Analysis of Eastern Mediterranean Oak Forests Over the Period 1965–2003 Using Landscape Indices on a Patch Basis

Analysis of Eastern Mediterranean Oak Forests Over the Period 1965–2003 Using Landscape Indices on a Patch Basis

Please download to get full document.

View again

of 10
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.


Publish on:

Views: 2 | Pages: 10

Extension: PDF | Download: 0

  This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institutionand sharing with colleagues.Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third partywebsites are prohibited.In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further informationregarding Elsevier’s archiving and manuscript policies areencouraged to visit:  Author's personal copy Landscape and Urban Planning 87 (2008) 67–75 Contents lists available at ScienceDirect Landscape and Urban Planning  journal homepage: Landscape spatial dynamics over 38 years under natural andanthropogenic pressures in Mount Lebanon Ihab Jomaa a , ∗ , Yves Auda b , Bernadette Abi Saleh c , 1 , Mou¨ın Hamz´e d , 2 , Samir Safi e , 3 a Remote Sensing Center (CNRS-CRS), National Council for Scientific Research, BP 11-8281, Beirut, Lebanon b Centre de T´el´ed´etection, CESBIO – LADYBIO, BPI 2801, 31401 Toulouse Cedex 9, France c Lebanese University, Museum Area, Beirut, Lebanon d National Council for Scientific Research, BP 11-8281, Beirut, Lebanon e Faculty of Sciences II, Lebanese University, Lebanon a r t i c l e i n f o  Article history: Received 10 May 2007Received in revised form 8 November 2007Accepted 9 April 2008Available online 21 May 2008 Keywords: FragmentationLandcoverLandsat TMSpot 5MediterraneanUrban encroachmentForest regenerationRemote sensing a b s t r a c t This research investigates spatial configuration changes of five forest types and introduces uncommonexploration of urban changes in Mount Lebanon. A forest map of 1965 and satellite images of Landsat TM(1987) and Spot 5 (2003) were used. Forest area has reduced by 7%, having low annual deforestation rate(<1%). Oak, cedar and cypress forests increased from 13% to 15%, 0.8% to 1% and 0% to 0.1% respectively;while pine and juniper forests decreased from 1% to 0.8% and 12% to 3% respectively. Grassland covered20%ofthesrcinalforestareafollowedbyagricultureandurban.Forestpatcheswereinvestigatedineach100mdistancefromtheedgeoftheoldforests.Regeneratedoakswerefoundwithinthefirst100m.Pineand cedar forests appeared at the 400m distance. Urban settlements showed 4myr − 1 moving progresstowardforestpatches.Landscapeindicesdemonstratedincreaseinthenumberofforestpatches(from169to432)andinthemeanproximityindex(from78.29to225.15)anddecreaseinmeanpatcharea(from219to 70ha) and largest patch index (from 15.45 to 10.32). The observed trends revealed major deforestationand fragmentation in 1965–1987, followed by a lesser deforestation tendency in 1987–2003. Each foresttype behaves differently to deforestation and therefore requires special management practices.© 2008 Elsevier B.V. All rights reserved. 1. Introduction Although forest ecosystems are among the vital environmentalresources on earth, they are facing severe loss and fragmentation.These are major problems in developing countries and importantenvironmental issues worldwide (Laurance, 1999; Teller´ıa et al.,2003; Wade et al., 2003; Matsushita et al., 2006; Boentje andBlinnikov,2007).Anychangesinlandcovercanhaveprofoundlocal,regional, and global consequences (Klemas, 2001).The Mediterranean region is one of the world’s 18 biologi-cal hot spots (Myers et al., 2000), where important biodiversityexists, and much of that biodiversity is under danger of extinc-tion.Mediterraneanforestshaveenduredlongperiodsofintensive ∗ Corresponding author. Tel.: +961 4 409845/46; fax: +961 4 409847. E-mail addresses: (I. Jomaa), (Y. Auda), (B. Abi Saleh), (M. Hamz ´e), Safi). 1 Tel.: +961 3 243192; fax: +961 1 612815. 2 Tel.: +961 1 850125; fax: +961 1 822639. 