Clearance and fragmentation of tropical rain forest in Xishuangbanna

BiodiversConserv(2009)18:3421–3440

DOI10.1007/s10531-009-9651-1

ORIGINALPAPER

Clearanceandfragmentationoftropicalrainforest

inXishuangbanna,SW,China

HongmeiLiÆYouxinMaÆWenjieLiuÆWenjunLiu

Received:6May2008/Accepted:13May2009/Publishedonline:24May2009

ÓSpringerScience+BusinessMediaB.V.2009

AbstractXishuangbanna,situatedinthenorthernmarginofthetropicalzoneinSoutheastAsia,maintainslargeareasoftropicalrainforestandcontainsrichbiodiversity.However,tropicalrainforestsarebeingrapidlydestroyedinthisregion.Thispaperana-lyzedspatialandtemporalchangesofforestcoverandthepatternsofforestsfragmentationinXishuangbannabycomparingclassi?edsatelliteimagesfrom1976,1988and2003usingGISanalyses.Thepatternsoffragmentationandtheeffectsofedgewidthwereexaminedusingselectedlandscapeindices.Theresultsshowthatforestcoverdeclinedfrom69%in1976tolessthan50%in2003,thenumberofforestsfragmentsincreasedfrom6,096to8,324,andthemeanpatchsizedeclinedfrom217to115ha.Itwasfoundthatfragmentsizedistributionwasstronglyskewedtowardssmallvalues,andfragmentsizeandinternalhabitatdifferstronglyamongforesttypes:lessfragmentedinsubtropicalevergreenbroadleafforest,butsevereinforeststhataresuitableforagriculture(suchastropicalseasonalrainforestandmountainrainforest).Duetofragmentation,theedgewidthwassmallerin2003thanthatin1976whenthetotalareaofedgehabitatexceededcorehabitatindifferentforesttypes.Thecoreareaoftropicalseasonalrainforestwassmallestamongmainforesttypesatanyedgewidth.Fragmentationwasseverewithin12.5-kmbuffersaroundroads.ThecurrentforestcoverwithinreservesinXishuangbannawascomparativelylargeandlessfragmented.However,thetropicalrainforesthasbeendegradedinsidereserves.Forconservationpurposes,theapproachestoestablishforestfragmentsnetworksbycorridorsandsteppingstonefragmentsareproposed.Thecon-servationeffortsshouldbedirected?rsttowardtheconservationofremainingtropicalrainforests.

KeywordsForestfragmentationÁEdgeeffectÁCorehabitatareaÁBiodiversityÁNaturereservesÁRoadbufferÁXishuangbanna

H.Li(&)ÁY.MaÁW.LiuÁW.Liu

XishuangbannaTropicalBotanicalGarden,ChineseAcademyofSciences,88XuefuRoad,

650223Kunming,People’sRepublicofChina

e-mail:lihm@http://wendang.chazidian.com;lihm080@http://wendang.chazidian.com

123

3422BiodiversConserv(2009)18:3421–3440Introduction

Disturbedbyhumanactivities,forestlossandhabitatfragmentationhavereceivedworldwideattention(Pimm1998;Lauranceetal.1997,1998;Parthasarthy1999).Habitatlossandfragmentationareamongtheprincipalcausesofbiodiversitylossandthecollapseofprimaryproductivityinthetropicalrainforests(Rantaetal.1998;Lauranceetal.1997,1998;DebinskiandHolt2000;DeFriesetal.2005).Forestfragmentationincludesbothareductionininteriorhabitatandanincreaseinedgelength,edgehabitatareaandthedegreeofisolationofforestpatches(Lauranceetal.2002;Cayuelaetal.2006).Forestfragmentationcanseverelymodifyhabitat’sphysicalorbiotaconditionsthatmanyspeciesliveonaffectthedistributionpatternofspecies,andeveninducesomespeciesloss(Maetal.1998;Lineraetal.1998;Lauranceetal.1998;Coxetal.2003).

