Forests in Southeast Asia are experiencing some of the highest rates of deforestation and degradation in the world. The main drivers of land cover change include expansion of agriculture and plantation estates, extraction of natural resources, infrastructure development, and small and large-scale deforestation. The underlying drivers of land cover change include a complex interaction of human populations within environmental, political, and economic dynamics. As a consequence, wildlife, hydrological and ecological functions, and local communities that rely on forest resources are experiencing significant negative impacts.

The expansion of monoculture plantations has long been one of the most powerful drivers of native forest degradation in this part of the world. Land use and forestry scientists and practitioners are working diligently to stem this tide of degradation but many challenges exist, one of which is clearly understanding where plantation expansion is occurring. As ground mapping is impractical, satellite imagery must be used.  However, persistent cloud cover in the tropics, as well as the similar appearance of different types of trees when fully vegetated, means that satellite images are often less useful than they should be. That’s where SIG came in. In order to better map how and where degradation is most rapidly and extensively occurring, SIG, in cooperation with WWF and IKEA, developed a method to accurately map rubber and palm oil plantations using a combination of imagery from Landsat-8, Sentinel 1 and 2 satellites.

In a study of forest plantations in Myanmar, SIG scientists created yearly composites for all three sensors and combined the data into a single composite image. Using statistical techniques and reference data of land cover classes –  including surface water, forest, urban and built-up, cropland, rubber, palm oil and mangrove – biophysical probability layers for each class of land cover were developed. Further statistical techniques then formed these layers into land cover and probability maps. The results were validated with high resolution imagery, and overall showed good accuracy (84%). After taking into account spatially explicit estimates of uncertainty to weight the validation points, the results were 91% accurate.

The study demonstrated how data fusion of different satellite images and spatially explicit error quantification in a cloud-based platform can be used for earth science applications. It also provided valuable insights into improving discrimination between natural and anthropogenic forest types. This is particularly applicable in regions where natural forest is quickly being replaced with plantations, and may not be detected with traditional remote sensing methods. The results are helping guide sustainable management of wood resources in Myanmar and present a way to improve sustainable supply chain analyses and associated environmental recommendations.