Pollut 1999 100 (1minus3) 179minus196(2) Gentry L E David M B Royer T V Mitchell C A StarksK M Phosphorus transport pathways to streams in tile-drainedagricultural watersheds J Environ Qual 2007 36 (2) 408minus415(3) Janke B D Finlay J C Hobbie S E Baker L A Sterner RW Nidzgorski D Wilson B N Contrasting influences of stormflowand baseflow pathways on nitrogen and phosphorus export from anurban watershed Biogeochemistry 2014 121 (1) 209minus228(4) Carpenter S R Booth E G Kucharik C J Lathrop R CExtreme daily loads role in annual phosphorus input into a northtemperate lake Aquat Sci 2015 77 (1) 71minus79(5) Royer T V David M B Gentry L E Timing of riverineexport of nitrate and phosphorus from agricultural watersheds inIllinois Implications for reducing nutrient loading to the MississippiRiver Environ Sci Technol 2006 40 (13) 4126minus4131(6) Melillo J M Richmond T C Yohe G W Eds ClimateChange Impacts in the United States The Third National ClimateAssessment US Global Change Research Program 2014(7) Havens K E James R T East T L Smith V H NP ratioslight limitation and cyanobacterial dominance in a subtropical lakeimpacted by non-point source nutrient pollution Environ Pollut2003 122 (3) 379minus390(8) Ferber L R Levine S N Lini A Livingston G P Docyanobacteria dominate in eutrophic lakes because they fixatmospheric nitrogen Freshwater Biol 2004 49 (6) 690minus708l

(9) No ges T Laugaste R No ges P To nno I Critical NP ratiofor cyanobacteria and N2-fixing species in the large shallow temperatelakes Peipsi and Vo rtsjarv North-East Europe Hydrobiologia 2008599 (1) 77minus86(10) Vrede T Ballantyne A Mille-Lindblom C Algesten GGudasz C Lindahl S Brunberg A K Effects of N P loading ratioson phytoplankton community composition primary production andN fixation in a eutrophic lake Freshwater Biol 2009 54 (2) 331minus344(11) Paerl H W Controlling cyanobacterial harmful blooms infreshwater ecosystems Microb Biotechnol 2017 10 (5) 1106minus1110(12) Correll D L Jordan T E Weller D E Transport of nitrogenand phosphorus from Rhode River watersheds during storm eventsWater Resour Res 1999 35 (8) 2513minus2521(13) Pionke H B Gburek W J Schnabel R R Sharpley A NElwinger G F Seasonal flow nutrient concentrations and loadingpatterns in stream flow draining an agricultural hill-land watershed JHydrol 1999 220 (1minus2) 62minus73(14) Richards R P Baker D B Eckert D J Trends in agriculturein the LEASEQ watersheds 1975minus1995 J Environ Qual 2002 31(1) 17minus24(15) Daloglu I Cho K H Scavia D Evaluating causes of trends inlong-term dissolved reactive phosphorus loads to Lake Erie EnvironSci Technol 2012 46 (19) 10660minus10666(16) Renwick W H Vanni M J Zhang Q Patton J Waterquality trends and changing agricultural practices in a Midwest USwatershed 1994minus2006 J Environ Qual 2008 37 (5) 1862minus1874(17) Zibilske L M Bradford J M Smart J R Conservation tillageinduced changes in organic carbon total nitrogen and availablephosphorus in a semi-arid alkaline subtropical soil Soil Tillage Res2002 66 (2) 153minus163(18) Baker D B Johnson L T Confesor R B Crumrine J PVertical Stratification of Soil Phosphorus as a Concern for DissolvedPhosphorus Runoff in the Lake Erie Basin J Environ Qual 2017 46(6) 1287minus1295(19) Stow C A Cha Y Johnson Y T Confesor R Richards RP Long-term and seasonal trend decomposition of Maumee Rivernutrient inputs to western Lake Erie Environ Sci Technol 2015 49(6) 3392minus3400(20) Renwick W H Vanni M J Fisher TJ Morris E L Streamnitrogen phosphorus and sediment concentrations show contrastinglong-term trends associated with agricultural change J Environ Qual2018 47 (6) 1513minus1521(21) Vanni M J Renwick W H Headworth J L Auch J DSchaus M H Dissolved and particulate nutrient flux from threeadjacent agricultural watersheds A five-year study Biogeochemistry2001 54 (1) 85minus114(22) Knoll L B Vanni M J Renwick W H Dittman E KGephart J A Temperate reservoirs are large carbon sinks and smallCO2 sources Results from high-resolution carbon budgets GlobalBiogeochem Cycles 2013 27 (10) 52minus64(23) Kelly P T Gonzalez M J Renwick W H Vanni M JIncreased light availability and nutrient cycling by fish provideresilience against reversing eutrophication in an agriculturallyimpacted reservoir Limnol Oceanogr 2018 63 (6) 2647minus2660(24) Price B J Preble County Soil and Watershed ConservationDistrict Eaton OH Personal communication December 2018(25) Vanni M J Arend K K Bremigan M T Bunnell D BGarvey J E Gonzalez M J Renwick W H Soranno P A SteinR A Linking landscapes and food webs Effects of omnivorous fishand watersheds on reservoir ecosystems BioScience 2005 55 (2)155minus167(26) Kelly P T Vanni M J Renwick W H Assessing uncertaintyin annual nitrogen phosphorus and suspended sediment loadestimates in three agricultural streams using a 21-year datasetEnviron Monit Assess 2018 190 91(27) Rakovan M T Renwick W H The role of sediment supply inchannel instability and stream restoration J Soil Water Conserv 201166 (1) 40minus50

