Données Landsat

Landsat 7 ETM+ est un capteur multi-spectral permettant de collecter des données thermiques infrarouges à haute résolution spatiale. L’instrument (Enhanced Thematic Mapper Plus) fournit des images à une résolution spatiale de 15 m à 60 m et un angle de visée au nadir. La bande Infrarouge thermique a une résolution spatiale de 60 m, les bandes visibles et les bandes moyennes infrarouges ont une résolution spatiale de 30 m. A la différence d’ASTER, Landsat est un capteur opérationnel qui acquiert des images en routine tous les 16 jours. Dans la Fig. 6.7 on présente un schéma comparatif entre ASTER et Landsat en termes de résolution spectrale. Dans ce travail, un ensemble de 12 images Landsat obtenues entre 2007-2008 ont été utilisées. Ces images correspondent aux données brutes à corriger plus tard. Les dates de chaque image Landsat sont présentées dans la Fig. 6.5.

Figure6.7 – Comparaison des bandes spectrales d’ASTER et Landsat ETM+. Les boîtes rectangulaires (rouge: ASTER, noir: Landsat ETM+) montrent les canaux des capteurs. La résolution spatiale est indiquée au-dessus des boîtes. La courbe de couleur montre la transmission atmosphérique en fonction de la longueur d’onde (Kääb et al, 2003).

PARTIE IV

Développement et exploration d’un

modèle TSVA pour la vigne

Introduction à la Partie IV

Après la révision des formalismes mathématiques utilisés pour modéliser localement les transferts sol-végétation-atmosphère dans des couverts complexes comme la vigne, on propose dans cette quatrième partie un modèle parcimonieux simulant le bilan hydrique d’une vigne partiellement enherbée en termes d’évapotranspiration totale et de teneur en eau du sol. Ce modèle considère trois sources d’échanges: le feuillage principal (vigne), le sol enherbé et le sol nu. Un module basé sur le bilan d’énergie est utilisé pour décrire les transferts d’énergie et de masse entre le continuum sol-plante et l’atmosphère avec un pas de temps horaire. Il est couplé avec un module de bilan hydrique du sol à pas de temps journalier. Le modèle utilise comme entrées des données météorologiques standard ainsi que des paramètres décrivant les caractéristiques de développement du feuillage, de l’herbe et du sol. Le modèle est calibré à l’aide de la méthode Multi-objective Calibration Iteractive Process (MCIP)et ensuite il est validé pour l’évapotranspiration et l’humidité du sol sur un ensemble de données collectées dans un vignoble enherbé du sud de la France. L’exercice de validation est double. Il se concentre d’abord sur l’évolution horaire de l’évapotranspiration provenant du module de bilan d’énergie uniquement, forcé avec des variables météorologiques, le rayonnement net et l’humidité du sol. Ensuite, le module de bilan d’énergie de surface couplé avec le module de bilan hydrique du sol est validé avec des mesures Eddy Covariance et d’humidité du sol.

Dans une deuxième étape (Chapitre 8), des méthodes d’analyse de sensibilité et de calibration multi-critère sont utilisées pour explorer le modèle et approfondir la pertinence des valeurs des paramètres obtenues lors d’une calibration interannuelle. Par la suite, dans une optique de spatialisation par télédétection, nous envisageons différents scénarios de calage en fonction des observations disponibles. Puisque au niveau de la démarche on travaille sur des parcelles agricoles relativement petites, on s’intéresse à des observations satellite hectométriques. Nous considérons à la fois les données Landsat disponibles tous les 15 jours, et aussi l’instrument ASTER qui fournit des observations sur demande.

