ebook img

Hartmann, AJ, Kobler, J., Kralik, M., Dirnböck, T., Humer, F., & Weiler, M. PDF

17 Pages·2017·5.73 MB·German
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Hartmann, AJ, Kobler, J., Kralik, M., Dirnböck, T., Humer, F., & Weiler, M.

Hartmann, A. J., Kobler, J., Kralik, M., Dirnböck, T., Humer, F., & Weiler, M. (2016). Model-aided quantification of dissolved carbon and nitrogen release after windthrow disturbance in an Austrian karst system. Biogeosciences, 13(1), 159-174. DOI: 10.5194/bg-13-159-2016 Publisher's PDF, also known as Version of record License (if available): CC BY Link to published version (if available): 10.5194/bg-13-159-2016 Link to publication record in Explore Bristol Research PDF-document This is the final published version of the article (version of record). It first appeared online via EGU at http://www.biogeosciences.net/13/159/2016/. Please refer to any applicable terms of use of the publisher. University of Bristol - Explore Bristol Research General rights This document is made available in accordance with publisher policies. Please cite only the published version using the reference above. Full terms of use are available: http://www.bristol.ac.uk/pure/about/ebr-terms.html Biogeosciences,13,159–174,2016 www.biogeosciences.net/13/159/2016/ doi:10.5194/bg-13-159-2016 ©Author(s)2016.CCAttribution3.0License. Model-aided quantification of dissolved carbon and nitrogen release after windthrow disturbance in an Austrian karst system A.Hartmann1,2,J.Kobler3,M.Kralik3,T.Dirnböck3,F.Humer3,andM.Weiler1 1FacultyofEnvironmentandNaturalResources,FreiburgUniversity,FreiburgimBreisgau,Germany 2DepartmentofCivilEngineering,UniversityofBristol,Bristol,UK 3EnvironmentAgencyAustria,Vienna,Austria Correspondenceto:A.Hartmann([email protected]) Received:8June2015–PublishedinBiogeosciencesDiscuss.:31July2015 Revised:4December2015–Accepted:13December2015–Published:15January2016 Abstract. Karst systems are important for drinking water 1 Introduction supply. Future climate projections indicate increasing tem- perature and a higher frequency of strong weather events. Both will influence the availability and quality of water provided from karst regions. Forest disturbances such as Karst systems contribute around 50% to Austria’s drinking windthrow can disrupt ecosystem cycles and cause pro- water supply (COST, 1995). Karst develops due to the dis- nounced nutrient losses from the ecosystems. In this study, solvabilityofcarbonaterock(FordandWilliams,2007)and we consider the time period before and after the wind dis- itresultsinstrongheterogeneityofsubsurfaceflowandstor- turbance period (2007/08) to identify impacts on DIN (dis- age characteristics (Bakalowicz, 2005). The resulting com- solvedinorganicnitrogen)andDOC(dissolvedorganiccar- plex hydrological behaviour requires adapted field investi- bon) with a process-based flow and solute transport simu- gation techniques (Goldscheider and Drew, 2007). Future lation model. When calibrated and validated before the dis- climate trajectories indicate increasing temperature (Chris- turbance,themodeldisregardstheforestdisturbanceandits tensen et al., 2007) and a higher frequency of hydrological consequencesonDINandDOCproductionandleaching.It extremes(Dai,2012;Hirabayashietal.,2013).Bothwillin- canthereforebeusedasabaselinefortheundisturbedsystem fluence the availability and quality of water provided from andasatoolforthequantificationofadditionalnutrientpro- karstregionsbecausetemperaturetriggersnumerousbiogeo- duction. Our results indicate that the forest disturbance by chemicalprocessesandfastthroughflowwaterhasadispro- windthrow results in a significant increase of DIN produc- portional effect upon water quality. Furthermore, forest dis- tion lasting ∼3.7 years and exceeding the pre-disturbance turbances(windthrows,insectinfestations,droughts)posea average by 2.7kgha−1a−1 corresponding to an increase of threat to water quality through the mobilization of potential 53%.TherewerenosignificantchangesinDOCconcentra- pollutants,andthesedisturbancesarelikelytoincreaseinthe tions.Withsimulatedtransittimedistributionsweshowthat future(Johnsonetal.,2010;Seidletal.,2014). the impact on DIN travels through the hydrological system One way to quantify the impact of changes in climatic within some months. However, a small fraction of the sys- boundaryconditionsonthehydrologicalcycleisthroughuse