Changing Agriculture:
How the Ataturk Dam Impacted Sanliurfa’s Agricultural Production
Ground Cover: Sanliurfa Province, Turkey
Ground Cover: Sanliurfa Province, Turkey
1990
1998
Scale 1 : 850,000
Scale 1 : 850,000
Figure 1.
Background
TheAtaturkDamisthecenterpieceoftheSoutheasternAnatolia
Project,amassivepublicworksprogramthatextendsacross9provinces
andincludestheconstructionof22
damsand19hydro-electricstations
ontheTigrisandEuphratesRivers.1
ThegovernmentoftheRepublicof
Turkeyhasspent$32billiononthis
project,whichaimstoincreasedomesticenergyproduction,boostagriculturalproductivity,andresolvethe
economicimbalanceinthesoutheasternprovincesofTurkey.
OncetheAtaturkDambeganoperatingin1992,theman-made48.7
km3reservoirprovidedasteady low
ofwatertoirrigatehundredsofthousandsofhectaresofcroplandinneighboringprovinces,likeSanliurfa,via
massiveundergroundtunnels.
Typically,irrigatedcropslikecottonaremorepro itableforsmallfarmers,manyofwhombeganreplacing
theirnon-irrigatedwheat ieldswith
irrigatedcashcrops.Asaresult,the
regionfacesanintensifyingriskof
foodinsecurity.2
ThisprojectusessatelliteimageryandTurkishAgriculturalCensusdatatomeasuretheimpactthatthe
AtaturkDamhashadonagricultural
productioninsoutheasternTurkey.
UsingtheprovinceofSanliurfaasan
example,thesemapsdemonstrate
howrapidlythelandscapehas
changedoverarelativelyshortperiod
oftime.
Initial Observations
Methods
Thesemapswereconstructed
fromhighresolutionimagerytakenby
theUSGSLandsat7satellite.Images
fromcomparabledatesinSeptember,
whencottonisnearingharvestinTurkey,wereselectedwiththeGlobalVisualizationViewer.Theyears1990and
1998representarelativelyshortpassageoftimebeforeandaftertheopeningoftheAtaturkDamandthesatellitecapturedhigh-qualityphotographs
inSeptemberoftheseyears.
Landsat7ThematicMappercapturessevendistinctwavelengthsof
visibleandinfraredlightastheyrelectofftheearth'ssurface.3After
compilingthedifferentbandsintoa
singlerasterimage,twoadjacentrastersfromeachtimeperiodwere
'mosaicked'together.
AftermaskingtheassembledrasterimagestotheborderoftheSanliurfaprovince,itwasapparentthat
severalpixelsweremissingdata.This
slightde iciencywasdisregardeddue
tothelimitedscopeoftheproject,and
theexcessiveresourcesthatwould
havebeenrequiredtodownloadand
mosaicthemissingsegmentofthe
Landsatimagery.
Next,eachyear'sderivedraster
underwentseveralexperimentalIsoClustersandMaximumLikelihood
Classi ications.Aftermuchtrialand
error,therastersweredigitallyorganizedinto25classesaccordingtosimilaritiesacrossthesevenbandwidths.
Theclassi icationsweremanually
comparedwiththefalse-colorRGB
Change in Crop Yield By Province
Cotton 1991-1998, Wheat 1991-1998
Scale 1 : 8,700,000
Figure 3.
compositesforaccuracyandthen
painstakinglyreclassi iedinto ivegeneralcategories:
Water
TheEuphratesRiver,irrigationchannels,reservoirs,etc.
AgriculturalVegetation
Greencropland,readytobeharvested
Non-VegetativeCropland
Inactivefarmlandor ieldsgrowing
non-irrigatedcropsonadifferentcycle
Non-ArableLand
Rock,mountain,pavement,buildings,
towns,etc.
Salt
Saltaccumulatesasaresultofthe
river'snatural lood/droughtcycle.
ThemapsinFigure1representa
staticpictureofgroundcoverbefore
andafterthecompletionofthedam.
Toobservechangeovertime,the1998
mapwasreclassi iedagaintoaspeci ic
numericalsystem.TheRasterCalculatormultipliedthenewmapwiththe
1990mapandcreated25newclassi ications,showninthetabletotheright.
