Sie sind auf Seite 1von 10

DeterminingtheExtentofOilSpilledinthePersianGulfUsingLandsatImagery

andDensitySlicing
AlixHartmann
January2016
ESCI442

Abstract
InJanuary1991,largeamountsofoilwerereleasedintothePersianGulfasaresultofanactof
environmentalwarfare.Thegoalofthisanalysiswastomaptheextentoftheoilspillusinga
processdescribedbyDavidWallin.Iaccomplishedthisbyapplyingspectralanalysistoolsin
ENVI,primarilydensityslicing,toLandsatimageryoftheregion.Ifoundthatresultofthe
densityslicingtoolwassubstantiallyimprovedbymultiplyingtheimagebyaland/watermask.
Afinalresultrevealedthatoilwasspilledon61.6hectares,or2.6percent,ofthestudyarea.
Whilethisprocesscanprovidevaluableinsightintoenvironmentalvariables,itshouldnoted
thattheprocessofchoosinglandclassesislargelysubjective.

Methods
ThestudyareaislocatedinDawhatAdDaffialongtheSaudicoastline.TheLandsatimagewas
acquiredonMarch4,1991.Theimageis512by512pixels,witheachpixelrepresenting30
squaremeters.

Figure1:Thestudyareaisdisplayedhereusingband5.
Iexaminedthedistributionofreflectancevaluesinband5oftheLandsatimage(TM5).There
arethreedistinctclustersinthedata.Usingatruecolorimagetovisualizelandcoverinthe
studyarea,Ideterminedthattheserepresentedwater,oil,andland.


Figure2:ThedistributionofbrightnessvaluesinTM5.Ideterminedthatthethreebumpsinthe
dataevidenthererepresented(fromlefttoright)water,oil,andland.

Icloselyexaminedthedistributionofreflectancevaluesandseparatedthemintofourdistinct
landcovertypes:water,lightoilspillage,heavyoilspillage,andland.Usingatechnique
describedinasimilaranalysisbyPaulBaumann,heavyoilwasdeterminedtohaveahigher
reflectancevaluethanlightoil.Watercoverwastheleastreflective,andlandcoverwasthe
mostreflective.Watercovertypeshadreflectancevaluesbetween0and25,lightoilcover
typesbetween26and50,heavyoilcovertypesbetween51and90,andlandcovertypes
between91and183(thehighestrecordedreflectancevalueinTM5).

Table1:UsingthedatashowninFigure2,fourdistinctcovertypesweredeterminedusing
brightnessvaluesinTM5.

CoverType

MinimumBrightness

MaximumBrightness

Water

25

LightOil

26

50

HeavyOil

51

90

Land

91

183

UsingthesecategoriesandtheDensitySlicingtool,Icreatedapseudocolorschemetoapply
toTM5.Thisallowedmetovisualizetheresultsofmysimplecovertypeclassification.


Figure3:TheresultsofapplyingapseudocolorschemetoTM5.
Ianalyzedtheresultsofthisclassificationschemetodeterminethespatialextentofoilspillage
intheregion.Sinceeachpixeloftheimagerepresents90squaremeters,pixelcountswere
multipliedby90,thendividedby10,000tofindthespatialextentinhectares.Thepercentage
ofthestudyareawascomputedusingthenumberofpixelswithinacovertypedividedbythe
totalnumberofpixelsintheimage.
Thepreliminaryresultsofmyanalysisrevealtheextentoftheoilspillinthestudyarea.Lightoil
spillageaccountedfor151.7hectaresand6.4%ofthestudyarea.Heavyoilspillageaccounted

for430.9hectaresand18.3%ofthestudyarea.Thetotalextentofoilspillageequatesto582.6
hectaresand24.7%ofthestudyarea.
Table2:ThespatialstatisticsderivedfromresultsofapplyingapseudocolorschemetoTM5.

CoverType
Water

Pixels

Hectares

PercentofStudyArea

151,311

1,361.8

57.7

LightOil

16,853

151.7

6.4

HeavyOil

47,877

430.9

18.3

Land

46,103

414.9

17.6

Toimprovetheaccuracyoftheseresults,IfollowedtheprocessoutlinedbyWallin.A
land/watermaskwascreatedusingtheBuildMasktool.Landcovertypesaredividedinto
twocategories:landandwater.Landisgivenavalueof0andwaterisgivenavalueof1.Using
theBandMathtool,TM5andtheland/watermaskweremultipliedtogether.Thisproduced
anoutputwherealllandareahasavalueof0andwaterarearetainsitsoriginalvalue.The
advantagetovisualizingthestudyareathiswayisthatbrightnessvaluesonnonlandcover
typesarestretchedout,allowingeasierdifferentiationbetweenwaterandoil.


Figure4:TheproductofTM5andaland/watermask.Oilcover,showninthebrightwhite
values,ismoreeasilydiscernedhere.
Finally,thepseudocolorschemewasappliedtotheproductofTM5andtheland/watermask.
Thepseudocolorschemewasadjustedtotakeintoaccountnewlandandwatervaluescreated
bythemaskingprocess.Landcovervalueswerenoweither0or91through182.Watercover
valuesaresetto1through25.

Results
Thefinalresultsofmyanalysisrevealtheextentoftheoilspillinthestudyarea.Lightoil
spillageaccountedfor49.7hectaresand2.1%ofthestudyarea.Heavyoilspillageaccounted
for11.9hectaresand0.5%ofthestudyarea.Thetotalextentofoilspillageequatesto61.6
hectaresand2.6%ofthestudyarea.

Figure5:TheresultsofapplyingapseudocolorschemetotheproductofTM5andaland/water
mask.

Table3:Thespatialstatisticsderivedfromresultsofapplyingapseudocolorschemetothe
productofTM5andaland/watermask.

CoverType
Water

Pixels

Hectares

PercentofStudyArea

140,948

1,268.5

53.8

LightOil

5,518

49.7

2.1

HeavyOil

1,324

11.9

0.5

114,354

1,029.2

43.6

Land

Discussion
ThisanalysisreliedonthespectralsignaturesofasingleLandsatscene.Theresultscouldbe
improvedbyincorporatinggroundtruthdatatoimproveconfidenceinlandcovertype
determination.Itisalsoimportanttoconsidertheeffectofsubjectivityontheseresults.The
classificationoflandcovertypesreliesonaneyeballingtechnique.WhileIdeterminedthe
landcovertypestofallundercertainrangesbasedonspectralreflectance,otheranalystsmay
usedifferingrangesbasedontheirownuniquejudgement.Furtheranalysisinthisstudyarea
wouldbenefitfromamorescientificallyrobustandrepeatableapproach.

LiteratureCited
Antonova,N.,D.Wallin.2016.LabII:EnvironmentalWarfareinthePersianGulf.
http://myweb.facstaff.wwu.edu/wallin/envr442/ENVI/442_lab2_ENVI.html
Baumann,PaulR.2001.EnvironmentalWarfare:1991PersianGulfWar.
http://employees.oneonta.edu/baumanpr/geosat2/Environmental_Warfare/ENVIRONMENTAL

_WARFARE.htm

Das könnte Ihnen auch gefallen