temoutflow(<5%)exceedsmeantransittimesof>1year. ofsimulationmodels.Specialmodelstructureshavetobeap- plied for karst regions to account for their particular hydro- logicalbehaviour(Hartmannetal.,2014a).Arangeofmod- els of varying complexity are available from the literature that deal with the karstic heterogeneity, such as groundwa- terflowintherockfracturematrixanddissolutionconduits (Jourdeetal.,2015;Kordillaetal.,2012),varyingrecharge areas(Hartmannetal.,2013a;LeMoineetal.,2008)orpref- PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion. 160 A.Hartmannetal.:Model-aidedquantificationofdissolvedcarbonandnitrogenrelease erentialrechargebycracksinthesoilorfracturedrockout- weusedvirtualtracerexperimentstocreatetransittimedis- crops(RimmerandSalingar,2006;Tritzetal.,2011). tributionsthatexpressedhowtheimpactofthestormsprop- Nitrate and dissolved organic carbon (DOC) have both agated through the variable dynamic flow paths of the karst beenconsideredindrinkingwaterdirectivesandwaterprepa- system. This allowed us to assess the vulnerability of the rationprocesses(Goughetal.,2014;Mikkelsonetal.,2013; karstcatchmenttosuchimpacts. Tissier et al., 2013; Weishaar et al., 2003). Though nitrate pollution of drinking water is usually attributed to fertil- ization of crops and grassland, an excess input of atmo- 2 Studysite spheric nitrogen (N) from industry, traffic and agriculture into forests has caused reasonable nitrate losses from for- Thestudysite,LTERZöbelboden,islocatedinthenorthern est areas (Butterbach-Bahl et al., 2011; Erisman and Vries, part of the national park “Kalkalpen” (Fig. 1). Its altitude 2000;Gundersenetal.,2006;Kieseetal.,2011).TheNorth- ranges from 550 to 956ma.s.l. and its area is ∼5.7km2. ern Limestone Alps area is exposed to particularly high ni- Mean monthly temperature varies from −1◦C in January trogen deposition (Rogora et al., 2006) and nitrate leaching to 15.5◦C in August. The average temperature is 7.2◦C occursatincreasedrates(Jostetal.,2010).Inadditiontothis, (at 900ma.s.l.). Annual precipitation ranges from 1500 to forest disturbances such as windthrow and insect outbreaks 1800mm and snow accumulates commonly between Octo- disrupttheNcycleandcausepronouncednitratelossesfrom berandMaywithanaveragedurationofabout4months.The the soils, at least in N-saturated systems that received ele- mean N deposition in bulk precipitation between 1993 and vated N deposition due to elevated NOx in the atmosphere 2006was18.7kgNha−1yr−1,outofwhich15.3kgN(82%) − (Bernal et al., 2012; Griffin et al., 2011; Huber, 2005). In was inorganic (approximately half as NO -N and half as 3 contrast to N deposition, atmospheric deposition of DOC is NH+-N;Jostetal.,2010).Duetothedominanceofdolomite, 4 low (Lindroos et al., 2008) and thus has not been identified the catchment is not as heavily karstified as limestone karst as major driver of DOC leaching from subsoil (Fröberg et systems, but shows typical karst features such as conduits al.,2007;KaiserandKalbitz,2012;Verstraetenetal.,2014). and sink holes (Jost et al., 2010). The site can be split into Moreover, studies show contrasting results but point to in- steep slopes (30–70◦, 550–850ma.s.l.) and a plateau (850– creasedDOC(TOC)leachingfromsoilandcatchmentsafter 950ma.s.l.), with the plateau covering ∼0.6km2. Chromic forestdisturbances(Huberetal.,2004;Löfgrenetal.,2014; Cambisols and Hydromorphic Stagnosols with an average Meyeretal.,1983;Mikkelsonetal.,2013;Wuetal.,2014). thickness of 50cm and Lithic and Rendzic Leptosols with WhilemanystudiesidentifyNandDOCassourceofcon- an average thickness of 12cm can be found at the plateau taminationinkarstsystems(Einsiedletal.,2005;Jostetal., and the slopes, respectively (WRB, 2006). Both the plateau 2010;Katzetal.,2001,2004;Tissieretal.,2013)orprovide andtheslopesaremainlycoveredbyforest.Norwayspruce static vulnerability maps (Andreo et al., 2008; Doerfliger et (Piceaabies(L.)Karst.)interspersedwithbeech(Fagussyl- al., 1999), only very few studies use models to quantify the vaticaL.)wasplantedafteraclearcutaroundtheyear1910. temporal behaviour of a contamination through the systems The vegetation at the slopes is dominated by semi-natural (ButscherandHuggenberger,2008).SomestudiesuseNand mixed mountain forest with beech (Fagus sylvatica) as the DOC to better understand karst processes (Charlier et al., dominant species, Norway spruce (P. abies), maple (Acer 2012; Mahler and Garner, 2009; Pinault et al., 2001) or for