Understandingthepatternsandprocessesofhabitatchangesisessentialforstudyingtherelationshipbetweenforesthabitatfragmentation,humanimpact,reservenetworks,andbiodiversityconservation(Revillaetal.2001;Ferrazetal.2005).Foresthabitatsizes,extentofedge,andthepastdisturbanceofremnantcanstronglyin?uencespeciesresponsestofragmentation(DebinskiandHolt2000;Euskirchenetal.2001;Harperetal.2005).Researchontwocomparableareasofevergreenmontaneforestreportedthattheareaswithrelativelylargepatches,someconnectivityandfewhumanactivitiescouldsupportpopulationsoflargemammalsandfrugivorousbirdsextirpatedinotherareaswithsmall,isolatedpatchesandmorehumaninterference(PattanaviboolandDearden2002).Researchersalsofoundthatspecies’abundancesinfragmentsdifferedfromthoseinintactforest,withsomedecliningandothersbecominghyperabundant(Lauranceetal.2002).Small(1–10ha)andisolatedfragmentscanlosespeciesinitiallyataremarkablyhighrate(Lauranceetal.2002).Increasedamountofedgehabitatandedgein?uencearethemostimportantconsequenceoffragmentation.Theforestedgein?uencescanleadtothedeg-radationofforestsfragments(Lauranceetal.2002).Theedgeeffectsofforestfragmentsarethemostimportantproximatecauseofelevatedtreemortality,damage,andturnoverbyalterationsinforestmicroclimateandgreaterwindturbulencenearedges(Lauranceetal.1998).Evaluatingedgeeffectswithinremnantforestsanddelineatingareaofedgein?u-encesoflandscapeareparticularlyimportantforresourceassessments,biodiversitystudies,landscapedesign,andwildlifehabitatmanagement(ZhengandChen2000).Themoreseverethepressureofhumanencroachment,themorefragmentedtheexistinghabitatbecomes.Roadconstructionasconcentratedhumanactivitypromoteslandscapemodi?-cation(McGarigaletal.2001).Theecologicaleffectsofroadscanresonatetosubstantialdistancesfromtheroad,creatinghabitatfragmentationandfacilitatingfragmentationthroughsupportofhumanexploitativeactivities(TrombulakandFrissell2000).Theincreaseinroaddensityaccountedformostofchangesinlandscapecon?gurationasso-ciatedwithmeanpatchsize,edgedensity,andcoreareametrics(McGarigaletal.2001).Buildingprotectionareaalleviatedforestfragmentsandpromotedbiodiversityconserva-tion(http://wendang.chazidian.comrgerforestpatchescanprovideeffectiveprotectivehabitatforspeciessurvivalandislesssusceptibletoedgeeffects(DebinskiandHolt2000).Forestreservedesignsfrequentlytakeintoaccountfragmentationpatternstopreservelargerandlessisolatedforestfragments(Rantaetal.1998;PattanaviboolandDearden2002).Sincesmallreservesmightrepresenthighqualityremnantsintropicalrainforest,theyarealsoemphasizedasprotectiontargetsforfutureexpansionofreservenetworks(Piessensetal.2005).Morestudieshavereportedthattheecologicalconsequencesoffragmentationmaydifferdependingonthepatternsorspatialcon?gurationimposedonalandscapeandhowitvariesbothtemporallyandspatially(Cayuelaetal.2006;Guiradoetal.2006).123

BiodiversConserv(2009)18:3421–34403423Understandingthepatternsofforestlandscapechangeandtheprocessesisessentialformanagingandconservingforestfragmentsanddiversityconservation(Rantaetal.1998;Lauranceetal.2002;PattanaviboolandDearden2002;Guiradoetal.2006).

InthisstudyweanalyzespatiallyexplicitinformationonforestcoverchangeandforestfragmentationstatethroughoutXishuangbanna,southwest,China,overtheperiod1976–2003.XishuangbannawasoneoftherichestbiodiversityspotinChina(CaoandZhang1997).Itrepresentsonly0.2%oftheareaofChina,butitcontainsapproximately5,000speciesofhigherplants(16%ofthenation’stotal),102speciesofmammals(21.7%),427speciesofbirds(36.2%),98speciesofamphibiansandreptiles(14.6%),and100speciesoffreshwater?sh(2.6%)(ZhangandCao1995).Withhumanpopulationgrowth,traditionalslash-and-burnagriculturalactivitiesandrubberplantationexpansion,deforestationwasdramaticinthepastdecades.Humanpopulation,increasedfrom220,000in1953to990,000in2000inXishuangbanna,suggestedthattheincreaseofpopulationforexploitingthenatureresourcemaybeoneofimportantfactorsaffectinglandscapechange.Ontheotherhand,http://wendang.chazidian.comrgeareaoftropicalrainforestandshiftingcultivationlandsatloweraltitudeshasbeenconvertedtorubberplantationsoverlast50years,therebyinducingtheclearanceofforestdistributionathighaltitudesorsteepslopesfornewarablelanddemand(Lietal.2007).

NaturereserveswerebuiltinXishuangbannaasthelossofforestandbiodiversityreceivedlocalgovernment’sattention.Atthesametime,manyforestrylawsandpoliciesrelatedtoforestconservationwereformulated,suchasNaturalForestConservationPro-gramin1998(Longetal.1999),andSlopingLandConversionProgramin1999(Zhangetal.2000).However,thesepolicieshadmixedimpactwithinXishuangbannaandnaturalforestcoverwasstilldeclining.Thetotalforestscannotre?ectthestatusofhabitatsregardingbiodiversitysincerubberplantationisincludedinforestsbyde?nition.Althoughabout12%ofthetotalareaofXishuangbannaarebuiltfornaturereserves(Guoetal.2002),eachsub-reservehadbecomeanisolatedislandbecausemostareasoutsidetheprotectedareabecamefarmlandsandplantations(Fig.1).Forestlossandfragmentsin?uencethespeciesdynamics.Forexample,theincidentsofhuman-elephantcon?ictsinXishuangbannaoftenappearedinpublicmedia(http://wendang.chazidian.com/a/20090112/000106.htm).ThemainthreatstothesurvivalofAsianelephantsarehabitatalterationandavailabilityoffoodcausedbyincreasedhumaninterference(ZhangandWang2003;FengandZhang2005).ItisimportanttoidentifyandunderstandAsianelephants’habitatstructure,continuityofhabitat,availabilityoffood,andthemovementpatternsofherdstoensurethecontinuedexistenceofAsianelephantsinChina(ZhangandWang2003).Aresearchonspeciesdiversitychangewithtropicalrainforestfragmentationalsoreportedthatthemostdominantspecies(i.e.,Barringtoniamacrostachya)inprimarynaturerainforestinXishuangbannadisappearedwithforestfragmentationandtheplantspeciesdiversityisgenerallylowerinthefragmentedforeststhanintheprimaryforest(Zhuetal.2004).Withnaturalprimaryforestscoverdecrease,localecologicalenvironmentwasalsodegraded,suchasfogformationanddurationreduction(Huangetal.2000;Liuetal.2004,2007).

AlthoughsomeoftheecologicalconsequencesofforestfragmentationhavebeeninvestigatedinXishaungbannan,nosystemicstudyhasbeenundertakentounderstandthetemporalandspatialchangesofforestfragmentationandhumanimpactsonfragmentationinXishuangbannatoprovideconservationistsandenvironmentalmanagerswithinfor-mationonthelastremnantsofforestfragmentsandthreatstobiodiversity.Theobjectivesofthepaperistoanalyze(1)temporalandspatialpatternofforestschangeduring1976and

123

3424BiodiversConserv(2009)18:3421–3440Fig.1ThelocationofXishuangbannainthesouthernpartofYunnanprovinceofChina

2003basedonthreeLandsatimagesandgeographyinformationsystemtechnology;(2)thepatternsofforestfragmentationandtheeffectsofvaryingedgewidthinforestsbyusingselectedlandscapeindices;and(3)thedegreeofforestfragmentationunderhumaninterference.

Methods

Studyarea

Xishuangbanna(21°080–22°360N,99°560–101°500E)locatedinYunnanProvince,southwestChina,covers19,150km2andbordersLaostothesouthandMyanmartothesouthwest(Fig.1).Theregionhasmountain-valleytopographywiththeHengduanMountainsrunningnorth-south,andabout95%oftheregioniscoveredbymountainsandhill.Thealtitudevariesfrom2,430to475mabovesealevel.Theclimateofthisregionisin?uencedbywarm-wetairmassesfromtheIndianOceaninsummer,includingmon-soons,andcontinentalairmassesofsubtropicalorigininwinter,resultinginarainyseasonfromMaytoOctober,andadryseasonfromNovembertoApril.ThecombinationofgeographyandclimateinXishuangbannahascreatedatransitionzonebetweenthe?oraandfaunaoftropicalSouthEastAsiaandsubtropicalandtemperateChina,resultingintheregionwiththehighestbiodiversityinChina(ZhangandCao1995;CaoandZhang1997).The?veprimaryforesttypesinXishuangbannaare:tropicalseasonalrainforest,tropicalmountainrainforest,evergreenbroad-leavedforest,monsoonforestoverlimestone,andmonsoonforestonriverbanks(Wuetal.1987

).