Environmental Science amp Technology Article

DOI 101021acsest8b05152Environ Sci Technol 2019 53 5613minus5621

5620

(28) Kunkel K E Easterling D R Redmond K Hubbard KTemporal variations of extreme precipitation events in the UnitedStates 1985minus2000 Geophys Res Lett 2003 30 (17) 1900(29) Pryor S C Scavia D Downer C Gaden M Iverson LNordstrom R Patz J Robertson G P Ch 14 Midwest In ClimateChange Impacts in the United States the Third National ClimateAssessment Melillo J M Richmond T C Yohe G W Eds USGlobal Change Research Program 2014 pp 418minus440(30) Stainton M P Capel M P Armstron F A J The ChemicalAnalysis of Freshwater Fisheries and Environment Canada Fisheriesand Marine Service 1997 Vol 25(31) R Core Team R statistical environment for statistical computingR Foundation for Statistical Computing 2014(32) Appling A P Leon M C McDowell W H Reducing biasand uncertainty in watershed flux estimates the R package loadflexEcosphere 2015 6 (12) art269(33) Wood S N mgcv Mixed GAM computation vehicle withautomatic smoothness estimation R package version 18minus24 2017httpsCRANR-projectorgpackage=mgcv(34) Pohlert T trend Non-parametric trend tests and change-pointdetection R package version 110 2018 httpsCRANR-projectorgpackage=trend(35) Sharpley A N Depth of Surface Soil-runoff Interaction asAffected by Rainfall Soil Slope and Management Soil Sci Soc Am J1985 49 1010(36) Daniel T C Sharpley A N Lemunyon J L AgriculturalPhosphorus and Eutrophication A Symposium Overview J EnvironQual 1998 27 251minus257(37) Chen N Mo Q Kuo Y M Su Y Zhong YHydrochemical controls on reservoir nutrient and phytoplanktondynamics under storms Sci Total Environ 2018 619minus620 301minus310(38) Carpenter S R Booth E G Kucharik C J Extremeprecipitation and phosphorus loads from two agricultural watershedsLimnol Oceanogr 2018 63 (6) 1221minus1233(39) Peterjohn W T Correll D L Nutrient Dynamics in anAgricultural Watershed Observations on the Role of a RiparianForest Ecology 1984 65 (5) 1466minus1475(40) Kronvang B Laubel A Grant R Suspended sediment andparticulate phosphorus transport and delivery pathways in an arablecatchment Gelbaek Stream Denmark Hydrol Processes 1997 11 (6)627minus642(41) Borah D K Bera M Shaw S Water sediment nutrient andpesticide measurements in an agricultural watershed in Illinois duringstorm events Trans ASAE 2003 46 (3) 657minus674(42) Oeurng C Sauvage S Sanchez-Perez J M Dynamics ofsuspended sediment transport and yield in a large agriculturalcatchment southwest France Earth Surf Processes Landforms 201035 (11) 1289minus1301(43) Sharpley A N Kleinman P J A Heathwaite A L GburekW J Folmar G J Schmidt J R Phosphorus loss from anagricultural watershed as a function of storm size J Environ Qual2008 37 (2) 362minus368(44) Arbuckle K E Downing J A The influence of watershed landuse on lake N P in a predominantly agricultural landscape LimnolOceanogr 2001 46 (4) 970minus975(45) Butler J S Coale F J Phosphorus leaching in manure-amended Atlantic Coastal Plain soils J Environ Qual 2005 34 (1)370minus381(46) Baker J L Laflen J M Water quality consequences ofconservation tillage J Soil Water Conserv 1983 38 (3) 186minus193(47) Kleinman P J A Sharpley A N Saporito L S Buda A RBryant R B Application of manure to no-till soils Phosphorus lossesby sub-surface and surface pathways Nutr Cycling Agroecosyst 200984 (3) 215minus227(48) Arshad M A Franzluebbers A J Azooz R H Componentsof surface soil structure under conventional and no-tillage innorthwestern Canada Soil Tillage Res 1999 53 (1) 41minus47(49) Vogeler I Rogasik J Funder U Panten K Schnug EEffect of tillage systems and P-fertilization on soil physical and

chemical properties crop yield and nutrient uptake Soil Tillage Res2009 103 (1) 137minus143(50) Smith D R King K W Johnson L Francesconi WRichards P Baker D Sharpley A N Surface runoff and tiledrainage transport of phosphorus in the Midwestern United States JEnviron Qual 2015 44 (2) 495minus502(51) Inamdar S P Mostaghimi S McClellan K M Brannan MBMP impacts on sediment and nutrient yields from an agriculturalwatershed in the coastal plain region Trans ASAE 2001 44 (5)1191minus1200(52) Soileau J M Touchton J T Hajek B F Yoo K HSediment nitrogen and phosphorus runoff with conventional-andconservation-tillage cotton in a small watershed J Soil Water Conserv1994 49 (1) 82minus89(53) (a) Twombly et alin preparation (b) Vanni M Junpublisheddata

Environmental Science amp Technology Article

DOI 101021acsest8b05152Environ Sci Technol 2019 53 5613minus5621

5621