HYDROPERIOD OF DOÑANA MARSHES:NATURAL OR ANTHOPIC ORIGIN OF INUNDATION REGIME?
Díaz-Delgado,R.(1),Bustamante, J.(1),Pacios,F.(1)and Aragonés, D.(1)
(1)RemoteSensingGIS Laboratory, EstaciónBiológica deDoñana -CSIC. Avda. MariaLuisa s/n,Seville, 41013, Spain,Email:
ABSTRACT
Thereisamajorconcern oncurrentrestorationand management programmesbeingapplied over Doñana RamsarSite.Historicalinundation regimehasbeen largely changed according to human transformations. Onsitemeasurementsalongshort andrecentperiods have providedscantlocalinformationoninundation variability.Historicalinundationpatterns ona spatial basisare necessarytounderstandshortandlongtime effectsof humantransformations on naturalregimes.
Bycompilingalongtimeseriesof LandsatMSS, TM and ETMimages(1975-2006) we havemapped inundationlevelsforeveryavailablescene over Doñana wetlands. After coregistration, radiometric correction andtimeseriesnormalization, weappliedthebest semiautomathicmethod for discriminatinginundation levelsaccordingtoground-truth data.Then,we reconstructed for each hydrological cycle the hydroperiod value for every pixel. The value is calculatedastheminimum numberofdaysdetectedas floodedfortheavailable datesduring everycycle. Resultsrevealasignificant increase on theaverage numberof daysinundatedforthewholemarshandlocal hydroperiodchanges related tohumantransformations and rainfall variability. Hydroperiod maps are therefore helping to understand hydrological implications of futuremanagement decisions.
1. INTRODUCTION
Doñana NationalandNaturalParksarethelargest protected wetlandsinEurope(500km2). They have undergonestrong human-inducedchanges alongthe 20th century,asaconsequenceofdesiccation foragricultural land practices,waterchannelredraw, biodiversity conservation and environmental protection against the recent mineralspill waste in 1998.
Overall, changes in the original marsh (estimated around1500km2, seeFig.1)at thebeginningof20th century have led to the current 270 km2 of shallow water(lostof 82%).
Inadditiontowetlandareareduction,tidalinfluence, one ofthemaininundationdrivers,wasalsolimitedby theconstruction of a wall along the GuadalquivirRiver (Fig.2).Amongsuchdramaticchanges,thetoxicspill,
occurredin April 1998,led todevelopanambitious restorationprojectofDoñanamarshes,named‘Doñana
2005’.Itsmaingoalsaretorecoverthe quantityand qualityoftributarywaterstothefreshwatermarsh.In
thiscontext,morethanever,itisnecessarytoknowthe
historical inundation patterns, either spatial ortemporal, andtheir relationships withnaturalandanthropic conditions.
Details oninundation process, such as interannual and seasonalvariation,aswell asinfluenceofhuman transformationshave been longtimedemanded by decision-makersin orderto apply a scientific-based managementtoDoñanamarshes.Sofar, hydrological management has beenconducted on an“event-reaction” basis, leading totemporal solutions that became lately partof new problems.
Remotesensing has becomeavery usefultooltoeasily deriveinundationmapsand recentlytoreconstruct historicalhydroperiod [1]
Ground-water Runoff water flow Tidal water flow
SurfaceInlandWaterInput
Figure 1.Doñanawater inflows at the earlyXX century.
Inthiswork wepropose 2methods toestimate hydroperiodfrominundationmasksgeneratedbymeans
ofLandsatimageryforthelast30years.Hydroperiodis a critical parameter dependent on topography, wetlands connectivityandproximity toinflows [2].Itisthena suitable variabletoasses the ecologicaleffectsofany transformation sinceitconditions vegetationtypeand feedbackssedimentationprocesses.
Accordingtotheavailable timeseriesweanalyse changes on the hydroperiod and evaluate the main originof suchchanges.
Channelling & Rice cropping
areas coveringall the rangeof reflectancevalues present inthescene. ForMSSimageswefollowthesame approach usingasreferenceoneLandsat4 TMsummer scene acquired in1985by bothsensors MSS and TM aboardofthesame platform.RMSerrorinthiscasewas independently estimatedin23.01m.
2.2.Flood discrimination
AccordingtoresultsfromBustamanteetal[5]wetested the ability of several methods and indices to discriminatelandcovered bywater.Thebest one consistedonsimpleslicingofTMband5andMSS band 4enablingtodetect different inundationlevels. Final inundationmaskswereproduced forevery availabledateonthisbasisbymergingon onehandlow inundatedand drysoilcategoriesandonthe otherhand wet and fully flooded pixels.
