Exploring Functional Similarity in the Export ... - Western University

WATER RESOURCES RESEARCH, VOL. 34, NO. 11, PAGES 3079-3093, NOVEMBER

1998

Exploring functional similarity in the export of nitrate-N from forested catchments:A mechanistic modeling approach I. F. Creed • and L. E. Band 2 Departmentof Geography,Universityof Toronto, Toronto, Ontario, Canada

Abstract. Functionalsimilarityof catchmentsimpliesthat we are able to identify the combinationof processes that createsa similar responseof a specificcharacteristicof a catchment.We applied the conceptof functionalsimilarityto the export of NO•--N from catchmentssituatedwithin the Turkey Lakes Watershed,a temperateforest in central Ontario, Canada. Despite the homogeneousnature of the forest, these catchmentsexhibit substantialvariability in the concentrationsof NO•--N in dischargewaters,over both time and space.We hypothesizedthat functionalsimilarityin the export of NO•--N can be expressedas a function of topographiccomplexityas topographyregulatesboth the formationand flushingof NO•--N within the catchment.We testedthis hypothesisby exploringwhether topographicallybasedsimilarityindicesof the formation and flushingof NO•--N capture the observedexport of NO•--N over a set of topographicallydiverse catchments.For catchmentswith no elevatedbase concentrationsof NO•--N the similarity indicesexplainedup to 58% of the variancein the export of NO•--N. For catchmentswith elevatedbase concentrationsof NO•--N, predictionof the export of NO•--N may have been complicatedby the fact that hydrologywas governedby a two-componenttill, with an ablationtill overlyinga basaltill. While the similarityindicescapturedpeak NO•--N concentrationsexportedfrom shallowflow pathsemanatingfrom the ablationtill, they did not capturebaseNO•--N concentrationsexportedfrom deep flow pathsemanatingfrom the basaltill, emphasizingthe importanceof includingshallowand deep flow pathsin future similarityindices.The strengthof the similarityindicesis their potential ability to enableus to discriminatecatchmentsthat have visuallysimilar surfacecharacteristics but showdistinctNO•--N export responsesand, conversely,to group catchmentsthat have visuallydissimilarsurfacecharacteristics but are functionallysimilar.Furthermore,the similarityindicesprovide a potentiallypowerful method to scaleand generalizeNO•--N export responsesto other regions. of NO•--N in the dischargewaters, over both time and space [Nicolson,1988;Creed,1998].In our applicationof the concept Catchmentsare consideredto be functionallysimilar if dif- of functionalsimilarityto the exportof NO•--N we attemptto ferent combinationsof the processcontrolsresult in the same understandthe underlying mechanismsthat determine why response; the changein oneprocesscontrolis compensated for NO•--N exportfrom thesecatchmentsvariesso dramatically. by a changein anotherprocesscontrol.The uniquenessof the In a recentpaper [Creedet al., 1996]we derivedsimilarity concept of functional similarity is that first, it provides a indicesfor representingthe dominantprocesses controllingthe methodthat will enableus to state,a priori, the interactionsof exportof NO•--N. Theseindiceswere testedon a singlecatchprocessesthat will create the same responseand, second,it ment within the Turkey Lakes Watershed.The indicescapprovidesa mechanisticbasisfor interpretationof the observed tured processesregulating both the formation of NO•--N differencesin the response(E. M. O'Loughlin,unpublished within the catchmentsurfaceand the flushingof NO•--N from manuscript,1991). the catchmentto the receivingwaters.For the singlecatchment In thispaper,we applythe conceptof functionalsimilarityto the export of NO•--N from catchmentssituatedwithin the the indiceswere ableto capturethe NO•--N peaksin discharge waters under distinct catchment functional states. The indices TurkeyLakesWatershed(TLW), a temperateforestin central Ontario, Canada.Despite a narrowrangein the loadingof N capturedNO•--N peaksthat occurredduring springmelt or to the individualcatchments,there is a wide range in the loss autumn storms,when a risingwater table rose into previously of N from the forest to surfacewaters [Creed,1998]. These unsaturatedparts of a NO•--N-rich soil profile. The indices 1.