pseudoplatanus),andash(Fraxinusexcelsior).Attheslopes advanced karst model calibration (Hartmann et al., 2013b, no forest management has been conducted since the estab- 2014b), but to our knowledge there are no applications of lishmentofthenationalpark. suchapproachestoquantifythedrainageprocessesofNand DOC,andparticularlysoafterstrongimpactsonecosystems 2.1 Availabledata (e.g. windthrow) that release reasonable amount of nitrate fromtheforestsoils. A10-yearrecordofinputandoutputobservationswasavail- In this study, we consider the time period before and af- able. Starting from the hydrological year 2002/03, it en- ter storm Kyrill (early 2007) and several other storm events velops well the stormy period that began in January 2007. (2008)thathitcentralEurope.Thestorms,fromnowonre- Itincludeddailyrainfallmeasurementsandstreamdischarge ferredtoasthewinddisturbanceperiod,causedstrongdam- measurements from stream sections 1 and 2 (Fig. 1). We agetotheforestsinourstudyarea,adolomitekarstsystem. obtained the discharge of the entire system with a simple Weapplyanewtypeofsemi-distributedmodelthatconsiders topography-based up-scaling procedure that is described in the spatial heterogeneity of the karst system by distribution more detail in Hartmann et al. (2012a). Irregular (weekly functions. We aimed at comparing the hydrological and hy- to monthly) observations of DOC, DIN and SO2− concen- 4 drochemicalbehaviour(DOC,DIN)ofthesystembeforeand trations are available for precipitation and at weir 1. DOC, duringthewinddisturbingperiod.Inparticular,wewantedto NO−, SO2− and NH+ (since January 2010) samples were 3 4 4 + understandifandhowDOCandDINinputtothehydrologi- filtered (0.4µm) before the analysis. NH concentrations 4 calsystemchangedbytheimpactofthestorms.Furthermore, were measured by spectrophotometry (Milton Roy Spec- Biogeosciences,13,159–174,2016 www.biogeosciences.net/13/159/2016/ A.Hartmannetal.:Model-aidedquantificationofdissolvedcarbonandnitrogenrelease 161 Figure 1. Study site and location of measurement devices (Hart- mannetal.,2012a;modified). tronic).WeeklyDOC,SO2− andNO− sampleswerepooled 4 3 to provide volume weighted biweekly (until March 2009) Figure2.Intra-annualandinterannualvariationsin(a)DOCcon- centrations,(c)DINconcentrationsand(e)discharge,andrelation andmonthly(thereafter)samples.DOCsampleswereacidi- betweendischargeand(b)DOCand(d)DINbeforeandduringthe fied with 0.5mL HCl 25% and were measured with a Mai- winddisturbanceperiod. hakTOCOR100andaCPNTOC/DOCanalyser(Shimadzu Corp., Japan). NO− and SO2− concentrations were deter- 3 4 mined by ion chromatography with conductivity detection. − DIN input was then calculated as the sum of NO3-N and 6mgL−1 during high discharge (similar to Hagedorn et al., + + − NH -N. Since NH is either transformed into NO or ab- 2000). However, there was no obvious difference between 4 4 3 + sorbed in the soil, NH concentrations in runoff are very thepre-disturbanceperiod(Fig.2b). 4 small or not detectable. Therefore we calculated DIN out- − putsasNO -N.Additionally,irregularobservationsofsnow 3 waterequivalentattheplateauallowedforindependentsetup ofthesnowroutines. 3 Methods 2.2 Recentdisturbances 3.1 Themodel Kyrillintheyear2007andsomesimilarlystrongstormssub- sequent to 2008 caused some major windthrows as well as singletreedamages.Awindthrowdisturbanceof∼5haoc- 3.1.1 Modelhydrodynamics curred upstream of weir 1. Though no direct measurements existastothetotalextentofthewindthrowareaweestimate The semi-distributed simulation model considers the vari- that5–10%ofthestudysitehasbeensubjecttowindthrow ability in karst system properties by statistical distribution (Kobleretal.,2015).Wedidnotobserveasignificantchange functionsspreadoverZ=15modelcompartments(Fig.3). in intra- and interannual variability in DOC concentrations Thatwayitsimulatesarangeofvariablydynamicpathways anddischargebeforeandduringthewinddisturbanceperiod throughthekarstsystem.Thedetailedequationsofthemodel (Fig. 2a,e). Runoffconcentrations ofDIN showedclear re- hydrodynamicsaresimilartoitspreviousapplications(Hart- sponsestothedisturbances.Withthefirstwindthroweventit mannetal.,2013a,c,2014b).TheyaredescribedintheAp- startedtoincreaseuntil2008/09andslowlydecreasedagain pendix. Since in our case the model is used to simulate the in 2010/11 (Fig. 2c). Comparison of DOC concentrations dischargeoftheentiresystemandaweirwithinthesystem, with discharge before and during the wind