Water-pumping wells(1974) Water transfer(1981)
RiverWall (1984)
River Dredging
2.3.Hydroperiodcalculations
Twowaystoestimatehydroperiodwereappliedover the time series. The first one estimatedthe number of daysapixelremainedinundatedalong acomplete hydrological cyclebyassumingmaintenance of inundation among two consecutive scenes. This first approach was reachedbyassigningJuliandaytoevery inundationmaskandcalculating the difference in days withthefirstavailablescene foreverycycle.Thus, final imagesgiveinformation onthenumberof daysapixel remained inundatedper cycle, i.e.annualhydroperiod.
Onthe otherhand, wegenerateddecadalhydroperiod compositesbyaddingeveryavailabledatepermonth
Prolongation of the RiverWall (1998)
Figure 2.Doñana marshesafter the toxicspill (1998).
For the legendsee Figure1.
2. MATERIALANDMETHODS
2.1.Time series of images
Weacquiredalltheavailable cloudfreeLandsatMSS, TMandETM+scenesforDoñana.Intotalmorethan
300imagescoveringthe1975-2004period composethe timeseries.Inordertomakethemcomparableweapply
a consistentpre-treatment.
A recent Landsat 7 ETM+ reference image from summerwithclearskyconditions wasgeoreferenced using 100groundcontrol points overaerialorthophotos oftheJuntadeAndalucía[3].The remainingTMand ETM images were co-registered to this reference image. Resamplingmethodwascubicconvolution. RMSerror calculatedwith anindependent test point sampleonco- registeredimages was 19.7m.Imageswerethen radiometrically corrected and transformed into reflectancevaluesapplyingthePonsSolé-Sugrañes [4]model.Finallyreflectanceimageswerenormalized tothereferenceimageusingasetofpseudo-invariant
and subsequently averagingtheresults per decade.
The lattermethodwas designedtoallow inter-decadal comparisonafterimportant transformationsinthe marshlandandthe formeraimingto detect quantitative regionalchangesinhydroperiodtrendswhensubjected to naturalvariability andanthropic transformations.
3. RESULTS
Inundation masks were produced for more than 300
Landsatimages.Annual hydroperiodwascalculatedfor every hydrologicalcycleofthe1975-2004period, startingthe 1stSeptemberandending 31stAugust. We discardedcycleswithlessthan 224days (7months) betweenfirst andlastavailableimage.Annual composites were then visually compared in order to assesforlocal changes duringthelast30years. Average hydroperiodtrendforthenaturalmarshdidnotshow anysignificant increaseordecreaseforthestudyperiod (r2=0.02, p0.05,Fig.3). However, visual interpretation didshowasoftincreaseinrecentyears of hydroperiodforhydrological cycleswithsimilarvalues of rainfall.
Decadalhydroperiodaidedto assesaboutlocalchanges alongthe 3analyzed decades.Forthispurpose, final decadal compositeswere subtracted onetoeach otherto evaluate the areas that have undergone dramatic changesin hydroperiodand reclassified to show the percentofchange(Fig.4).Theseareaswerespatially
identifiedandtheprocessesinvolvedwereaddressed 2 1
accordingtothehistoricalandcurrenttransformations.
Severalprocessesweredetectedtobethemaincauseof 3
the trendsobserved for inter-decadal comparison: 4
3
a) Hydroperiod decrease: Northern areas have been
isolatedfromtheoriginalwaterinflows(number1
inFigure4)andacceleratedsiltationprocessesare 3
altering topography and consequently reducing 3 4
hydroperiod (number3 in Figure4),
b) Hydroperiod increase: Siltation processes are 3
promotingupstreamareastoremainforlongertime
inundated, sothat they are acting as casual dams (number 2 inFigure 4); on the other hand, appearance of new inundable areas such as fish
culturesandsaltworkshaveincreasedhydroperiod
in places where inundation used to be shorter
(number4 in Figure 4).
250
y=1.1958x+91.702
R2 =0.0228
Figure 4. Areas withrelevantlocal changes ininter-
decadal comparison ofhydroperiod(1985-95versus
1995-2004).Red andyellowzonesindicatehydroperiod reduction and bluepixelsaccount foraconsistent increase(see text fordetails).
200
100
50
0
250
200
150
100
50
y=0.2597x-41.484
R2=0.7039
HydrologicalCycle
Figure 3.Averagetrendof Doñana naturalmarshes hydroperiodalongthelast 30 years.
Inordertoaddressthequestiononthemaindriving forceforhydroperiodchangeswe analysedthe relationship betweenaverageannual hydroperiodand accumulated rainfall per cycle. Results show a significant and positive relationexplaining upto 70% of variancein hydroperiodduringthe last30years(Fig.5). Theunexplained30% of variancemightaccountfor other driving forcesrelated to humantransformations.