Introduction

alsocapturedN peaksthat occurredduringsummerdroughts, when a risingwater tableroseinto previouslysaturatedpartsof •Nowat Departments of PlantSciences andGeography, University a NO•--N-poor soil profile following a period of enhanced of Western Ontario, London, Ontario, Canada. rates of nitrification[Creedet al., 1996].Although the catch2Nowat Departmentof Geography, Universityof North Carolina, ment functional statesunder these two scenarioswere distinct,

catchments

exhibit

substantial

variation

in the concentrations

Chapel Hill.

the NO•--N exportswere similar. The similarity indiceswere based on the doctrine that topographyregulatesNO•--N formation and NO•--N flushing processes.Topographycan be an important regulator of nu-

Copyright1998 by the American GeophysicalUnion. Paper number 98WR02102. 0043-1397/98/98WR-02102509.00 3079

3080

CREED AND BAND: CATCHMENT NO•--N EXPORT BEHAVIOR

trient cyclingand routing dynamicsas it influences(1) the formation of nutrient sink areas (areas that retain N) and sourceareas(areasthat releaseN to adjacentwaters)within the catchment,(2) the hydrologicflushingof thesenutrient sourceand sink areas,and (3) the exportof nutrientsaccompanyingthis formationand flushing[Murdochand Stoddard, 1992;Stoddard,1994;Hornbergeret al., 1994;Boyeret al., 1995, 1996; Creedet al., 1996]. In this paper,we hypothesizethat NO•--N exportis a function of the topographiccomplexityof the catchmentas topographyregulatesboth NO•--N formationprocesses and NO•--N flushingprocesses within the catchment.To test this hypothesis,we explorewhether the similarityindicescapturethe observedvariation in the export of NO•--N over a set of catchmentsrepresentingthe rangein topographiccomplexitywithin the watershed.

2.

Test Area

lands of central Ontario, Canada. Situated on the Canadian

Shield,this 10.5km2 watershedcontainsa headwaterchainof lakes fed by small catchmentsthat ultimately drain into BatchawanaBay on the eastern shorelineof Lake Superior (Figure 1). The TLW was selectedas the test area becausepotential sourcesof variation in the export of NO•--N were few. The selectioncriteria mandatedthat the test area should(1) be within a singleclimaticregion; (2) be within a singleforest region;(3) containa forestdominatedby a singlespecies;(4) contain a forest within a stable stage of development;(5) contain a substantialrange in the hypothesizedtopographic controlson the exportof NO•--N; and (6) havesufficientdata to testthe hypotheses. The TLW is locatedwithin the Superior ClimaticRegion (criterion1) and the AlgomaSectionof the Great Lakes-St.LawrenceForest Region (criterion 2). The forestis dominatedby sugarmaple (Acersaccharum Marsh.), with smallstandsof yellowbirch(Betulaalleghaniensis Britton) and other minor species(criterion3). The forestis an unevenaged(120-180 years)matureforestthat with the exceptionof a light harvestof yellow birch over 35 years ago, remains essentiallyundisturbed(criterion4). Given the homogeneous character of the catchments in terms of criteria 1-4, we were

able to focus on the topographiceffectson the export of' NO•--N. The TLW topographyis complex.Controlledby bedrock, the undulatingtopographyresultsin small, sharplydefined subwatersheds for each lake and providessubstantial heterogeneityof topographicfeaturesboth within and among the catchments(criterion5). has served as a contin-

Methods

In our conceptualmodel (Figure 2) the exportof NO•--N is a functionof the hydrologicflushingof NO•--N [Hornberger et al., 1994; Creedet al., 1996]. On the basisof this conceptual model, when the water table is low, N accumulateswithin the

soil profile, resultingin small export of NO•--N to adjacent waters.As the water table rises,it "flushes"the soilprofile, and NO•--N within the soil profile is availablefor export.As the water table reaches the soil surface, the water table flushesthe

organicallyenrichedsurfacelayersof the soilprofile,resulting in potentiallylargeexportof NO•--N to adjacentwaters,unless denitrificationprocesses becomesignificant.To test our conceptualmodel,we adopteda "hybrid"modelingapproachin which we simulatedindicesthat representthe hypothesized controlson the exportof NO•--N andrelatedtheseindicesto the observedexportof NO•--N. 3.1.

The test area is the Turkey Lakes Watershed(TLW), an old-growthsugarmaple forest locatedin the Algoma High-

Since the autumn of 1980 the TLW

3.

