Mapping and monitoring deforestation and forest degradation in Sumatra (Indonesia) using Landsat time series data sets from 1990 to 2010

Mapping and monitoring deforestation and forest degradation in Sumatra (Indonesia) using Landsat time series data sets from 1990 to 2010
of 17
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.
Related Documents
  This content has been downloaded from IOPscience. Please scroll down to see the full text.Download details:IP Address: content was downloaded on 10/10/2013 at 15:41Please note that terms and conditions apply. Mapping and monitoring deforestation and forest degradation in Sumatra (Indonesia) usingLandsat time series data sets from 1990 to 2010 View the table of contents for this issue, or go to the  journal homepage for more 2012 Environ. Res. Lett. 7 034010( usMy IOPscience  IOP P UBLISHING  E NVIRONMENTAL  R ESEARCH  L ETTERS Environ. Res. Lett.  7  (2012) 034010 (16pp) doi:10.1088/1748-9326/7/3/034010 Mapping and monitoring deforestationand forest degradation in Sumatra(Indonesia) using Landsat time series datasets from 1990 to 2010 Belinda Arunarwati Margono 1,2 , Svetlana Turubanova 3 ,Ilona Zhuravleva 4 , Peter Potapov 3 , Alexandra Tyukavina 3 ,Alessandro Baccini 5 , Scott Goetz 5 and Matthew C Hansen 3 1 Geographic Information Science Center of Excellence (GIScCE), South Dakota State University,Brookings, SD 57007, USA 2 The Ministry of Forestry (MoF) of Indonesia, Jakarta 10270, Indonesia 3 Department of Geography, University of Maryland, College Park, MD 20742, USA 4 Greenpeace Russia, 125040 Moscow, Russia 5 Woods Hole Research Center, Falmouth, MA 02540, USAE-mail: and Received 6 April 2012Accepted for publication 2 July 2012Published 19 July 2012Online at Abstract As reported by FAO (2005  State of the World’s Forests  (Rome: UNFAO), 2010  Forest Resource Assessment (FRA)2010/095  (Rome: UNFAO)), Indonesia experiences the second highest rate of deforestation among tropicalcountries. Hence, timely and accurate forest data are required to combat deforestation and forest degradation insupport of climate change mitigation and biodiversity conservation policy initiatives. Within Indonesia, SumatraIsland stands out due to the intensive forest clearing that has resulted in the conversion of 70% of the island’sforested area through 2010. We present here a hybrid approach for quantifying the extent and change of primaryforest in Sumatra in terms of primary intact and primary degraded classes using a per-pixel supervised classificationmapping followed by a Geographic Information System (GIS)-based fragmentation analysis. Loss of Sumatra’sprimary intact and primary degraded forests was estimated to provide suitable information for the objectives of theUnited Nations Framework on Climate Change (UNFCCC) Reducing Emission from Deforestation and ForestDegradation (REDD and REDD+) program. Results quantified 7.54 Mha of primary forest loss in Sumatra duringthe last two decades (1990–2010). An additional 2.31 Mha of primary forest was degraded. Of the 7.54 Mha cleared,7.25 Mha was in a degraded state when cleared, and 0.28 Mha was in a primary state. The rate of primary forestcover change for both forest cover loss and forest degradation slowed over the study period, from 7.34 Mha from1990 to 2000, to 2.51 Mha from 2000 to 2010. The Geoscience Laser Altimeter System (GLAS) data set wasemployed to evaluate results. GLAS-derived tree canopy height indicated a significant structural difference betweenprimary intact and primary degraded forests (mean height 28 m ± 8.7 m and 19 m ± 8.2 m, respectively). The resultsdemonstrate a method for quantifying primary forest cover stand-replacement disturbance and degradation that canbe replicated across the tropics in support of REDD + initiatives. Keywords:  deforestation, forest degradation, change detection, remote sensing, Landsat, Indonesia Content from this work may be used under the termsof the Creative Commons Attribution-NonCommercial-ShareAlike 3.0 licence. Any further distribution of this work must maintainattribution to the author(s) and the title of the work, journal citation and DOI. 1. Introduction Deforestation and forest degradation are the second leadingcauses of anthropogenic greenhouse