The inalmap(Figure2)required
anotherreclassi ication,groupingtogetherthepixelsthatdidn'tchangebetween1990and1998andgrouping
therestaccordingtotheir'changed'
statusin1998.Becauseeachpixel
measures30mx30m,calculatingthe
totalareaofeach'class'simplyrequiresmultiplyingthequantityofpixelsby900andthendividingby
1,000,000togetaresultinkm2.
Thesigni icantchangeinground
coveroveraneight-yearspanisimmediatelyevidentwhencomparingthe
mapsinFigure1.Theeyeinstantlynoticesthelossofbrown,non-vegetative
croplandandthematerializationofa
distinctpatchofgreeninthesouthern
portionoftheprovince.Thealmost
rectangulargreenpatchistheHarran
PlainanditistheprimaryareairrigatedbythetwinSanliurfaTunnelsthat
usegravity- lowtodeliverwaterfrom
LakeAtaturk.4Onlyoneofthetunnels
hadopenedby1998,soitispossible
Figure 4.
Comparative Analysis & Conclusions
Thisproject’sobjectivewasnot
simplytovisualizethechanginglandscape,butalsotoquantifytheimpact.
Inthetableabove,theshadedboxes
indicatelandthatwasleftunchanged
bytheconstructionofthedam.Theremainingboxes,whichtotal6,996km2,
revealhow36%ofSanliurfa’slandscapechangedinjusteightyears.
Thetotalamountofagricultural
land(vegetativeandnon-vegetative)
declinedby2,961km2,buttheboxes
highlightedinredshow1,210km2of
agriculturalvegetationwasadded.Asa
result,vegetativeagriculturalland
makesupamuchhigherproportionof
thetotalagriculturallandin1998.
Inshort,lesstotallandisbeing
cultivated,butinSeptember—whenirrigatedcropslikecottonarereadyfor
harvest—moreofitisvegetative. Thisresultiscorroboratedbythe
resultsofthecropyieldanalysisinFigures3and4,ontheleft.Figure3indicatesthataveryclearincreaseincottonproductionoccurredinprovinces
wheretheSoutheasternAnatoliaProjectimprovedirrigationfromtheTigrisandEuphratesRivers.
Simultaneously,cottonproduction
decreasedinmostotherprovinces,signifyingageographicshiftinthecountry'scottonproduction.
TheUSDAForeignAgricultural
Itisdif iculttouseLandsatimagesto
preciselydifferentiatebetweencrops,but
theTurkishAgriculturalCensusrecordsannualcropproductionandorganizesthedata
intodifferentlevelsbeginningin1991.Figures3and4demonstratethechangeover
sevenyearsinnationalcropproductionaccordingtoprovincewhereproductionis
measuredintons.
Wheatandcottonwerestudiedtoobserveiffarmerswerereplacingstaplefoodcropswithcash-cropsinresponsetotheincreasingavailabilityofirrigation.Increasing
cash-cropproductionchangestheeconomic
andsocialclimatebyencouraginglarger
monocultureoperationsandnonsubsistencefarmingpractices.
DatafromtheAgriculturalCensuswas
reformattedandjoinedtothegeocodedvectorshape iles.Percentagechangewascalculatedbytheformula:
Classeswerethenselectedtobest
demonstratechangingtrendsandthecolor
schemeimportedfromColorBrewerwas
modi iedtoemphasizethedifferencebetweenpositiveandnegativechange.Provincesthatdidnotgrowthespeci iedcropin
bothyearswereclassi iedandcodedseparatelyas"nodata."Majorriversandsigni icantcitieswereselectedfromlargerESRI
datasetstoprovidevisualcontext.
1. Republic of Turkey Ministry of Development Southeastern Anatolia Project Regional Development Administration. www.gap.gov.tr
2. Mustafa Yilmaz, “The Effects and Impacts of the Southeastern Anatolian Development Project.” State University of New York at Binghamton, (2012).
3. For more information see “FAQ about Landsat Missions,” www.landsat.usgs.gov/band_combination_browse_images.php
4. NASA Visible Earth: Ataturk Dam (2003). www.visibleearth.nasa.gov/view.php?id=3796
5. Dogan Altinbilek, “Development and Management of the Euphrates-Tigris Basin.” International Journal of Water Resources Development. vol. 20, no. 1 (2004).