RItESSys Simulation System

The RegionalHydroEcologicalSimulationSystem(RHESSys)(Figure3) is a distributedmodeldesignedto computethe distributionof biogeochemicalstores and the fluxeswithin catchments.The distributedmodel capturesthe variability in biogeochemicalprocessesby subdividingthe catchmentinto hillslopes. Within eachhillslopea probabilitydistributionfunction (pdf) of catenarypositionsis calculated.Catenarypositionsdiffer in their propertiesbecauseof downslopetransport processes that include(1) differentialdrainageconditions,(2) differentialtransportand depositionof suspendedmaterials, and(3) differentialleaching,translocation andredeposition of solublematerials[Hall and Olson,1991]. Catchmentbiogeochemical fluxes are determined through simulation at two scales:the hillslopescale,where simulationover the pdf of catenarypositionsconsidersthe variabilityin biogeochemical fluxesrelatedto the nonlinearimpactsof wetter and drier parts of the hillslope,and the catchmentscale,where simulation over the populationof hillslopesconsidersthe variability in biogeochemicalfluxes related to differencesin the average slope,aspect,or elevationof the hillslopeswithin the catchment. For a more detaileddescriptionof RHESSysthe reader is referred to Running and Coughlan [1988], Running and Gower[1991],RunningandHunt [1993],andBandet al. [1991, 1993]. 3.1.1. RItESSys data inputs. A combination of distributed and nondistributeddata inputsis requiredto drive the RHESSyssimulationsystem(Figure 3). For eachcatchment the drainage outlet was located within a Global Positioning System-derivedUniversal TransverseMercator coordinate system.The drainagearea was extractedand partitionedinto hillslopes,ridge lines,and streamlinesaccordingto the methods of Band [1989],Lammersand Band [1990], and Mackay and Band [1994]. Once delimited,hillslopeswere parameterized for forest,topographic,and soil attributes. Distributedcanopyleaf areaindex(LAI) datawere derived from LANDSAT TM data [Nemani et al., 1993] that were calibratedusinga combinationof DECAGON SunfleckCeptometer-derived and allometricallyderived LAI data. Topographicattributes,includingboth primary (slope,aspect,elevation, contributingarea) and secondary(catenary index) attributes,were derived from digital terrain analysis[Band,

uousmonitoringsitefor the studyof catchmenthydrologyand biogeochemistry (see referencesin a specialissueon TLW publishedin CanadianJournalof Fisheriesand Aquatic Sciences,45, supplement1, 1988). A catchmentmonitoringnetwork was strategicallyestablishedto encompass the range in the complexityof the catchmenttopography, with somecatchments characterizedby relativelygentle, concaveslopesand other catchmentscharacterizedby relatively steep, convex slopes.Previouslyunpublisheddata collectedfrom this monitoring network during the 10-year period from 1981 to 1990 1986,1989].Soil attributes,includingeffectivemineralsoil providedthe basisfor this investigationon the topographic depth(or rootingdepth),saturatedsurfacehydraulicconductivityKs, and the rate of changeof saturatedsurfacehydraulic effectson the exportof NO•--N (criterion6).

CREED AND BAND: CATCHMENT

NO•--N EXPORT BEHAVIOR

'"' '"-

3081

/

x x _1

.. r

/

I \

1 km Watershed Boundary Lake Boundary, River

Contour

- •-

ß

Index

ershed

and Streams Interval

Contour

Catchment

Interval

Weir

Figure 1. Locationof the Turkey LakesWatershed(TLW) (centeredat 47ø03'00"Nand 84ø25'00"W).The analysespresentedin this paper focuson the continuouslygaugedcatchmentswhich representthe range in topographicconditionswithin this region.

conductivitywith soil depth m, were estimatedfrom field- and literature-baseddata. Distributionsof K s and m in catchmerits that range from 5 to 70 ha were difficult to estimate,particularly as field data were collectedfrom sparsesites.Although a realistic range of these parameterswas estimated from the field and the literature, the actualvalueswithin this rangewere determined during model calibration in which the simulated

dischargewas matched to the observeddischarge.Using the sameset of parametersfor Ks and m, we producedacceptable model simulationsof dischargehydrographs(simulatedversus observedestimatesof annual dischargewere within +_15%, with no apparentbiasamongthe years)for eachcatchmentfor the period from 1981 to 1990 [Creedet al., 1996;I. F. Creed, unpublisheddata, 1996).