emissions following 11748-9326/12/034010 + 16$33.00  c  2012 IOP Publishing Ltd Printed in the UK  Environ. Res. Lett.  7  (2012) 034010 B A Margono  et al fossil fuel combustion, accounting for over 17% of globalcarbon dioxide emissions (IPCC 2007). Consequently,deforestation and forest degradation have become animportant issue concerning climate change mitigation,highlighted in the Intergovernmental Panel on ClimateChange (IPCC) Fourth Assessment Report in 2007. About75% of the emissions from tropical deforestation andforest degradation have been from developing countriescontaining large extents of tropical forest, including Brazil,the Democratic Republic of the Congo and Indonesia(IPCC 2007, MoF 2008a). Global initiatives such as the United Nations Framework Convention on ClimateChange (UNFCCC)’s program on Reducing Emissions fromDeforestation and Forest Degradation (REDD and REDD + )aim to mitigate climate change by reducing tropical forestcover loss and forest degradation.Indonesian forests account for 2.3% of global forestcover (FAO 2010) and represent 39% of Southeast Asia forest extent (Achard  et al  2002). As home to the thirdmost extensive humid tropical rainforest, Indonesia playsa significant role in overall REDD objectives. In addition,Indonesia’s forests feature high floral and faunal biodiversity(FWI/GWF 2002, MoF 2003a), the maintenance of which wouldbeanimportantco-benefitofreducingforestcoverloss.Besides the high biodiversity, almost 65 million, or about 27%of the Indonesian population depends directly on these forests(FWI/GWF 2002) for their livelihoods. As a consequenceof economic and population pressure, Indonesia experiencesone of the world’s highest deforestation rates, second onlyto Brazil (FAO 2001, 2006a, Hansen  et al  2008b, 2009) with an estimated annual gross emission from deforestationof 502 million t CO 2  equivalent (MoF 2008a). Indonesia is thus faced with a challenge appropriate for REDD insimultaneously sustaining key forest ecosystem services aswell as the livelihoods of local populations that rely on them.Forest ecosystems, notably primary forests of the humidtropics, shelter a major portion of terrestrial biologicaldiversity (MacKinnon 1997) including an estimated 80%of all terrestrial species (Carnus  et al  2006), and contain70–90% of terrestrial aboveground and belowground biomass(Houghton  et al  2009). These forests are often converted tomonoculture forest plantations (Carnus  et al  2006, Stephensand Wagner 2007) and agro-industrial estates such as oil palm (Barlow  et al  2007, Koh and Wilcove 2008), greatly reducing forest biodiversity and carbon storage of forest biomass. Theforests on the island of Sumatra in Indonesia are home to over10000 plants species, 201 mammal species, and 580 avifaunaspecies (Whitten  et al  2000, MoF 2003a). However, rapid land use conversion in support of agro-industrial development haveled to the removal of natural forest cover with a correspondingloss of biodiversity and forest carbon stocks (Whitten  et al 2000, Casson 2000). The most significant REDD initiative to date in Indonesiais the $1000000000 program of the Norway–Indonesiapartnership that has mandated a moratorium of loggingwithin Indonesia’s primary forests (Letter of Intent (LOI2010) of Norway and Indonesia, Presidential InstructionNo. 10 2011). While there are some exceptions to themoratorium (Murdiyarso  et al  2011), the principal objectiveis to reduce emissions by limiting the clearing of primaryforests. To confirm the program’s success, timely and accurateinformation on primary forest extent and its change dueto both deforestation and forest degradation is required.Improved national forest monitoring methods for Indonesiaare currently being undertaken by the Ministry of Forestryand the Indonesia Space Agency as part of the IndonesiaNational Carbon Accounting System (INCAS). However,no national-scale products have been publicly released yet.Developing cost-effective operational algorithms for primaryforest monitoring is important for verifying the performanceof the moratorium.Forest degradation has been emphasized within theinternational forestry community, including the UnitedNations Forum on Forest (UNFF), and the 2010 Targetof the Convention on