Non-linearmodels were notattemptedtobefitted althoughaloglinearcurvewouldbe apparentlybetter fittedto data, revealinginflexionpoints betweenthe amount ofrainfallandthe correspondinginundation time.
0
020040060080010001200
Rainfall(mm)
Figure 5.Relationship between hydroperiod and rainfall.
4. DISCUSSION
Doñanawetlands are being monitoredthroughasemi- automatic approachleadingtoreveal historical patterns of inundation inthe marsh. The approachis time- consuming since pre-treatment of the time series of imageshasto beappliedovermorethan300Landsat images.
Inundationmapping iseasily achieved by applyinga simplethresholdtoTM band 5and MSSband 4 revealingthespatialpatternsof watercoverfor3 decades. Agreementwithground-truth datavalidatesthe methodology for accurately discriminate inundated
areas, either turbid or clean, plant-covered or bare water.
Historicalreconstruction has been summarizedthrough thecalculation of hydroperiod following 2 proposed methods.Assumedtrendsofhydroperiodchangesby thelocalmanagers couldbe re-assessed withthe helpof such valuable information. Annual hydroperiod show notsignificanttrendforthe wholemarshalthoughlocal changes are fairlyvisibleon ainter-annual comparison basis. Forinstance, hydrologicalcycleswithsimilar rainfallamountyieldedquite differentspatial patterns of annual hydroperiods. Furtherworkis ongoingto evaluate the effects of short-termtransformations.
Complementarily, decadal hydroperiod composites allowedtolocallycomparechanges onthetime ofwater persistedafter largewetlandtransformation.Channeling ofTraviesostream,oneofthemaintributaryinflows hasinitiatedanacceleratedsiltationprocess attheNorth ofthemarsh. Suchdynamicsareresponsibleforthe increaseofhydroperiodupstreamandthedecreasein thenewdeposedsands.Alternatively,loss of Northern inflowsconnection has beenslowly performedby wetland drainage for agricultural practices. Other humantransformationand exploitations such as fish cultures located at the South-eastern part of the marsh andrehabilitation ofsalt-worksinthelastdecade have caused anartificial increaseon localhydroperiod.
Besides, during the last decade an important phenomenon of overinundationhastakenplaceasa consequence oftheconstructionand prolongation ofthe Riverwall. Apartofsuchregionaleffect,newinflows derived from management decisions have increased local siltation processes reducing hydroperiod.
Finally,resultsfrom theanalysis ofannualrainfallshow thatrainfall accounts for70% ofthe hydroperiod variability. This fact leads tointerpret that themain driving force in the process of filling in and maintenance ofinundationis rainfall.However,thereis stillaremaining30% of variance whichcould be explainedby local andconsistent changingtrends.
Asaconcludingremark,currentdecisionsonDoñana
2005restorationprogrammearetaking intoaccount resultsfrom thishistoricalreconstructionandhelpfully aidingtovalidateexpectedresultsfrom its implementation.
5. AKNOWLEDGEMENTS
Thisstudywasfundedby DoñanaNationalPark administration(Ministry of Environment)throughthe research project "Reconstructionof bird population dynamics during the last three decades" and by the
Ministry ofScienceandEducationthroughtheresearch project HYDRA(#CGL2006-02247/BOS). F. Pacios was funded by an I3P fellowship from the CSIC. Doñana National Park and Natural Park provided permitsforfieldworkin protectedareaswithrestricted access. A. Travaini, H. Le Franc, D. Paz, A. Polvorinos, M.JiménezandI.Románhelped withfiledwork. J.C. Gilabert,J.L. Pecharrománand P.L. Porta helpedwith imageprocessing.
6. REFERENCES
1. Bourgeau-Chavez, L.L., Smith, K.B., Brunzell, S.M., Kasischke,E.S., Romanowicz, E.A.and Richardson, C.J. Wetlands Vol.25, No.1, 176-191,
2005.
2. GarcíaNovo,F.andMarínCabrera,C. Doñana: AguayBiosfera. Doñana2005, Confederación Hidrográficadel Guadalquivir,MinisteriodeMedio Ambiente.,Madrid,2006.
3. Junta de Andalucía. Ortofotografía digital de Andalucía(Color).Colorfotogrametricflight 1:60.000 (1998-1999). Juntade Andalucía, 2003
4. Pons,X.andSoléSugrañes,L.RemoteSensingof
EnvironmentVol. 48, 191-294,1994.
5. Bustamante,J.,Díaz-Delgado,R.andAragonés, D. RevistadeTeledetección,Vol. 24, 107-111,2005