3082

CREED AND BAND: CATCHMENT NO•--N EXPORT BEHAVIOR

Depth

Dissolved

to

N Species

Bedrock

profile.The RHESSys-derived N indicesprovidea meansof exploring N dynamics bothat the subcatchment scale,in terms of potentialsinksandsources of NO•--N withinthe catchment, and at the catchmentscale,in termsof the potentialexportof NO•--N from the catchment.As the focusof this paperis to

A

compare result• among different catchments, wederived N

Time

indicesfrom catchment-average dynamics. A shortdescription of the derivation of the N source and flush indicesused by

Depth

Creedet al. [1996]is givenbelow,andthe readeris referredto

to

Drainageof

Bedrock

that reference

Dissolved

N Species

NExport

is reasonable for the TLW

Moderate

Saturated

Subsurface Flow

N Export Large

Surface

detail.

the availabilityof NO•--N in the soilprofile.A simpleN source index is basedon the assumptionthat N cyclingactivityis stoichiometrically relatedto C cyclingactivity.Thisassumption

Small

Baseflow

for additional

3.1.2.1. N sourceindex: The N sourceindex represents

N Export

andSaturated Subsurface Flow

soils which are characterized

as

havingan almost1:1transformation of organicN to inorganic N [Foster,1985,1989].The N sourceindexis computedasthe ratioof C supplyto C demand.The C supplyis an exponential functionof the maximumrate of grossorganicC decomposition modifiedby a moisturefactor (MF) and a temperature factor (TF) [after Partonet al., 1987].The C demandis a functionof net carbonfixation(equation(1)). Source

B

-

Supply Cpotentialsoil pool(1 -- exp[MF X TF]) Cphotosynthesi s-- Crespiratio n

Demand

(1)

Figure 2. A conceptualmodelof the hypothesized controls on the exportof NO•--N. The flushingexportof NO•--N is a functionof two groupsof interactiveprocesses: (a) processes that control the net accumulationof N in the soil, and (b) processes thatcontrolthenettransportof N fromthesoil.The flushingexport of NO•--N is presentedin three diagnostic stages.At stage1 the water table is low and water-soluble formsof N accumulate withinthe soilprofile,resultingin small concentrations of NO;--N in dischargewaters.At stage2, as

Distributed

Data

Generated by GeographicInformation System

[

Drainage Nemørk ExtraCfiøn ,:,:,'," ]

the water table startsto rise, water-solubleforms of N in the

soilprofileareflushed,resultingin moderateconcentrations of NO•--N in dischargewaters.At stage3, as the water table continuesto rise,water-solubleforms of NO•--N in the organ-

icallyenrichedsurfaceof the soilprofileare flushed,resulting in largeconcentrations of NO•--N in discharge waters.

Nondistributed forestphysiological parameters werederived from field- and literature-baseddata and are summarizedby

Creedet al. [1996].The meteorological variablesdrivingthe simulationsystem,includingdailyprecipitationanddailyminimum and maximumtemperaturedata, were collectedat an AtmosphericEnvironmentService(AES) meteorological recordingstationlocatedat the southeast boundaryof the TLW. Missingdatawere estimatedby linearregression on the basis of datafrom the closestmeteorological stationswithinthe area for whichcompleterecordswereavailable(MontrealFallsand Sault Sainte Marie, Ontario). The meteorologicalvariables wereextrapolated from the recordingstationto eachhillslope usingthe RHESSysclimatemodel,MTCLIM [Running et al., 1987],enablingus to capturevariationsdue to slope,aspect, and elevationbut not variationsdue to topographicshadows. 3.1.2. RHESSysdata outputs. RHESSyssimulationoutputsprovidethebasisfor computation of similarityindicesthat representthe hypothesized controlson the exportof NO•-N (Figure3): (1) a N sourceindex,whichconsiders the availability of NO•--N to be flushedfromthe soilsurface,and(2) a N flushindex,whichconsidersthe flushingof NO•--N causedby the interactionsof the water table with NO•--N within the soil

]NOn'distrib uted Data[ I

Climate oa•and •

I physiø10gicdl Parameters I

I f6•eac• Førest species I

Simulation System

MiCroClimatological Biogeochemical &

HYdrOlo•calMOdel

Figure 3. A schematicdiagramof the RegionalHydroEcologicalSimulationSystem(RHESSys).The engineof RHESSysis the simulationsystem.Data inputsto the simulation systemincludedistributedforest,terrain, and soil data and nondistributedclimate data and forestphysiological parameters. Data outputsfrom the simulationsystemincludewater andcarbonbudgets,fromwhichare derivedthe N sourceand N flush indices.