Biological Diversity (CBD) (Simula2009). While the definition of forest varies considerablywithin the global forestry community (Fuller 2006), progresshas been made and nominal definitions defined in supportof REDD monitoring objectives. Quantifying and mappingforest degradation, on the other hand, is less mature andarriving at a common standard is a challenge (Simula2009). The lack of a universal definition of forestdegradation causes complications when REDD+ projects areimplemented (Sasaki and Putz 2009). Indeed, quantifying forest degradation is more difficult compared to deforestation,as deforestation represents a stand-replacement disturbanceand a permanent conversion of land use, while forestdegradation does not represent a change in land use andthe outcome is by definition still a forest land cover (FAO2004, 2007). This study reports on the quantification of  forest cover change in support of deforestation and forestdegradation mapping. Two methods are used separatelyand combined to yield a spatially explicit time seriesof stand-replacement disturbance and forest degradationfor the island of Sumatra in Indonesia. Primary forestexpanse is estimated using per-pixel classification methods.Subsequently, geographic information system-based methodsare used to incorporate the presence of human disturbancesas an indicator of forest degradation. Forest cover loss wasmapped for 2000–10 time interval using per-pixel changedetection approach pioneered by Broich  et al  (2011b).Cloud cover is a major problem in working withoptical remotely sensed data sets in humid tropical forestenvironments such as Indonesia and Brazil (Hoekman 1997,Asner 2001, Hansen  et al  2008b, 2009). Unlike Brazil, for example, Indonesia does not have a seasonally cloud-freewindow, requiring more data intensive methods to overcomepersistent cloud cover (Broich  et al  2011b). For Brazil,the regular acquisition (Fuller 2006, INPE 2012) of annual cloud-free imagery over the ‘arc of deforestation’ facilitatesthe application of advanced methods in detecting selectivelogging in quantifying degradation. Direct per-pixel methodslike those of Souza  et al  (2003) and Asner  et al  (2005)employ Landsat data to map degradation in the AmazonBasin. However, for such methods to work, images mustbe acquired within weeks of the logging event due to the 2  Environ. Res. Lett.  7  (2012) 034010 B A Margono  et al ephemeral nature of the signal in time series multi-spectralimagery. For regions with persistent cloud cover, such asIndonesia, timely data for mapping degradation using suchdirect methods are not viable.Another difference concerns the dynamics of forest coverin Indonesia, the vast majority of which does not result,strictly speaking, in deforestation. Most forest cover loss isquickly followed by forest cover gain in the form of timberplantations and palm estates (Uryu  et al  2008, Hansen  et al 2008b). Fast growing tree species (e.g.  Acacia mangium ) usedfor industrial tree plantations (MoF 2008a) grow three to five meters annually during the first five years (Matsumura 2011,Jones 2012). The combination of rapid recovery of forestcanopies and the paucity of viable cloud-free observationsposes a unique monitoring challenge.Current land cover and land use maps of Indonesia aremade via photo-interpretation methods (MoF 2011). Forest is broadlyclassifiedintoprimaryandsecondary/degradedforest,identified by the appearance of human disturbance (Adeney et al  2009, FAO 2010). The secondary/degraded forest class represents forests fragmented or affected by commerciallogging, while primary forest represents undisturbed or intactforests (MoF 2003b, 2005). Boundaries between primary and secondary/degraded forests are manually delineated bymultiple operators. The approach is time-consuming and theuse of multiple interpreters across space and through timecompromises the consistency of the output map products.Regardless, the accuracy of the forest cover classes is reportedto be high ( > 90%), based on field verification and localknowledge of the operators (MoF 2011).The objective of our research was to map the forestcoverdisturbancewithintheprimaryforestsofSumatraIsland(Indonesia) from 1990 to 2010 using Landsat data. Primaryforests were