123

BiodiversConserv(2009)18:3421–34403425Datasourcesandmethods

Land-use/land-coverchangewasdeterminedusingtwoLandsatMultiSpectralScanner(MSS)images(24February1976—#139/45,and25April1975—#140/45),aLandsatThematicMapper(TM)image(2February1988—#130/45)andaLandsatEnhancedThematicMapper(ETM)image(7March2003—#130/45).Twoimageswereusedtocreatethe1976cover,withinformationfrom1975usedto?llinareaswithcloudcoverinthe1976image.AllimageswereacquiredduringthedryseasonbetweenFebruaryandApril.Twoland-usemapsdevelopedbytheXishuangbannaDepartmentofLandandResource(XishuangbannaLand-useStatusMap1982,1991)andavegetationmapdevelopedbytheXishuangbannaForestryBureau(XishuangbannaVegetationDistributionMap1993)wereusedasreferencesfortheclassi?cationandaccuracyestimationoftheMSSandTMimages,respectively.Topographicmaps(scale=1:50,000)anddigitaltopographicdatawithacontourintervalof100mpublishedbytheStateBureauofSurveyingandMappingofChinawereusedtobuildadigitalelevationmodel(DEM).TheTMsatelliteimageswererecti?edtoAlbersConicalEqualAreaprojectionsystemwitha35-mpixelsize.TheETMandMSSimageswereregisteredtotheTMimagesusinganimage-to-imageregistrationtechnique:recti?cationRMSerrorswere\0.5pixelsand\1pixels,respectively.Allnon-thermalchannelsoftheTMandETMimagesandallchannelsoftheMSSimageswereusedtocreateclassspectralsignaturesforclassi?cation.Theimageswereclassi?edusingthesupervisedmaximumlikelihoodclassi?cationmethod.Trainingareasforeachland-coverclasswereidenti?edforeachimage.FortheETMimage,trainingareaswereidenti?edinthe?eldduringFebruary–March2003.FortheTMandMSSimages,trainingareasweregeneratedfromtheDepartmentofLandandResourcemapsof1982and1991,andtheForestryBureau’svegetationmapof1993,respectively.Weselectedlargehomogeneousareasforthetrainingareas.Foreachland-usetype,weincludedatleast10trainingareastore?ectthevariationwithinalanduseduetotopographyandslopeeffects.Initiallyweusedthesame15land-useclassesdevelopedbytheNationalAgriculturalZoningCommittee(1984).Forestswereclassi?edintofourclasses.Itwasdif?culttodistinguishdifferentforesttypesfromtheimages.ThecommonforesttypesXishuangbanna(i.e.,tropicalseasonalrainforest,mountainrainforestandsubtropicalevergreenbroadleafforest)wereseparatedbasedonelevation(Guoetal.1987).Tropicalseasonalrainforestisforestedareaswithgreaterthan30%closedcanopydominatedbybroadleaftrees,andatanaltitudelessthan800m.Mountainrainforestisforestedareaswithgreaterthan30%closedcanopydominatedbybroadleaftrees,andatanaltitudebetween800and1,000m.Subtropicalevergreenbroadleafforestisforestedareaswithgreaterthan30%closedcanopydominatedbybroadleaftrees,andatanaltitudegreaterthan1,000m.Conifersandbambooscouldbedistinguishedbasedondifferencesintextureandspectralcharacteristics.Rubberplantationswereeasytoclassifybecausethetreesaredeciduousduringthedryseason,andmostnativeforestspeciesareevergreen.Shrublandisacommonland-useclass,butitisoftenatransitionbetweenabandonedagriculturallandandforestorplantations.Arablelandsincludedareasofactiveagricul-ture,shiftingcultivation,grassland,teagardens,andpaddyrice.Thelandusepolygonthemesfor1976,1988and2003,obtainedfromthedigitalclassi?cationofsatellitedateandsubsequentGISanalyseswereoverlaidandintersectedtoderivelanduse/coverchanges.

Theaccuracyofourclassi?cationwasveri?edbyground-truthing.Speci?cally,wecomparedourclassi?cationofthe2003ETMimagewith?eldobservationsinDecember2004.Atotalof286pointswereveri?ed.Ineachpoint,wedeterminedthecurrentland-use

123