CREED AND BAND: CATCHMENT NO•--N EXPORT BEHAVIOR

To examinethe variabilityof the N sourceindex,both within and amongcatchments,a relativeN sourceindexis computed that normalizesthe sourceindexto the watershed'stemporal and spatialmaximumN sourceindex(equation(2)).

3083

rn a parameterproportionalto the rate of changeof saturatedhydraulicconductivitywith depth;

X hillslope-average catenary index(f In (aTe/Titan13)da).

3.1.2.2. N flush index: The N flushindex representsthe interactionsof the water table with NO•--N within the soil profile. The water table dynamicsare basedon the soil water

Biogeochemicalstores and fluxes, includingevaporation, transpiration,photosynthesis, and respirationare computed for each catenarypositionwithin a catchment.At the end of each day the saturationdeficit for each catenarypositionis area-weightedto update the saturationdeficit for each hillslope(Sh) and,in turn, thissaturationdeficitis area-weighted to updatethe saturationdeficitfor the catchment(S), which

saturation

forms the basis of the N flush index.

Source

Relative Source Index = Sourcemax (2)

deficit.

Soil water

saturation

occurs when

all the

where

The N flush index is computedas S/S3o, the ratio of the currentsaturationdeficitto the previous30-dayaveragesaturation deficit.While S considersonly the currentday saturation deficit,S/S3o incorporatesS, the risingand recedingoscillationsof S, and the frequencyof the oscillationsof S. The N flushindexprovidesthreepossiblescenarios: (1) If S/S3o< 1, the current day water table is higher than before, and saturatedthroughflowis risinginto previouslyunsaturatedparts

Sh hillslope-average soil saturationdeficit,m;

the currentdaywater table is lower than before, and saturated

pores in the soil are filled with water. Although a deficit is typicallyrepresentedon a negativescale,in this paper, we presentits absolutevalue, on a positivescale.The hillslopeaveragesaturationdeficit Sh is computedon the basisof the followingwater balanceequation(equation(3)):

Sh • Sp- Rn + Q

(3)

of thesoilprofile(i.e.,catchment isflushing), (2) if S/S3o> 1,

Sp previous dayhillslope-average soilsaturation deficit,m; R n net rechargeto the saturatedzone, a functionof the throughfallor melt that exceedsthe interception capacityof the litter and the unsaturatedsoil water dynamicsas computedby BIOME-BGC, m; Q hillslopedischargeto the stream,m.

In catchmentswith significanttopographicrelief the distribution of soil water can be highlyheterogeneous, leadingto nonlinear effectson the biogeochemicalcyclingrates within the catchment[Band, 1993;Band et al., 1993]. Topographic attributes

have been used to estimate

the distribution

of soil

water within a hillslope[Bevenand Kirkby, 1979; O'Loughlin, 1981].In TOPMODEL [Bevenand Kirby, 1979],the distribution of water within the hillslopeis dependenton catenary position.Catenarypositionswith larger contributingdrainage areasand smallersurfaceslopes(i.e., near streams)will have smallsaturationdeficits,while catenarypositionswith smaller contributing drainageareasandlargersurfaceslopes(i.e., near ridges)will have large saturationdeficits.For each point i within the hillslopea catenaryindex is computed(equation (4)) on the basisof the assumption of a steadystatesubsurface throughflowsystemin whichrechargefrom the unsaturatedto the saturatedzone in the contributingarea abovea unit length of contouris equal to the saturatedthroughflowacrossthe contourlength[Bevenand Kirby,1979;Beven,1986].

CI = In (a Te/Ti tan/3)

(4)

where

throughflowis receding(i.e., catchmentis draining),and (3) if S/S3o = 1, there is no changein the water table relativeto the previous30-dayperiod.The 30-dayaveragesaturationdeficitis an appropriatelength of time betweenconsecutive flushings for the accumulationof substantialN in the soilprofilefor the Turkey Lakes Watershed [Creedet al., 1996]; however,the optimal lengthof time will vary over the year. In addition,the optimallengthof time will varyfor differentregions.Note that sinceS representsthe catchment'saveragesaturationdeficit, intersectionof a risingwater table with the NO•--N within the soil profile can occureither when S is small (water table is high, and large flushing areas within the catchment are present)or whenS is largebut one tail of the distributionof soil saturationdeficitsis small (water table is low, but small flushingareaswithin the catchmentare present). 3.2.

Chemistry of Discharge Waters

For eachcatchment,total dailywater discharge(mm/d) was derived from a continuouslymeasuredstream-gaugestation equippedwith a 90øor 120øV notchweir. Missingdata (usually wintersmallflows)were estimatedby linearregression with an adjacentcatchmentfor which there was a completerecord. Streamwater sampleswere collectedevery2 weeksduringthe winter, daily duringspringsnowmelt,and everyweek or every 2 weeks during the summerand autumn. Sampleswere collectedat the samesamplingpoint, in the centerof the stream, duringeachvisit. To removeparticulatematter, each sample was filtered through a Whatman No. 41 filter that had been rinsed with distilled water. To determine

CI catenary index,m2; a contributing area,m2; Ti local soil transmissivity, m/d; Te hillslope-average soil transmissivity, m/d; /3 local gradient,degrees.

The catenaryindexis usedto computethe saturationdeficitat point i, Si (equation(5)):

Si = Sh nt-m(X - CI) where

Si local soil saturationdeficit, m; Sh hillslope-average soil saturationdeficit,m;

(5)

the concentration

of

NO•--N, a 100-mL subsample was filtered througha 0.45-/xm Millipore membranefilter that had been washedwith distilled water andstoredwithoutpreservativeat 2øCin glasscontainers that had been washedwith acid water (0.1 N H2804) and rinsedwith distilledwater. The subsamplewas analyzedfor NO•--N within 48 hoursof collectionby a cadmiumreduction procedureusinga TechniconAutoAnalyzer II-C+. To determine the concentrationof calcium(Ca), a 100-mLsubsample wasstoredwithoutpreservativeat 2øCin polyethylenecontainers that had beenwashedwith acidwater (10% HNO3) and rinsedwith distilledwater. The subsample wasanalyzedfor Ca within 7 daysof collectionby atomicabsorptionon a Varian 1275 spectrophotometer [Nicolson,1988].

3084

CREED AND BAND: CATCHMENT NO•--N EXPORT BEHAVIOR

c50 TYPE

A

c37, c50

Jan

Mar

Jun

Sep

Dec

2.5

c46 TYPE B

c42, c44, c46, c47, c49

'Z•0.5 .

0

Jan

Mar

Jun

Sep

Dec

2,

TYPE C

c31, c33, c34

Jan

Mar

Jun

Sep

Dec

TYPE

D

c32, c35, c39

•0.5 0•

Jan

Mar

Jun

Sep

Dec

Figure 4. Naturalvariationin the exportof NO•--N. The graphsdepictthe averageconcentration of NO•--N in discharge waters(10-yearaverage_+1 standarderror). The TLW catchments showfour typesof responses in NO•--N export:(a) smallbaseconcentrations with peakconcentrations restrictedto springmelt, (b) small baseconcentrations'with peak concentrations occurringthroughoutthe year, (c) large baseconcentrations with peak concentrations occurringthroughoutthe year, and (d) largebaseconcentrations with peak concentrationsrestrictedto springmelt. This groupingof the NO•--N exportsdoesnot infer that the processes generatingthe NO•--N exportsare similarwithin eachgroup. 4.

4.1.

Results

and Discussion

Nitrate-N Export Behavior

and extrinsicprocesses involvedin the exportof NO•--N from catchments.