characterized into primary intact and primarydegraded subclasses using a hybrid approach. Total extent of primary forest was derived from a per-pixel direct mappingapproach and coupled with a fragmentation analysis using theIntact Forest Landscape (IFL) method (Potapov  et al  2008).The difference in the extent of primary forest and intact forestlandscape was taken as the extent of primary degraded forest.Forest cover changes for both forest cover loss due tostand-replacement disturbance and forest degradation from1990 to 2010 were mapped independently and trends of change within primary intact and primary degraded forestswere quantified. Our research aims to answer the followingresearch questions: (a) what is the extent of primary intact andprimary degraded forests in Sumatra; (b) what are the rates of primary forest cover loss, both stand-replacement disturbanceand degradation; and (c) in what official forest land use zoneshave these changes occurred? 2. Materials and methods 2.1. Definitions and rationale According to the FAO Global Forest Resource Assessment(FAO 2006a), forest is defined by the presence of trees with land covering more than 0.5 ha. The trees should be able toreach a minimum height of 5 m  in situ  and a canopy cover of at least 10%. For the purposes of national forest accountingand management, Indonesian forest is defined as an area witha minimum mapping unit of 0.25 ha that is covered by treeshigher than 5 m with a canopy cover of more than 30% (MoF2008a). For this study, we employed the Indonesian definitionof forest with a focus on forests composed of indigenous treespecies and lacking near-term evidence of stand-replacementdisturbance (FAO 2005). Forest timber and pulp plantations,oil palm estates and secondary forest are excluded from theanalysis. Our definition of primary forest includes intact anddegraded states, or natural forests consisting of native treespecies that have not been cleared and converted to other landuses.Intact forest consists of native tree species where thereare no clearly visible indications of human activities (MoF1989) and the ecological processes are not significantlydisturbed (FAO 2006b). We employed the IFL method to map the extent of primary intact forest, which is definedas an unbroken expanse of natural ecosystems within areasof current forest extent, without signs of significant humanactivity, and having an area of at least 500 km 2 (Potapov et al  2008). Degraded primary forest is a natural forestwhich has been fragmented or subjected to forest utilizationincluding wood and or non-wood forest product harvestingthat alters the canopy cover, and overall forest structure (ITTO2002). Forest management practices leading to degradation,such as selective logging, are evidenced by the presence of logging roads, logging patios, or forest canopy gaps. Primaryintact forest is mature forest absent of and removed fromsuch disturbance features. For simplicity, we define forestcover loss as an area having experienced a stand-replacementdisturbance, and define forest degradation as an area havingexperienced a transition from primary intact forest to primarydegraded forest. 2.2. Data2.2.1. Satellite imagery.  Satellite data inputs includedLandsat 7 Enhanced Thematic Mapper Plus (ETM + ) andLandsat 5 Thematic Mapper (TM) imagery downloadedfrom the US Geological Survey National Center for EarthResources Observation and Science via the GLOVIS dataportal ( ). Both Landsat archive data and Global Land Survey (GLS) data were used. All imagesfrom 1985 to 2010 with cloud cover less than 50%for 37 Landsat scene footprints covering Sumatra wereselected and downloaded. In total, 3129 ETM +  and 193TM images from 1999 to 2011, and 54 archival and 37Global Land Survey (GLS) TM images from 1985 to 1995were used in our analysis. Images were resampled to a60 m spatial resolution to reduce false change detection dueto residual misregistration effects. To remove cloud/cloudshadow affected observations, a per-pixel quality assessmentwas implemented using a set of pre-defined cloud/cloudshadow detection rules. All images were normalized usingMODIS atmospherically corrected reflectance data (Hansen et al  2008a, Potapov  et al  2012) as a normalization target 3  Environ. Res. Lett.  7  (2012) 034010 B A Margono  et al Figure 1.  (a) Eight provinces of Sumatra (Nanggroe Aceh Darussalam, North Sumatra, West Sumatra, Bengkulu, Riau, Jambi, SouthSumatra, and Lampung). (b) Forest land use zones for Sumatra. over pseudo-invariant land cover features. Red (630–690 nm),near-infrared (760–900) and short-wave infrared (1550–1750,2080–2350 nm) spectral bands were used for analysis.Source images were used to create time-sequential imagecomposites nominally centered on 1990, 2000, 2005 and2010. Additionally, a set of multi-temporal metrics (for2000–5 and 2005–10 time intervals) representing surfacereflectance change within the analyzed time intervals weregenerated, as previously described by Potapov  et al  (2012).Multi-temporal metrics are derived from sets of viableland surface observations and include minimum, maximum,median and selected percentile reflectance values per band,as well as the slope of linear regression per band versusobservation date. Due to incomplete cloud-free coverage for1990, all but 7.4% of the Sumatran land area was coveredby the resulting image composite. The time-sequentialimage composites were used for visual image interpretation,classification training and IFL mapping while multi-temporalmetrics, together with digital elevation data and slope derivedfrom Shuttle Topography Radar Mission (SRTM) (Rabus  et al 2003), were used as inputs for the supervised classification.As part of the primary intact and degraded forest covermapassessment,weusedLiDAR(lightdetectionandranging)data from the GLAS (Geoscience Laser Altimetry System)instrumentonboardtheIceSat-1satellite.GLASwaslaunchedin January 2003 and collects laser pulses in an ellipsoidalfootprint of approximately 65 m, spaced about 172 m apartalong the orbital track. We acquired the GLAS Release 28(L1A Global Altimetry Data and the L2 Global Land SurfaceAltimetryData)datasetoverSumatrafromtheNationalSnowand Ice Data Center (NSIDC, vertical waveforms of returned energy and associateddata on elevation, signal beginning, signal end and noise wereused to initially screen the data sets; additional screening wasconducted to remove the effects of cloud cover and a series of other factors before the calculation of canopy height and theheight of median energy (HOME), as described in Goetz  et al (2010) and Goetz and Dubayah (2011). 2.2.2. GIS data sets.  Official provincial boundaries(figure 1(a)), forest land use zones, and land cover digitalmaps of Sumatra were obtained from the Ministry of Forestry of Indonesia (MoF 2010). According to IndonesianForestry Law (article 6 UU-41, 1999) forest land isofficially divided into three major land use zones basedon purpose and function: protection forest ( hutan lindung ),conservation forest ( hutan konservasi ) and production forest( hutan produksi ). Production forest is further subdividedinto regular production forest ( hutan produksi tetap ), limitedproduction forest ( hutan produksi terbatas ) and convertibleproduction forest ( hutan produksi konversi ). A summary of the Indonesian forest land use zones is shown in table 1 alongwith a map of Sumatran forest land use zones in figure 1(b). 2.3. Mapping the extent of primary forest  Primary forest cover mapping employed Landsat compositesand multi-temporal metrics as input data and was performedusing a two-step supervised classification. The first stepof classification included mapping areas with tree canopycover of 30% and above for the 1990 and 2000 referenceyears. We used a decision tree algorithm, a hierarchicalclassifier that splits independent data (Landsat inputs) intomore homogeneous subsets regarding class membership(Breiman  et al  1984). The training data were a binarytraining data set of tree cover and non-tree cover, createdusing photo-interpretation of the circa 1990 and 2000 imagecomposites, respectively.The resulting tree canopy cover class was subsequentlyclassified into primary forest and other tree cover classes ina second procedure using a similarly created training data setrepresenting primary forest and other tree cover classes. The 4
Similar documents
View more...
Related Search
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

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!