disturbance pe- some small modifications had to be performed. Preceding riodrevealedasimilarpattern.Asshownbyotherstudieson studiesshowedthatweir1(Fig.1)receivesitsdischargepar- DOCmobilization(e.g.,RaymondandSaiers,2010),apos- tially from the epikarst and partially from the groundwater, itive correlation between concentrations and discharge (on reaching it partially as concentrated and partially as diffuse log10 scale) occurred for DOC with concentrations up to flow (Hartmann et al., 2012a). Consequently we derive its www.biogeosciences.net/13/159/2016/ Biogeosciences,13,159–174,2016 162 A.Hartmannetal.:Model-aidedquantificationofdissolvedcarbonandnitrogenrelease in the rock matrix, additional processes were included in themodelstructure.Similartoprecedingstudies(Hartmann etal.,2013a,2014b)SO2− dissolutionG [mgL−1]for 4 SO4,i compartmenti iscalculatedby (cid:18)Z−i+1(cid:19)aGeo G =G · , (2) SO4,i max,SO4 Z wherea [–]isanothervariabilityparameterandG Geo max,SO4 [mgL−1]istheequilibriumconcentrationofSO2−inthema- 4 trix.DOCismostlymobilizedattheforestfloor(Borkenet al., 2011). Stored in the soil or diffusively and slowly pass- ingdownwards,largepartsoftheDOCareabsorbedorcon- sumed by microorganisms, but when lateral flow and con- centratedinfiltrationincrease,netleachingofDOCincreases aswell.ForthatreasonourDOCtransportroutineonlypro- videswatertotheepikarstwhenitissaturated(Eq.A4)with Figure3.Diagramofmodelstructure;itisassumedthatdischarge increasing DOC net production toward the more dynamic andhydrochemistryatthetwoweirsiscomposedbydifferentmix- modelcompartments(Fig.3).ItsDOCconcentrationP turesofdiffuserecharge(green),concentratedrecharge(red),dif- DOC,i fuse groundwater flow (blue) and concentrated groundwater flow [mgL−1]foreachmodelcompartmentisfoundby (purple). P =P ·(cid:18)Z−i+1(cid:19)−aD1OC, (3) DOC,i DOC Z dischargeQ [Ls−1]by weir where a [–] is the DOC variability constant and P DOC DOC Q (t)= f [mgL−1] is the DOC net production at soil compartment weir Epi 1. Similar to other studies that assessed N input to a karst h XZ XZ i · fEpi,conc· Rconc,i(t)+(1−fEpi,conc)· Rdiff,i(t) system (Pinault et al., 2001), we used a trigonometric se- i i +(1−f )·(cid:2)f ·Q (t)+(1−f ) riestoassessthetimevariantnetproductionofDIN,PDIN,i Epi GW,conc GW,Z GW,conc [mgL−1],tothesoil: XZ−1 i · Q (t) , i GW,i (1) PDIN,i =PDIN+ADIN·sin(cid:18)365.25·(cid:0)JD+SPH,DIN(cid:1)(cid:19). (4) 2π where fEpi is the fraction from the epikarst and (1−fEpi) Here, PDIN is the mean amount of dissolved inorganic N in the fraction from the groundwater. fEpi,conc and fGW,conc thesoilsolution,whileAN [mgL−1]andSPH,DIN[d]arethe represent the concentrated flow fractions of the epikarst amplitude of the seasonal signal and the phase shift of sea- andgroundwatercontributions,respectively.Table1listsall sonalDINuptake(immobilizationbyplantsandsoilorgan- modelparameters,withashortdescriptionforeach. isms) and release (net DIN in the soil water) cycle, respec- tively.J istheJuliandayofeachcalendaryear.Duetoits D 3.1.2 Modelsolutetransport seasonalvariation,P canalsobenegative,meaningthat DIN,i uptakeofDINtakesplace. To model the non-conservative transport of DOC and, DIN andSO2−,weequippedthemodelwithsolutetransportrou- 4 3.2 Modelcalibrationandevaluation tines. SO2− was included as an additional calibration vari- 4 able because it proved to be important to reduce model With 14 model parameter that controlled the hydrodynam- equifinality(Beven,2006)byaddingadditionalinformation ics and 7 parameters that allow for the non-conservative about groundwater dynamics (Hartmann et al., 2013a, b). solute transport, the calibration of the model was a high- Theinclusionofthesethreesolutesallowedforamorereli- dimensional problem. For that reason we have chosen the ableestimationofmodelparameters(Hartmannetal.,2012b, Shuffled Complex Evolution Metropolis (SCEM) algorithm 2013a)and,lateron,theevaluationofpossiblechangesinthe (Vrugt et al., 2003), which proved to be capable of explor- dynamic of solute concentrations during the stormy period. ing high-dimensional optimization problems (Fenicia et al., For most of the model compartments solute transport sim- 2014;Feyenetal.,2007;Vrugtetal.,2006).Asperformance ulations simply followed the assumption of complete mix- measureweusedtheKling–Guptaefficiency(KGE;Guptaet ing. However, to represent net production and leaching of al.,2009).Forcalibration,KGEwasweightedequallyamong