3 Tel.: +961 3 394 962. human-induced perturbations which have exacerbated fragmen-tation processes and caused the appearance of forest patchesimmersed in matrices of humanized landscapes (G´omez-Aparicioet al., 2005). Human activities such as fire, grazing, cutting, andcoppicing have shaped the Mediterranean vegetation cover andturned it into a complex assemblage of highly diverse vegetationstructures(Qu´ezelandM´edail,2003).Inaddition,urbanencroach- ment is occurring in chaotic patterns in Mediterranean landscapessurrounding forests. The Mediterranean region is a mosaic-likelandscape characterized by intense fragmentation (Dufour-Dror,2002). This fragmentation has contributed to the impoverishmentof Mediterranean forests and grasslands (Margaris et al., 1996).Although Mediterranean forests are considered resistant to thenegative effect of fragmentation (Lazanta et al., 2006), neverthe-less fragmentation is rapidly changing the landscape. Forests of the Mediterranean region have become spatially scattered and areoften distributed as small mountain populations.Throughouthistory,therehavebeencyclesofforestdegradationand regeneration as human pressure intensified and then receded.Vegetationregrowthfollowingareductioninhumanpressuremayformacontinuousshrubcanopy,coveringareasthathadpreviouslybeenbare(SeligmanandHenkin,2000).Mediterraneanvegetation 0169-2046/$ – see front matter © 2008 Elsevier B.V. All rights reserved.doi:10.1016/j.landurbplan.2008.04.007  Author's personal copy 68  I. Jomaa et al. / Landscape and Urban Planning 87 (2008) 67–75 recovers quite rapidly thanks to its adaptability to human inter-vention (Savorai et al., 2003). However, each forest type respondsdifferently to external pressures. One forest type will adapt to theenvironment, while other types could become spatially rare.The present study examines five forest types to determine theirspatial pattern of fragmentation over a 38-year period in Lebanon.The aim of this study is: (1) to estimate the deforestation rate dur-ing the 1965–2003 period, using forest map (1965), Landsat TMimage (1987) and Spot 5 image (2003), (2) to assess the impact of changes of forest spatial configuration on class and landscape lev-els, and (3) to identify urban spatial changes with regard to theoriginal(1965)forestpatches.Thisresearchprovidedvaluableper-spectiveandreferenceonthebehaviorofthefiveforesttypeswhichcould assist in future management strategies and conservationprograms. 2. Materials and methods  2.1. Study area SituatedontheeasternshoreoftheMediterraneanSea,Lebanonoccupies the junction between Europe, Asia, and Africa and hasa surface area of 1,045,200ha. The present study was conductedon 136,318ha in the mid-northern part of the country between35 ◦ 38  –36 ◦ 21  Eand34 ◦ 9  –34 ◦ 22  N,includingthecountry’shighestpeak, Mount Lebanon (Fig. 1). The study area extends in a hori-zontal section 25km wide at the seashore and 61km wide inland.Elevationrangesfromsealevelto3088matapoint41kmfromthecoast. The topography is rough and characterized by steep slopeswith an average inclination of 85m/km. Soils are mostly of brownrendzina and terra rossa characteristics developed upon Eocenechalkylimestone(Dubertret,1953;G`eze,1956).Theclimateistyp- icallyMediterranean,withmeanannualtemperaturerangingfrom12 ◦ C inland to 19 ◦ C near the coast. Annual rainfall ranges from900mm in the lowlands to 1400mm in the highlands (Plassard,1971).RainfalloccursmostlybetweenOctoberandMay,withahotand subhumid to dry summer.  2.2. GIS database 2.2.1. 1965 Data Forest mapping in Lebanon started with the aerial mission of 1962. Consequently, the 1965 forest map is the first of its typefor the country (El Husseini and Baltaxe, 1965). Since then, it hasnever been updated for more than 38 years. It has a spatial scale of 1:50,000andconsistsof27sheetscoveringthewholecountry.Thestudy area covers six sheets that were digitized using a geographicinformation system (GIS). Five forest types were recorded in thestudy area: (1) oak forest (OF), (2) pine forest (PF), (3) cedar forest(CF), (4) juniper forest (JF), and (5) cypress forest (CyF).Since a landcover map of 1965 does not exist for the study area,urban patches were extracted from a topographic map (1:50,000spatialscale)forthesameyear.Bothlayers(forestandurbandevel-opment in 1965) were resampled to 50m pixel size and convertedto a raster grid format for landscape analysis.  