Chapitre 7

A three-source SVAT modeling of

evaporation: Application to the

ContentslistsavailableatScienceDirect

AgriculturalandForestMeteorology jou rn al h om ep a ge :w w w . e l s e v i e r . c o m / l o c a t e / a g r f o r m e t

A three-sourceSVAT modelingofevaporation:Applicationtothe seasonaldynamicsofagrassedvineyard

CarloMontesa,∗,Jean-PaulLhommea,JérômeDemartyb,LaurentPrévotc,FrédéricJacoba

aInstitutdeRecherchepourleDéveloppement,UMRLISAH,34060Montpellier,France bInstitutdeRecherchepourleDéveloppement,UMRHSM,34095Montpellier,France cInstitutNationaldelaRechercheAgronomique,UMRLISAH,34060Montpellier,France

ar tic le info

Articlehistory: Received12June2013

Receivedinrevisedform5February2014 Accepted13February2014

Keywords: Latentheatflux Multi-source Sparsevegetation Soilwaterbalance Seasonalcourse

abstr act

AparsimoniousandversatileSoil–Vegetation–AtmosphereTransfer(SVAT)modelisproposedforthree componentvineyards,whichincludesvinefoliage,grassedsoilandbaresoil.Athree-sourceenergy balanceapproachdescribestheenergyandmasstransferbetweenthesoil–plantcontinuumandthe loweratmospherewithanhourlytimestep.Itiscoupledwithasoilwaterbalancemodulerunning withadailytimestep.Themodelmakesuseofstandardmeteorologicaldatatogetherwith parame-tersdescribingfoliagedevelopment,grassandsoilcharacteristics.Themodeliscalibratedbymeans oftheMulti-objectiveCalibrationIterativeProcess(MCIP)algorithmandnextvalidatedfor evapora-tionandsoilmoistureoveradatasetcollectedinaSouthernFrancegrassedvineyard.Thevalidation exerciseistwofold.Itfocusesfirstonthedailycourseofevaporationderivedfromthesurfaceenergy balancemoduleonly,forcedwithmeteorologicalvariables,netradiationandsoilmoisture.The compari-sonagainstEddyCovariancemeasurementsshowsagoodagreement(R2=0.96andRMSE=14.0Wm−2). Next,asimulationcouplingthesurfaceenergybalancemodulewiththesoilwaterbalancemoduleis val-idatedoverEddyCovarianceandsoilmoisturemeasurements.Simulationsthroughouttwocontrasting growingseasonsprovidegoodestimatesofdailyevaporation(R2=0.90andRMSE=0.43mmd−1)and soilwatercontent(R2=0.98andRMSE=6.95mm).Modelinaccuraciesarisemainlyunderconditionsof strongsurfacerunoff.Resultsalsosuggestthattheparameterizationsrelatingthesurface-atmosphere modulewiththesoilmodule(i.e.stomatalresistance)shouldbecarefullyexaminedunderwaterstress conditions.Finally,themodelversatilityisaddressedthroughasetofsimulations.Itappearsthatthe modelingapproachallowsassessingtheseasonalwaterbalanceofvineyardswithdifferentstructure (grassfractionordistancebetweenrows)andthatitcouldbeappliedtosimilarcroppingsystems.

©2014ElsevierB.V.Allrightsreserved.

1. Introduction

Progressintheoreticalandappliedresearchaimingat accu-rately assessing crop waterconsumption in both rain-fed and irrigated conditions is anessential issuefor agriculturalwater management.Sinceevaporationmeasurementsarescarce, oper-ationalformulationstoestimatewaterconsumptionatfieldscale are necessary(Trambouze etal.,1998;Spanoetal., 2009).For viticultureregionsinMediterraneanandsemi-aridenvironments, actualevaporationrepresentsamajorcomponentofsurfacewater balance,reachingupto70%oftheyearlyprecipitation(Moussa

∗ Correspondingauthorat:UMRLISAH,2placePierreViala,34060Montpellier, France.Tel.:+330785408326.

E-mailaddress:ccmontesv@gmail.com(C.Montes).

etal.,2007).Knowledgeofactualevaporationisalsoofinterest inviticulture,inordertoassessandhandletheinfluenceofsoil waterdeficitongrapevineyieldsandberrycomposition(Vaudour, 2003;Pellegrinoetal.,2005).Nevertheless,thephysical repre-sentationofthesoil–plant–atmospheresystemingrapevinesisa complexissue,becausethesparsestructureofvineyardsimposes toconsiderboththefoliageandtheunderstory,whichrequires multi-sourcemodeling.

