Deforestation rates in insular Southeast Asia between 2000 and 2010

Abstract

Insular Southeast Asia experienced the highest level of deforestation among all humid tropical regions of the world during the 1990s. Owing to the exceptionally high biodiversity in Southeast Asian forest ecosystems and the immense amount of carbon stored in forested peatlands, deforestation in this region has the potential to cause serious global consequences. In this study, we analysed deforestation rates in insular Southeast Asia between 2000 and 2010 utilizing a pair of 250 m spatial resolution land cover maps produced with regional methodology and classification scheme. The results revealed an overall 1.0% yearly decline in forest cover in insular Southeast Asia (including the Indonesian part of New Guinea) with main change trajectories to plantations and secondary vegetation. Throughout the region, peat swamp forests experienced clearly the highest deforestation rates at an average annual rate of 2.2%, while lowland evergreen forests declined by 1.2%/yr. In addition, the analysis showed remarkable spatial variation in deforestation levels within the region and exposed two extreme concentration areas with over 5.0% annual forest loss: the eastern lowlands of Sumatra and the peatlands of Sarawak, Borneo. Both of these areas lost around half of their year 2000 peat swamp forest cover by 2010. As a whole this study has shown that deforestation has continued to take place on high level in insular Southeast Asia since the turn of the millennium. These on-going changes not only endanger the existence of numerous forest species endemic to this region, but they further increase the elevated carbon emissions from deforested peatlands of insular Southeast Asia thereby directly contributing to the rising carbon dioxide concentration in the atmosphere.

Multiple-scale prediction of forest loss risk across Borneo

“…Land cover change data for [2000][2001][2002][2003][2004][2005][2006][2007][2008][2009][2010] We used land cover maps produced for the years 2000 and 2010 by the Centre for Remote Imaging, Sensing and Processing (CRISP; Miettinen et al 2011Miettinen et al , 2012. These maps were based on MODIS surface reflectance product images (NASA 2010), Shuttle Radar Topography Mission (SRTM) 90 m version 4 digital elevation information (Jarvis et al 2006) and several peatland distribution maps Subagjo 2003, 2004;Selvaradjou et al 2005;Wahyunto et al 2006).…”

“…These maps were based on MODIS surface reflectance product images (NASA 2010), Shuttle Radar Topography Mission (SRTM) 90 m version 4 digital elevation information (Jarvis et al 2006) and several peatland distribution maps Subagjo 2003, 2004;Selvaradjou et al 2005;Wahyunto et al 2006). Miettinen et al (2011) used a three step classification process to classify the landscape into 13 categories: water, mangrove, peat swamp forest, lowland forest, lower montane forest, upper montane forest, plantation/regrowth, lowland mosaic, montane mosaic, lowland open, montane open, urban and large scale palm plantation. Accuracy assessment for these maps found overall accuracies of 83% for 2000 and 85% for 2010, with class accuracies between 75 and 85% for the non-forest classes and up to 97% for the forest classes (Miettinen et al 2011(Miettinen et al , 2012.…”

“…Miettinen et al (2011) used a three step classification process to classify the landscape into 13 categories: water, mangrove, peat swamp forest, lowland forest, lower montane forest, upper montane forest, plantation/regrowth, lowland mosaic, montane mosaic, lowland open, montane open, urban and large scale palm plantation. Accuracy assessment for these maps found overall accuracies of 83% for 2000 and 85% for 2010, with class accuracies between 75 and 85% for the non-forest classes and up to 97% for the forest classes (Miettinen et al 2011(Miettinen et al , 2012. Overall, the accuracy assessment of these maps compares favourably to other studies (e.g., Clark et al 2010: 79.3%;Friedl et al 2010: 75%;Rozenstein and Karnieli 2011: 81%;Rodriguez-Galiano et al 2012: 92%).…”

“…The classes used in the analysis include: (1) water, (2) mangrove, (3) peat swamp forest, (4) lowland forest, (5) lower montane forest, (6) upper montane forest, (7) plantation or regrowth, (8) lowland mosaic, (9) montane mosaic, (10) lowland open, (11) montane open, (12) Urban (Miettinen et al 2011(Miettinen et al , 2012. For each of these classes we used FRAGSTATS 4.0 (McGarigal et al 2012) to calculate the percentage of the focal landscape covered by each class within the six focal window extents (1, 10, 20, 30, 40 and 50 km).…”

“…Each of these four metrics was also calculated for focal windows at the six spatial extents. FRAGSTATS metrics were calculated for the land-use maps in both the year 2000 and 2010 (Miettinen et al 2011(Miettinen et al , 2012, the data from the year 2000 was used to calibrate the models while the 2010 data was used to predict the risk of future forest loss. All spatial layers were resampled to a 500 m pixel size and projected to an Albers conformal conic projection for analysis.…”