Nitrate-N exportcouldbe classifiedaccordingto one of four 4.2. Similarity in N Dynamics By usingour hybrid modelingapproach,we generatedretypes of responses(Figure 4) [Creed,1988]: NO•--N peaks restrictedto springmelt with small NO•--N base concentra- sponsesurfacesthat representthe observedexportof NO•--N tions(Figure4a), NO•--N peaksoccurring throughout theyear as a function of the simulatedhypothesizedcontrolson the with small NO•--N baseconcentrations (Figure 4b), NO•--N export of NO•--N (Figure 5). An analysisof these surfaces conpeaks occurringthroughoutthe year with elevatedNO•--N providesinsightsto the importanceof the hypothesized baseconcentrations (Figure4c), and NO•--N peaksrestricted trols. If NO•--N export is restrictedto catchmentconditions to the springmelt with elevatedNO•--N baseconcentrations reflectinga combinationof NO•--N formation and NO•--N at capturingthe (Figure 4d). We hypothesized that the observedvariationin flushingactivities,then we were successful the export of NO•--N amongthese catchmentswas due to dominant controls on NO•--N export. However, if NO•--N factorsintrinsicto eachcatchment(e.g.,topographiccomplex- export is not restrictedto these catchmentconditions,then ity) rather than factorsextrinsicto the catchments.We at- other processesare important for NO•--N export. The assemblage of catchmentsshowssimilar responsesin tempted to capture the observedvariation in the export of of NO•--N (Figure 5, light NO•--N with similarityindicesthat captureboth the intrinsic the exportof peak concentrations

o

.

o

-•

Similarity in the "si in the N respons that occurs in respo N flushing a

, 1.5•

c46

N formation

•

•

act

.o ß

ß

•

0.5 •

Leg

..

• •

.

[ •o c39

08 •0

.

0.5 o '

•*

o o.

'

• by "peak" resp shallow hydro

•• by "base" deepresp hydrolo

Figure 5. Response surfaces of observed exportof NO•--N (mg/L) asa functionof the sim andflushindices(the response surfaces are for the four catchments depictedin Figure4).

3O86

CREED AND BAND: CATCHMENT

NO•--N EXPORT BEHAVIOR

The form of the responsesurfacessuggeststhat the concentration of NO•--N in dischargewatersmay be an exponential functionof the N sourceand flushindices(equation(6)):

Flushing States

•' •[• '• •

Oralran 'øg

•t.5'

t .•

ates

N = a exp (-/3FI) exp (+ ySI)

(6)

where

NO•--N concentration,mg/L; Source Index, dimensionless; Flush Index, dimensionless;

FI

empiricalcoefficients.

a, /3, 7

For c50, a catchment with minimal base concentrations of

NO•--N (wherebaseconcentrations reflectNO•--N exportduring a nonflushingstate), a NO•--N export model basedon an exponentialfunctionof the N indicesprovidesresultsthat are significant,but not with a large degreeof explanation(Figure

Figure 6. The flushingexportof NO•--N (mg/L) as a function of saturationdeficit S (mm). The water table is at the catchment

surface

when

S -

0 and below

the catchment

surfacewhenS > 0. Flushing(S/S3o < 1) occursboth when the water table is high (S - 0) duringlarge flow periodsand when the water table is low (S - 250 mm) during smallflow periods.In the dischargewaters the peak concentrationsof NO•--N occurduringflushingconditions,while the baseconcentrationsof NO•--N occurduringdrainingconditions[after Creedet al., 1996].

7, p < 0.001, multiple1'2 = 0.32). We attributethe small predictivepower largelyto limitationsof the N sourceindex that are described

below.

4.2.1. Iramobilizing versus mobilizing properties. A sea-

sonal analysisof the residualsof the NO•-N export model (Figure 8) indicatesthat the export of NO•--N during the autumn flushingtime period was overestimated.For the N sourceindex(equations(1) and (2)) a weaknessof the computationof the supplyterm is that in assuminga stoichiometric relationshipbetweenthe generationof carbon(CO2) and N, we did not considerthe role of microbial dynamicsin the generationof N. In particular,we did not discriminatebetween the immobilization(i.e., organicN) versusmobilization(i.e., NO•--N) fates of the mineralizednitrogen. Although the assumptionof a stoichiometricrelationshipfor the generationof CO2 andN (i.e., a 1:1transformation of organicN to inorganic NO•--N) appliesover most of the season[Foster,1989], the assumption doesnot applywith canopyturnoverin the autumn

areas).The occurrenceof a singlepeak in the responsesurfacesindicatesthat we were able to capture,with our simple indicesof N dynamics,the diverseconditionsof the catchment that result in peak NO•--N export. The diverse conditions includeflushingstateswithin the catchmentwhen it is characterizedby saturatedareasthat rangefrom relativelylargeareas (i.e., S - 0, duringmelt and stormevents)to relativelysmall season. areas(i.e., S