Themostfrequentlyusedmulti-sourceevaporationmodelis theonefirstdevelopedbyShuttleworthandWallace(1985)(S–W model) and extendedbyChoudhury andMonteith (1988) and

ShuttleworthandGurney(1990).Thismodelcorrespondstoan extensionofthebig-leafmodelofPenman–Monteith(Monteith, 1965)intotwointeractingevaporativelayers:themainfoliage andtheunderlyingsubstrate.Subsequently,theS–Wmodelwas upgraded by Brenner and Incoll (1997) (“clumped” model)to http://dx.doi.org/10.1016/j.agrformet.2014.02.004

C.Montesetal./AgriculturalandForestMeteorology191(2014)64–80 65

Ai available energy for each component vs and bs (Wm−2)

A_f availableenergyforthemainfoliage(Wm−2) c radiationextinctioncoefficientbycanopy cp specificheatofairatconstantpressure(Jkg−1K−1) CR3 capillaryriseintoreservoir(3)(mm)

d displacementheight(m)

D1 drainagefromreservoir(1)to(3)(mm) D2 drainagefromreservoir(2)to(3)(mm) D3 deeppercolationfromreservoir(3)(mm) Da vaporpressuredeficitatreferenceheight(Pa) Dm vaporpressuredeficitatmeancanopysourceheight

(Pa)

ea vaporpressureatreferenceheight(Pa)

e*(Ti) saturatedvaporpressureattemperatureTi(i=f,vs, bs)(Pa)

F_bs fractionofbaresoil(=1−Fvs) Fvs fractionofvegetatedsoil F₁ F_bs

F₂ Fvs

Gvs soilheatfluxofvegetatedsoil(Wm−2) Gbs soilheatfluxofbaresoil(Wm−2) Ivs infiltrationtermforvegetatedsoil(mm) I_bs infiltrationtermforbaresoil(mm)

K(z_h) turbulentdiffusivityatcanopyheight(m2s−1) L Monin-Obukhovlength(m)

LAI_f leafareaindexofmainfoliage(m2m−2)

CLAIvs clumpedleafareaindexofvegetatedsoil(m2m−2) n parameterwithvalueof1foramphistomatousand

2forhypostomatousfoliage

ra aerodynamicresistancebetweenthemeansource height(zm)andthereferenceheight(zr)(sm−1) r_a,i aerodynamicresistancebetween theevaporative

source(i=vs,bs)andmeansourceheight(z_m,sm−1) r_a,f,h bulkboundary-layerresistanceof thefoliagefor

sensibleheat(sm−1)

rs,i surfaceresistance(stomatalorsoilsurface)foreach source(i=f,vs,bs)(sm−1)

Rn netradiationofthewholecanopy(Wm−2) ua windspeedatreferenceheight(ms−1) z_h heightofthemainfoliage(m) zm meansourceheight(m) zr referenceheight(m)

z₀ roughnesslengthformomentumofmainfoliage(m) zi

0 roughness length for momentum of vegetated (i=vs)orbaresoil(i=bs)(m)

z1 depthofsoilreservoir(1)(m) z2 depthofsoilreservoir(2)(m) zR vinesrootingdepth(m) zG watertabledepth(m)

psychrometricconstant(PaK−1)  latentheatofvaporization(Jkg−1)