DOC and DIN in the soil, as well as dissolution of SO2− all solutes, one-third for the discharge of the entire system 4 Biogeosciences,13,159–174,2016 www.biogeosciences.net/13/159/2016/ A.Hartmannetal.:Model-aidedquantificationofdissolvedcarbonandnitrogenrelease 163 Table1.Modelparameters,description,rangesandcalibratedvalueswithKGEperformancesforthecalibrationandvalidationsamples. Parameter Description Unit Ranges Optimizedvalues Lower Upper Sample1 Sample2 Vmean,S meansoilstoragecapacity mm 0 1500 450.18 599.13 fvar,S fractionofthespoilthathasavariabledepth – 0 1 0.06 0.02 Vmean,E meanepikarststoragecapacity mm 0 1500 1495.49 1233.98 aSE soil/epikarstdepthvariabilityconstant – 0 2 1.69 1.91 Kmean,E epikarstmeanstorageconstant d 1 50 2.65 8.27 afsep rechargeseparationvariabilityconstant – 0 2 0.88 1.44 KC conduitstorageconstant d 1 10 1.37 1.03 aGW groundwatervariabilityconstant – 0 2 2.00 1.88 fEW fractionofweirdischargeoriginating – 0 1 0.56 0.72 fromtheepikarst fWE,conc fractionofweirdischargeoriginating – 0 1 0.57 0.47 fromtheepikarstasconcentratedflow fWGW,conc fractionofweirdischargeoriginating – 0 1 0.01 0.06 fromthegroundwaterasconcentratedflow PDOC DOCproductionparameter mgL−1 0 15 1.79 1.57 aDOC DOCvariabilityconstant – 0 2 0.92 1.05 PDIN DINproductionparameter mgL−1 −5 10 −1.35 0.11 SPH,DIN phaseofannualDINproduction d 0 365 0 2 ADIN amplitudeofannualDINproduction mgL−1 0 10 3.36 1.84 Gmax,SO4 equilibriumconcentrationofSO4inmatrix mgL−1 0 50 2.74 3.07 aGeo equilibriumconcentrationvariabilityconstant – 0 2 0.11 0.04 ∗ ∗ KGEweighted weightedmulti-objectivemodelperformance – 0 1 0.56/0.49 0.52/0.53 ∗ ∗ KGEQ,tot modelperformancefordischargeofentiresystem – 0 1 0.41/0.33 0.35/0.42 ∗ ∗ KGEQ,W modelperformancefordischargeofweir – 0 1 0.67/0.62 0.61/0.66 ∗ ∗ KGEDOC modelperformanceforDOCconcentrations – 0 1 0.38/0.35 0.43/0.32 ∗ ∗ KGEDIN modelperformanceforNO3concentrations – 0 1 0.48/0.40 0.48/0.45 ∗ ∗ KGESO4 modelperformanceforSO4concentrations – 0 1 0.74/0.62 0.64/0.65 ∗Calibration/validationwithothersample. andtwo-thirdsforthedischargeofweir1,whoseobservation viceversa.Aparametersetisregardedstablewhenthecal- precisionwasregardedtobemorereliablethantheup-scaled ibrationwithbothsamplesyieldssimilarparametersetsand discharge.KGEisdefinedas theirKGEconcerningdischargeandthesolutesdoesnotre- ducesignificantlywhenapplyingthemtotheothersample. q KGE=1− (r−1)2+(α−1)2+(β−1)2, (5) σ µ 3.3 Changeinhydrochemicalbehaviourwiththe withα= S and β= S, (6) stormyperiod σ µ O O wherer isthelinearcorrelationcoefficientbetweensimula- Afterthemodelevaluation,weusethedifferentcomponents tionsandobservations,andµ /µ andσ /σ arethemeans of the KGE in Eqs. (5) and (6) to explore the impacts of S O S O andstandarddeviationsofsimulationsandobservations,re- the storm disturbance period on the hydrochemical com- spectively.αexpressesthevariabilityandβ thebias. ponents. By assuming that the model is able to predict to Tocheckforthestabilityofthecalibratedparameters,we hydrochemical behaviour that prevailed without the impact perform a split-sample test (Klemeš, 1986). Since the pre- of the storms adapting the hydrochemical parameters of the disturbance time series was too short to be split into two model in Eqs. (3)–(4) and analysing the difference between equally long periods, we perform a both-sided split-sample theadaptedhydrochemicalsimulationsandthenon-adapted test by bootstrapping two independent 4-year time series of simulations,weareabletoquantifythechangeinsolutemass observations (first sample: discrete sampling of 50% of the balanceduetothestormimpact.Wedefinethetimespanfor valuesofeachobservedtimeseries;secondsample:remain- our adaption as the time when the different components of ing 50% of the observations). We calibrate our model with KGE exceed the range of their pre-disturbance variability. thefirstsampleandevaluateitwiththesecondsample,and Duringthistimeperiodwecompensatefortheapparentde- www.biogeosciences.net/13/159/2016/ Biogeosciences,13,159–174,2016 164 A.Hartmannetal.:Model-aidedquantificationofdissolvedcarbonandnitrogenrelease viations by adapting the hydrochemical parameters. This is in the soil and sulfides in the dolomite. Its variability con- donetwice–oncebymanualadaptionandanothertimeus- stant is quite low (<0.1). Weighted KGEs, as well as their ing an automatic calibration scheme. Their new values will values for the individual simulation variables, are relatively indicatechangesintheseasonality,productionorinterannual stable.Overall,calibrationonbothsamplesprovidedsimilar variations. parameter values. Due to its higher stability concerning the evaluation period, we chose the second sample for further 3.4 Transittimedistributions analysis. Thedischargesimulationsfollowadequatelythevariations The signal of the storm impact will travel at various veloc- intheobservations(Fig.4),althoughsomesmalleventsare ities and via pathways through the karst system. While fast not reproduced by the model and although the simulations flowpathsandsmallstorageswilltransportthesignalrapidly oftheweir’sdischargetendtounderestimatepeakflows.No to the system outlet, slow pathways and large storages will obviousdifferencescanbeseenbetweenthepre-disturbance delay and dilute the signal. Transit time distributions indi- andwinddisturbanceperiod.Thehydrochemicalsimulations catehowfastsurfaceimpactstravelthroughthehydrological tend to follow the observations as well (Fig. 5). However, system. We derive transit time distributions from the model thereissometimessomeunderestimationoftheDOCpeaks by performing a virtual tracer experiment with continuous for the pre-disturbance period. The DIN simulations appear injection over the entire catchment at the beginning of the to be more precise during the pre-disturbance period, but impact of the stormy period. When a model compartment there is a systematic underestimation when the disturbance reaches 50% of the tracer concentration is considered me- takesplace. diantransittime.Thethus-derivedtransittimeswillelaborate how the hydrological system propagates the signal through 4.2 Modelperformanceduringthewinddisturbance the system including all slow and fast pathways as defined period by Eqs. (12) and (18). As for DIN and DOC we assume completeandinstantaneousmixingwitheachmodelstorage Thereisadeviationbetweenpre-disturbanceanddisturbance (soil, epikarst, and groundwater) at each compartment; the period simulated and observed variability and bias for DIN timethatwerefertoas“meantransittime”ofamodelcom- (Fig.6).AsimilartendencycanbefoundforDOC.However, partmentisthetimethevirtualtracerneedstopassthrough onlyforDINarethedeviationsdifferenttothevariationsal- theparticularmodelstorage.Incombinationwiththefluxes readyfoundduringthepre-disturbanceperiod(whichisalso that are provided from each of the model compartments, it thecalibration/validationperiod).ThevariationsinDOCap- ispossibletoquantifythefractionalcontributionoffastand pear to be systematic, too, but they fall within its ranges of slowflowpaths,respectively.Wewillapplythevirtualtracer variabilityduringthepre-disturbanceperiod. from the previously assessed beginning of the impact un- til the end of the time series to assess the transit time dis- 4.3 AdaptionofNparametersforthewinddisturbance tribution. In addition, we apply a second virtual tracer that period also lasts only for the disturbance period (as estimated in Sect. 3.3) to evaluate the filter and retardation potential of The very first signs of the impact were found on 1 May thekarstsystem. 2007, and they lasted to the end of the hydrological year 2010/11. In a first trial (Table 2), the model parameters for the DIN production were adapted manually to compensate 4 Results for the changes in observed DIN concentrations with a fo- 4.1 Modelperformance cus on reducing the difference indicated by the bias, β, and variability,α,componentsofKGE .Inasecondtrial,we DIN Table1showsthecalibratedparametersforthetwosamples. useanautomaticcalibrationschemetoachievetheoptimum Theyindicateathicksoilandarelativelythinepikarst.The KGE .AsindicatedbythehighestKGE(Table2),theau- DIN dynamics expressed by the storage constants indicate days tomaticcalibrationprovidedthehighestKGE ,butthisis DIN andweeksfortheconduits(modelcompartmenti=Z)and achieved by improving variability α and correlation r. Al- the epikarst, respectively. The distribution coefficient of the mostnoimprovementisreachedforthebiasβ.Eventhough groundwaterislargerthanthesoil/epikarststorageconstant. resulting in a slightly lower improvement of KGE the DIN ForDOCandDINthereareanaturalproductionratesof1.6– manualcalibrationresultsinamuchmoreacceptablereduc- 1.8and−1.35−0.1mgL−1,respectively.TheDOCdistribu- tion in the bias (Fig. 6). Its parameter values showed a pro- tion coefficient is between 0.9 and 1.1. The phase shift and duction rate, P , of DIN almost 2mgL−1 larger than the DIN amplitudeforDINshowedthatthereisaseasonalvariation pre-disturbancevalue;anamplitude,A ,around1mgL−1 DIN in DIN net production, with its maximum release at April smaller; and a phase shift, S , towards a week earlier PH,DIN eachyearforbothofthesamples.SO2−isdominatedbythe intheyear,resultinginamoreacceptablesimulationofDIN 4 concentration in the precipitation input with some leaching dynamicsduringthedisturbanceperiod(Fig.7). Biogeosciences,13,159–174,2016 www.biogeosciences.net/13/159/2016/ A.Hartmannetal.:Model-aidedquantificationofdissolvedcarbonandnitrogenrelease 165 Figure4.Observedvs.simulateddischargesfortheentirekarstsystemandweir1. Figure5.Observedvs.simulated(a)DOCand(b)DINatweir1. 4.4 Transittimedistributions 5 Discussion Thetransittimedistributionsshowthatthesoilandepikarst 5.1 Reliabilityofcalibratedparametersandmodel systemreactsquiterapidlytothevirtualinjection.Fiftyper- simulations cent of the injection concentration is reached within ∼60 days (Fig. 8a), while most of groundwater system requires Most of the calibrated model parameters are in ranges that ∼100daystoreach50%oftheinjectionconcentration,with are in accordance with other modelling studies or field ev- afewflowpathsrequiringupto300days(Fig.8c).Asimilar idence. General differences between the calibrated parame- behaviour is found when the impact ends (Fig. 8b, d). This tervaluesoftheboth-sidedsplit-sampletestmaymostlybe behaviour also shows that some of the slowest flow paths due to the comparatively low resolution of the hydrochemi- justreachtheinputconcentrationbeforetheystarttodecline calvariables(SO ,DOCandDIN),whichactuallyincreased 4 again. asaresultofthebootstrappingprocedure.However,thegood multi-objectivesimulationperformanceofthemodel,aswell asitsevaluationbythesplit-sampletest,indicateanoverall acceptable performance of the model. At almost 3–8 days theepikarststorageconstantisinaccordancewithfieldstud- ies on the epikarst storage behaviour that found retention times of some days to a few weeks (Aquilina et al., 2006; www.biogeosciences.net/13/159/2016/ Biogeosciences,13,159–174,2016 166 A.Hartmannetal.:Model-aidedquantificationofdissolvedcarbonandnitrogenrelease Table 2. Calibrated pre-storm parameters for DIN dynamics and twoscenariosforadaptingthemodelatthestormyperiod. Parameter Unit Calibrationtype pre-storm manual automatic PDIN mgL−1 0.11 2.10 0.00 SPH,DIN d 2.00 9.00 23 ADIN mgL−1 1.80 0.70 2.63 ∗ KGE – 0.29 0.41 0.46 DIN ∗ variabilityα – 0.75 1.04 1.05 DIN ∗ biasβ – 0.70 1.01 0.83 DIN ∗ correlation – 0.40 0.41 0.49 DIN ∗For2006/07–2011/12. byplantsorsoilorganismsandthattheproductionperiodis Figure6.IndividualcomponentsoftheKGE:(a)ratioofsimulated shorterbutmorepronouncedduetoitslargervalueofampli- andobservedvariabilities,(b)ratioofsimulatedandobservedaver- tude.However,weexpectthesedifferencestobeminorsince agevalues,and(c)theircorrelationforthewinddisturbanceperiod; thephaseshiftS ofbothcalibrationsamplesisalmost forcomparisontheKGEcomponentsandtheirinterannualvariabil- PH,DIN the same, as well as their annual maximum (P +A ) ityarealsoshownforthepre-stormperiodandafterthecorrection DIN DIN of 2.01 and 1.95mgL−1. This behaviour indicates a maxi- oftheDINproductionmodelparametersduringthewindperiod. mumofDINproductionandleachingatthetimeoftheyear when snowmelt reaches its maximum (March to April) and whenDINuptakebyplantsisstilllow(Jostetal.,2010).The Perrinetal.,2003).Thesoilandtheepikarststoragecapac- dissolutionequilibriumconcentrationsof2.7–3.1mgL−1for ity are quite large. These high values may be explained by SO2− indicate the abundance of the precipitation input, ox- 4 structural errors of the model that result in unrealistic cali- idationofsulfides(e.g.pyrite)inthedolomiteandtracesof bratedparametervalues,inparticularpossibleparameterin- evaporatesinthesmallPlattenkalkoccurrences(Kraliketal., teractionsbetweentheirstoragecapacitiesandstoragecoef- 2006). ficients.Sincethesoilandthevegetationcontrolsthefraction ofrainthatislosttoevapotranspiration,thishighcalibrated 5.2 Impactofstorms valuemightbeduetotreerootsrangingthroughthesoilinto theepikarst(Heilmanetal.,2012)orrockdebris(Hartmann The deviation between simulated and observed time series etal.,2012a). (Fig. 5) already indicates that DIN is the only solute that Similar to the epikarst storage constant, the conduit stor- shows a clear impact of the storms. This is further corrob- ageconstant,K ,is,withitsvalueof1.1days,intherange oratedbyconsideringtheindividualcomponentsofKGEin C ofpreviousmodellingstudies(Fleuryetal.,2007;Hartmann Fig. 6. It is well known that nitrate leaching to the ground- etal.,2013a).Thehighvaluesoftheepikarstvariabilitycon- waterincreasessharplyaftertreedamage(dieback)inforests stant and the groundwater constant indicate a low develop- whereNisnotstronglylimited(Bernaletal.,2012;Griffinet ment of preferential flow pathsin the rock, which is typical al.,2011;Huber,2005).SuchdisturbancesdisrupttheNcy- fordolomiteaquifers(FordandWilliams,2007).Alowde- cle.ThelossoftreeNuptakefavoursnitrificationofsurplus + gree of karstification was already known for our study site NH bymicroorganisms.Moreover,above-(i.e.foliage)and 4 (Jostetal.,2010)andthecalibratedrechargeareasfallwell belowground(i.e.fineroots)litterfromdeadtreesenhances withintherangesfoundinpreviousmodellingstudies(Hart- themineralizationoforganicmatter,ammonificationandni- mannetal.,2012a,2013c). trification. Both processes are accelerated by increased soil The hydrochemical parameters mostly show realistic moisture and soil temperature due to the loss of the forest values. A DOC production parameter, P , of ∼1.6– canopy.Subsequently,leachingofNincreaseswithincreased DOC 1.8mgL−1 resulted in realistic simulated concentrations at seepage fluxes due to decreased interception and water up- the weir. For DIN production the two calibration samples takebytrees.SincethesimpleDINroutineofthemodelcan- result in values of −1.4 and 0.1mgL−1, going along with nottakeintoaccountsuchchanges,theunderestimatedDIN amplitudesof3.4and1.8,respectively.Hence,thereappears concentrationsandtheiramplitudeshowtheeffectofforest to be some correlation between the production and ampli- disturbance on the leaching of DIN from the studied catch- tude parameters, P and A . Negative values indicate ment.Thereisalsoanapparentlysystematicdeviationofthe DIN DIN that, during some periods of the year, all DIN is consumed DOCvariabilityα.Butitsvariationsduringthepre-stormpe- Biogeosciences,13,159–174,2016 www.biogeosciences.net/13/159/2016/ A.Hartmannetal.:Model-aidedquantificationofdissolvedcarbonandnitrogenrelease 167 Figure 7. Observed and simulated DIN dynamics using the pre-storm parameters (red line), the scenario 1 parameters derived from the deviationsassessedbytheKGEcomponents(orangeline),andthescenario2parametersderivedbysystematicvariation(dark-redline). Figure8.Meantransittimesfor(a)thesoilandepikarstand(c)thegroundwaterstoragederivedbyaninfinitevirtualtracerinjectionstarting withthebeginningofthewinddisturbanceperiod,and(b)thereactionofthesoilandepikarstand(d)thegroundwaterstorageastheimpact ends. riodaresimilarlylarge,thuspointingtoanegligibleeffectof pathsarelikelyduetothesteepnessofthecatchmentslopes forestdisturbanceonDOCleaching.Numerousstudieshave (Boyer et al., 1997; Sakamoto et al., 1999; Terajima and identified the forest floor as a DOC source (Borken et al., Moriizumi, 2013). The missing signal of forest disturbance 2011; Michalzik et al., 2001). Windthrow generally causes on DOC concentrations at weir 1 even shortly after the dis- a (short-term) pulse of above- and belowground litter (Har- turbancemaybeduetotheminorextensionofthedisturbed monetal.,2011).Thereby,mineralizationofthesurpluslit- area, the minor increase of surface and shallow subsurface ter input concurrent with improved soil climatic conditions flow due to the relative low slope of the disturbed area, the likely increased the leaching of DOC from the forest floor buffering of increased topsoil DOC leaching due to absorp- (Fröberg et al., 2007; Kalbitz et al., 2007). Concurrent, in- tion of DOC within the subsoil (Borken et al., 2011; Huber creasedsoilwater,surfaceandshallowsubsurfaceflowmay et al., 2004), missing DOC-rich riparian source areas (i.e. favour increased soil DOC leaching to downslope surface wetlands, floodplains) and the reduction in pre-disturbance waters(Monteithetal.,2006;NeffandAsner,2001;Sander- organic matter input to soil (i.e. litter, root exudates; Hög- manetal.,2009).Inmountainouscatchmentsthelatterflow berg and Högberg, 2002). Theoretically, hydrological pro- www.biogeosciences.net/13/159/2016/ Biogeosciences,13,159–174,2016

Description:
Publisher's PDF, also known as Version of record. License (if A. Hartmann1,2, J. Kobler3, M. Kralik3, T. Dirnböck3, F. Humer3, and M. Weiler1.
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.