2.2.2. Landcover classifications of 1987 and 2003 To achieve further understanding, image series from LandsatTM for 1987 and Spot 5 for 2003 covering the study area wereacquired. The 1987 series provided an intermediate landcoverdataset between 1965 and 2003, dating from about the end of thecivil war.The Landsat TM and Spot 5 series were resampled to have50m resolution similar to the 1965 raster map grids (Staus etal., 2002). The images were geometrically and topographicallycorrected. Atmospheric corrections were considered unnecessarybecause the final classification image is to be used for compari-son purposes rather than for change detection analysis (Bernstein,1983). Terrain elevations of 50m pixel size were obtained throughbuilding a digital elevation model (DEM) of the study area, using50m equidistant contour lines. The DEM was used as input to aLambertianReflectanceModeltoapplytopographicnormalizationsto the images (Colby, 1991).Sixclassesoflandcoverweredefined:(1)oakforest(OF),(2)pineforest (PF), (3) cedar forest (CF), (4) juniper forest (JF), (5) cypress Fig. 1.  Location of the study area in the north of Lebanon, traversing the mountains.  Author's personal copy I. Jomaa et al. / Landscape and Urban Planning 87 (2008) 67–75  69 forest (CyF), and (6) non-forest areas including agricultural fieldsand orchards, bare soil, grass areas, and quarries.Landcover classification was performed using landcover datafrom various sources such as the previously obtained 1965 land-cover information for the study area. Superimposing previous dataonto the satellite images assisted in performing visual interpre-tation and defining training sites. Fifty-three training sites werevisually defined. In addition, use was made of the empirical obser-vationthatsomeforesttypesareassociatedwithcertainaltitudinalranges (Abi Saleh et al., 1996; Hoersch et al., 2002). Ground land-cover data were gathered in 2003, when information could becollectedfrom197locations,ofwhich94pointswereconsideredtorepresent no change in landcover data (urban settlements, old for-est growth, conservation areas, etc.). These 94 ground data pointshelpedtodeterminefurthertrainingsitesforcompletingtheclassi-ficationofthe1987LandsatTMimages.Thelandcoverclassificationmap for 1987 was obtained by assigning training sites or pixels tothe category determined using the criteria discussed above. Thisprocedure is useful especially when working with earlier (belongto the past) satellite images (Cayuela et al., 2006). The entire set of ground control points (197 points) was used to classify the 2003Spot 5 images. Visual image manipulation was performed usingERDAS IMAGINE ® remote sensing software.  2.3. Accuracy assessment  Fieldandgroundverificationpoints(controlpoints)arethebasicelementsofaccuracyassessment.Observationfieldpointsarecom-paredaccordingtotheirexactlocationwiththeclassifiedlandcoverclassesonthemap.Thisaccuracyassessmentgeneratedaconfusionmatrix.Field surveys in 2003 helped to identify 86 control points withnochangeinlandcoverandtocollect174groundcontrolpoints.The86 points were used to construct the confusion matrix of the 1987classified image, while the other 174 points assisted in construct-ing the related 2003 confusion matrix. User’s accuracy, producer’saccuracy and overall accuracy were calculated based on the confu-sion matrices (Cayuela et al., 2006).  2.4. Fragmentation and landcover change analysis 2.4.1. Landscape-level metrics GIS layers of landcover were first converted into raster grids forthe application of landscape spatial indices which were calculatedusing the Spatial Analyst 2.0 of ArcView version 3.2. Small patches(lessthan5pixelsinsize)withinthehigherspatialresolutionlayers(1987 and 2003) were removed (Millington et al., 2003).Previous studies have used different landscape indices for frag-mentation analysis (Imbernon and Branthomme, 2001; Li et al.,2001;Stausetal.,2002;Echeverr´ıaetal.,2006;Cayuelaetal.,2006;Matsushita et al., 2006), which we chose the following: (a) patchnumber (NP), (b) mean patch area (MPA in ha), (c) total landscapearea (TA in ha), (d) class area (CA in ha), (e) mean shape index, (f)largest patch index (LPI, % of the landscape covered by the largestpatch), (g) area-weighted mean patch fractal dimension index,(h) mean proximity index (MPI), (i) mean nearest neighbor index(MNN);(j)Shannon’sdiversityindex(SDI),(k)Shannon’sevennessindex (SEI), and (l) interspersion and juxtaposition index (IJI).Number of patches (NP) and mean patch area (MPA) are com-puted together because they provide a direct interpretation of landscape-level fragmentation, e.g., high NP with low MPA corre-spondstoahighlyfragmentedlandscape(LeitaoandAhern,2002).For a more detailed definition of landscape indices, the reader isreferredtotheFRAGSTATSuserguide(McGarigalandMarks,1995).  