1 slopeofthesaturatedvaporpressurecurveatair temperature(PaK−1)

 airdensity(kgm−3) s solarzenithangle(radians)

accountforthreesourcesofevaporationafterdividingthe under-storyintoabaresoilfractionandasoilfractionbelowthemain foliage,andalsobyVerhoefandAllen(2000)toaccountforfour sourcesofevaporation.Thetwo-andthree-sourceformalismswere revisitedbyLhommeetal.(2012)toproposemoreconciseand accurateformulationsand toaccountforfoliagemorphological

characteristics(amphistomatous versushypostomatous leaves). All thesemodelsare based onthe diffusiontheory (K-theory) forenergyandmasstransferwithintheloweratmosphere.More complexmodelsbasedonhigherorderLagrangianandEulerian dispersionprocessescanbefoundintheliterature:theyallowa bet-terrepresentationofvegetation–atmosphereturbulenttransfers (Raupach,1989;Yi,2008),buttheircomplexityanddata require-mentmakethemdifficulttouseinapracticalmodelingframework. Ithasbeenshown,further,thatthediffusiontheoryisappropriate torepresentthemicroclimateatcanopyscaleincomparisonwith Lagrangianrepresentations(VandenHurkandMcNaughton,1995; Wuetal.,2001).

Oneofthefirstmodelstoestimatevineyardevaporationis theoneproposedbyRiouetal.(1989,1994).Itisnota multi-sourcemodel:vineyardevaporationunderunstressedconditions isexpressedasasimplefunctionofpotentialevaporationandsolar radiationinterceptedbythecanopy.Thismodelwasextendedlater byTrambouzeandVoltz(2001),whoderivedabilinearrelationship relatingtheratiobetweenvineyardactualandmaximum transpira-tiontotheaveragesoilwaterstorage.Subsequently,severalauthors haveapplied themulti-sourceresistance-basedformulationsto assessvineyardevaporation.First,wehavetomentionthework byRanaandKaterji(2008),whereasimplesingle-sourcemodel (Penman–Monteith)wasappliedtovineyardstrainedonoverhead system.InanearlierworkbySene(1994),themorecomplexS–W modelwasappliedwiththepurposeofinterpretingenergy bal-ancemeasurementsoverasparsevineyardinsouthernSpain.More recently,anappropriaterepresentationoftotallatentheatflux fromadrip-irrigatedvineyardincentralChilewasobtainedby

Ortega-Fariasetal.(2007)byapplyingthesameS–Wmodel.In addition,Poblete-EcheverriaandOrtega-Farias(2009)adaptedthe so-called“clumped”modeltodripirrigationoverthesameregion ofChilebydividingthesubstrate(baresoil)intoadryandawet (irrigated)portion.Zhangetal.(2008)comparedthesetwomodels (S–Wandclumped)againstBowenratioestimatesinasemi-arid vineyardofChina:theyconcludedthattheclumpedmodelwas moresuitabletoestimatetotalvineyardevaporationthantheS-W model.Onthesamebasis,Zhangetal.(2009)elaborateda multi-sourceS–Wtypemodeltosimulatetheevaporationfromavineyard underpartialroot-zoneirrigation,takingintoconsideration differ-entpatchesofsoil.

Allthesevineyardevaporationmodels,however,donottake intoaccountthecommonpracticeofmaintainingapermanentor semi-permanentgrasscover.Thisconsistsinaseededornatural grasscoverinbetweenvinerows,maintainingbaresoilonthe rows.Thispracticeisincreasinglyusedbecauseithasseveral posi-tiveimpacts,suchasthereductioninrainfallerosivepotentialand surfacerunoff,thereductioninnutrientlixiviation,thedecrease invinevigorandgrapeproduction(whichimprovesgrapes qual-ity)andtheimprovementsinsoilstructureandtrafficabilityafter rainfallevents(PradelandPieri,2000;MorlatandJacquet,2003; Celetteetal.,2005,2008;Gaudinetal.,2010).Ascomparedwith the traditional baresoil grapevine cultivation, the grass cover affectsenergyandwaterbalancesincesurfacealbedo,net radia-tionpartitioning,waterconsumptionandinfiltrationaremodified (Rodriguez-Iturbe,2000;ZhangandSchilling,2006;Centinarietal., 2012).Forinstance,in arecentwork byHollandetal.(2013)

ongrassedvineyard,significantdifferenceswerefoundbetween grassedandbaresoilenergypartitioning.Therefore,thisgrasscover componentshouldbeconsideredintoamodelingformulation.

Inaddition,mostofthevineyardevaporationmodelsmentioned aboveonlyconsiderabove-groundprocesses(i.e.vegetationand soilsurface),whichinteractwithsoilwaterthroughthe param-eterizationofastomatalorsubstrateresistancetoevaporation, inthebestcase.Thus,theydonotallowthetemporaldynamics ofvineyardevaporationtobeadequatelysimulatedthroughout

theseason. However,modelshavebeendeveloped tosimulate soilwaterbalanceofvineyards.Sene(1996)wasthefirstto com-bineasimplesoilmoisturemodelwithatwo-component(S–W) representationofvineyardevaporationinordertoestimatethe long-termwaterbalanceofasparsevinecropgrowingunder semi-aridconditions.Lebonetal.(2003)alsoperformedsimulationsof theseasonaldynamicsofsoilwaterbalanceinvineyardsbyusing asinglereservoirsoilmodelalongwiththeRiouetal.(1989,1994)

approachforgrapevinetranspirationcoupledwithastress func-tioninvolvingsoilwateravailability(TrambouzeandVoltz,2001).

Celetteetal.(2010)extendedthemodelofLebonetal.(2003)to simulatethewaterbalanceofanintercroppedvineyard consider-inganadditionalandseparatesoilcompartmentunderthecover crop.Galleguillosetal.(2011)alsousedthemodelofRiouetal. (1989,1994),butcoupledwiththeHYDRUS-1Dsimulationmodel ofsoilwatertransfers.Althoughrealisticresultshavebeenobtained withthistypeofsoilwaterbalancemodel,theevaporationprocess remainspoorlyrepresentedandamorerealisticapproachbased uponmicrometeorologicalresistance-typemodelscoupledwith soilwatermodelsappearstobenecessary.

With regards to the elements discussed above, the main objective ofthepresentworkistodevelopaSoil–Vegetation– Atmosphere Transfer (SVAT) modelwhich simulates the vine-yardevaporationdynamicsatseasonalscaleandaccountsforthe grasscoverasaviticulturalpractice.Itcombinesa comprehen-sivemicrometeorologicalthree-sourcemodelofevaporationwith athreereservoirsoilwaterbalancemodel.Theformulationis ver-satileenoughtoallowtheassessmentofevaporationratefrom differentmixedcroppingsystems,insofarasspecies-specific bio-physicalparametersandphysicalsoilpropertiesdescriptorsare adjustedtotheprevailingconditions.Theformulationisalsokept asparsimoniousaspossibletoforeseeitsapplicationattheregional extentwhileaccountingfortheinter-fieldvariability.Theplanis asfollows.InSection2,theSVATmodelisfullydescribed, sepa-ratingtheevaporationmodelfromthesoilwaterbalancemodel. Section3detailsthestudyarea,theexperimentaldata,themodel implementationandthestrategyforcalibration.Section4shows acomparisonofmodelsimulationsagainstgroundtruthdatato validatethemodelandsomesimulationsarepresentedtoshow theversatilityofthemodelthroughitsaptitudetorepresent dif-ferentviticulturalpractices(proportionofgrassedsoilanddistance betweenrows).Finally,modelresultsandlimitationsarediscussed inSection5.

2. Modeldevelopment

2.1. Representingthevine-grass-soilsystem

The soil–plant–atmosphere continuum is represented as a three-sourcesystemthatincludesthevinecanopy(mainfoliage) andacompositesubstratemadeofagrasscoverandbaresoil.This rainfedCabernetSauvignonvineyardisconductedinrows.Vine leafareaindex(LAIf)variesthroughoutthegrowingseasonfrom 0toamaximumvalueandnextfallsbackto0duringsenescence. Thegrasscoverispresentonlyononeinter-rowoutoftwo.Itis characterizedbytheconceptofclumpedLAI(CLAIvs),definedas grassleafareaindexperunitareaofgrasscover:CLAIvs=LAIvs/Fvs, Fvsbeingtheproportionofvegetatedsoil.Grassexhibitsaseasonal dynamics:itsgrowthisinitiatedbyautumnprecipitationsandit driesoffinearlysummerasaresultoflargewaterstress.Thebare soilfraction(Fbs)coverstherestoftheinter-rowandthesoilbelow thevines.Thevine-grass-soilsystemcorrespondstoathree-source systembetweenautumnandearlysummerandtoatwo-source systeminsummer,whenthesubstrateismadeofbaresoilanddry grassonly(F_bs=1).

Thethreecomponentsshouldbeconsideredseparatelybecause theyhavedifferentphysicalandgeometricalfeatureswhichaffect energyandmasstransfers.Nevertheless,thevinepatchesarenot largeenoughtoadoptapatchrepresentationofthewhole sys-temandconsequentlyalayerrepresentationispreferred(Boulet etal.,1999;LhommeandChehbouni,1999;Andersonetal.,2005). Giventhecompositenatureofthesubstrate,themodeling com-binesalayerapproachforthevine-substratesystemwithapatch approachforthesubstrate(grasscover+baresoil),asrepresented inFig.1andexplainedinLhommeetal.(2012,section3.1).Indeed, whilesubstrateandvineareinterrelatedintheverticaltransfer ofheatandwatervaporwithasoleaerodynamicresistanceabove thewholecanopy,grasscoverandsubstrateareassumedtoact separatelyvis-a-visthecanopysourceheight.

Soilmoisturedynamicsisrepresentedbyabuckettypemodel madeofthreereservoirsinrelationwiththethreecomponentsof theevaporationmodel:adeepreservoircorrespondingtothevine rootingsystem(∼2m)andtwoshallowreservoirscorresponding tothetwosubstratecomponents.Themaininputofthesystem cor-respondstotheinfiltrationofwaterfromprecipitation.Drainage processescontrolthewatertransfersbetweenreservoirs. Evapo-rationisthemainoutputanddeeppercolationactsasasecondary output.Capillaryrisefromthesaturatedzonebelowthedeep reser-voirisalsoconsidered,buthorizontalwatertransfers(runoff)are ignored.

TheSVATmodelconsistsofcouplingthesurfaceenergy bal-anceforplant-atmospheresystemandthesoilwaterbalancefor thesubsurfacesystem.Theevaporationmodelrunswithashort timestep(onehourorless)andisforcedwithmeteorologicaldata (airtemperatureandhumidity,windspeed,solarradiation)and vegetationdata(vineheightandleafarea,fractionofgrasscover). Thesoilwaterbalancerunsonadailytimestepandisforcedwith dailyprecipitation.Thewatercontentofeachsubsurfacereservoir isaninputtothecorrespondingevaporationcomponentsand con-verselytheevaporationcomponentsareinputsforthesoilwater balancemodule.

2.2. Evaporationmodel

Wedetailhereafterthesurfaceenergybalanceandsoilwater balancemodulesformingtheSVATmodelandtheircorresponding parameterizations.Forthenumerousformulationsconsideredin thissection,thevaluesofthecorrespondingparametersaregiven inTable1.

2.2.1. Formulationofevaporation

The total flux of latent heat (Et) is the sum of the con-tributionsfromthreesources:mainfoliage(Ef),vegetatedsoil (Evs)withrelativeareaFvsandbaresoil(Ebs)withrelativearea F_bs=1−F_vs.They areaggregatedfollowinga coupled(orlayer) approach(Fig.1):

Et=E_f+Evs+E_bs. (1) The three evaporation componentsreach themean canopy sourceheight(zm),assumedtobelocatedattheapparentsinkfor momentum(zeroplanedisplacementheightd+roughnesslength z0),wheretheymixtogetherformingthetotalevaporationat ref-erenceheight(zr)asitcanbemeasuredwithBowenratioorEddy