6. FAS online, “Southeastern Anatolia Becomes a Major Cotton Producing Region for Turkey,” (2001). www.fas.usda.gov/pecad2/highlights/2001/08/turkey_gap/pictures/
turkey_gap.htm
Thesaltpatches,visibleinpink
inFigure1,aretheresultofthenaturalseasonal lood/droughtcycleof
theEuphrates.5Becausethedamregulatesthe lowoftheriver,thebanks
nolongerover lowandthesaltpatcheshaveeitherbeen looded(inthe
north)orreplacedwithagricultural
landorimperviouscover(inthe
southwestcorneroftheprovince)as
indicatedinthe1998mapandthe
chartbelow.
Figure 2.
Scale 1 : 850,000
(Production1998—Production1991)
————————————————————————— x100
Production1991
Scale 1 : 8,700,000
toimaginethattheareawouldcontain
evenmoreagriculturalvegetationtoday.
Theexpandingblueareainthe
northisalsoeasilydiscernible.TheEuphratesRiver lowssouthwestalong
thenorthernborderoftheprovince
andtheAtaturkDamislocatedjustbelowthepointwherethewaterbegins
toaccumulate.HalfofLakeAtaturkis
situatedinSanliurfawhiletherestof
the817km2reservoirextendsnorthwardintotheprovinceofAdiyaman.
Servicenotesthattheregionexperienceda50%increaseinlandusedfor
cottonproductionbetween1994and
2001.Overthesametimeperiod,the
southeastregionwentfromproducing
25%ofthecountry'scottontoproducinghalfofthewholecountry'ssupply.6 Withalargeruralpopulationthat
reliesonsubsistencefarming,almost
everyTurkishprovincegrowswheat.
Again,thesoutheasternprovincesoutstrippedthegrowthoftherestofthe
country'swheatproduction,butnotto
thesameextentthatitdidwithcotton. Between1991and1998,sixprovinces(allinthesoutheast)morethan
doubledtheircottonproduction.Sirnakproduced18timestheamountof
cottonitproducedjustsevenyears
earlier!Meanwhile,Sanliurfawasthe
onlyprovincethatsawmorethana
100%increaseinwheatproduction.
Whilethisstudycannotconclude
thatfarmersreplacedtheirwheat
ieldswithcotton,thedatashowsthat
theincreaseintheproductionofcashcropsisoutpacingtheincreaseinthe
productionoffood-crops.
ThemapinFigure2accentuates
theincreasingconcentrationofagriculturallandinSanliurfaandsupports
thehypothesisthatlarger,commercialscaleagricultureisreplacingafragmentedsubsistencefarmingsystem.
Cartography, writing and analysis all executed by Eric Joseph Siegel according to requirements for ENV107.
A special thanks to Instructor Carl Zimmerman and TA Carolyn Talmadge for their guidance and support.
Maps created on ArcMap10.1 with data courtesy of USGS Global Visualization Viewer (www.glovis.usgs.gov),
Turkish Statistical Institute (www.turkstat.gov.tr), Global Administrative Areas (www.GADM.org), and ESRI.
Projection: WGS 1984 UTM Zone 37S
May 6, 2013
Limitations
Asexplainedpreviously,anarea
of~83km2ismissingfromthe inal
mosaickedrasters.Asaresult,the inal
dataforSanliurfaispartiallyincomplete.
TheIsoClusterreclassi ication,
thoughhighlyaccurateformostofthe
province,misclassi iedasubstantial
numberofpixelsalongthenortheast
borderoftheprovince.In1990and
1998thenon-vegetativecroplandarea
appearstobemountainousandrocky
inthefalse-colorRGBphotos.Because
itrepeatedtheerrorinbothmaps,this
shouldnothavegreatlyimpactedthe
datainthecomparativeanalysis.
Theprocessofreclassifyinginto
the ivegeneralcategorieswassomewhatsubjectivebecausethereclassi icationtheprocesswasdonemanually.
Theclassi icationsystemalsofailedto
distinguishbetweenman-madeimperviouscover(urbanization)and
‘natural’non-arableland(rock).The
IsoClusterwouldlikelybeunableto
distinguishthedifferenceifthemanmadestructuresweremadeoutofmaterialcutfromthe‘natural’non-arable
terrain.
Amoredetailedanalysisandlarger-scaleprojectwouldbene itfroma
closerexaminationofothercropsand
anexpansionofscaleintoneighboring
provinces.