2.4.2. Urban changes and urban spatial configuration with regardto 1965 forest patches The total urban area in 1965 was first compared to the urbancover of 1987 and 2003. Then we investigated urban spatialtransformation and coverage changes with respect to the spatialconfigurations of 1965 forest patches. Two separate GIS raster lay-ers were created, each including: (1) forest and urban patches in1965 and (2) forest patches in 1965 and urban patches in 2003.Mean nearest-neighbor (MNN) and mean proximity (MPI) indiceswerecomputedforeachlayeronthelandscapelevel.Thisapproachhelped to understand the degree of urban spread throughout thestudy area and how urban patches move toward areas of potentialforest use. Areas of forest potential were considered to be the oldforest spatial locations (1965 forest locations). This analysis mea-suresthedifferenceinproximitybetweenurbanandforestsin1965and between the forests of 1965 and urban patches in 2003. Therecognitionofhowurbanareasapproachedsrcinalforestlocations(1965) generated additional understanding of the forest changesthat had occurred by 2003.  2.5. Forest regeneration Regenerated forest patches were considered to be the new(2003)forestpatchesthathadappearedontheoutsideofthe1965forest patches (old forest locations within the landscape). For eachforest type, the outside forest patches were counted within each100m interval, starting from the patch edges of the 1965 forests.The area in hectares was also recorded to identify the forest typewiththemostregeneratedarea.Thisanalysisindicatedthedistancerelationship or the proximity of regenerated forests of each foresttype to the 1965 forest patches. 3. Results  3.1. Accuracy assessment  The classifications of the 1987 Landsat TM and Spot 5 imageshave an overall accuracy of 76% and 85% respectively (Table 1). Thelowerlevelofaccuracyforthe1987imageclassificationisrelatedtothelargersrcinalpixelsizeoftheimages(Mehneretal.,2004)andto the fact that the images were acquired 16 years before the fieldinvestigationsin2003.Thelowestvalueofproducer’saccuracycor-responded to cedar forest, which was under-classified. The lowestvalueofuser’saccuracywasforthepineandcypressforestclasses,which were overestimated in the classification. Cedar forest has alimitedgeographicaldistribution.Pineforestmightcontainoaksaslowerlayerorunderstorytrees,resultinginalessclearspectralsig-nature differentiation. Cypress has limited geographic distributionand underwent reforestation during the early 1990s.  3.2. Changes in landcover areas 3.2.1. Total area change of each forest type Changes in forests and landcover information were derived fol-lowingtheanalysisofthethreerastergridsof1965,1987,and2003(Fig. 2). The forest area decreased from 37,380ha (27% of the studyarea)in1965to27,321ha(20%)in2003.Thiscorrespondstoalossof only7%ofthenativeforestsduringtheperiod1965–2003.Althoughtotal forest area decreased, oak forest increased from 18,335 to20,421ha (13–15%), cedar forest from 1059 to 1069ha (0.7–0.8%)and cypress forest from 25 to 130ha (0–0.1%). Among all the foresttypes,juniperforestshowedthemostdramaticdecrease,from12%of the study area in 1965 to 3% in 2003. Pine forest decreased by0.5% during the period 1965–2003. The non-forest cover increasedfrom 73% to 80% from 1965 to 2003 (Table 2).  Author's personal copy 70  I. Jomaa et al. / Landscape and Urban Planning 87 (2008) 67–75  Table 1 Confusion matrix for classification of 1987 Landsat TM and 2003 Spot 5 satellite images using six categories of landcoverClassified map Ground control points1987 TM image 2003 Spot 5 imageOF PF CF JF CyF NoF User’s accuracy OF PF CF JF CyF NoF User’s accuracyOak forest (OF) 20 1 0 0 0 1 90 39 2 0 0 0 1 92Pine forest (PF) 3 9 0 4 1 0 52 3 20 1 4 1 0 69Cedar forest (CF) 0 1 3 0 0 0 75 0 1 3 0 0 0 75 Juniper forest (JF) 0 1 0 20 0 1 90 0 2 0 26 0 3 83Cypress forest (CyF) 0 0 1 0 3 0 75 0 1 1 1 5 0 62Non forest cover (NoF) 1 0 1 3 0 11 68 2 0 0 3 0 55 91Total 24 12 5 28 4 13 86 44 26 5 34 6 59 174Producer’s accuracy 83 75 60 71 75 84 76 88 76 60 76 83 93 85  Table 2 Estimated area of landcover types (class area, CA) in 1965, 1987 and 2003 in the study areaLand classes 1965 1987 2003ha % ha % ha %Oak forest 18,335 13 18,788 14 20,421 15Pine forest 1,592 1 1,067 1 1,023 1Cedar forest 1,059 1 1,048 1 1,069 1 Juniper forest 16,369 12 6,589 4 4,678 3Cypress forest 25 0 122 0 130 0Non forest cover 98,938 73 108,704 80 108,997 80Total forests only 37,380 27 27,614 20 27,321 20 Fig. 2.  Variation in the spatial configuration of forests in 1965, 1987, and 2003.
Related Search
Similar documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks