In the latter half of 2015, fires raged once again across Indonesia, creating a state of emergency due to poisonous smoke and haze across Southeast Asia as well as incurring great financial costs to the government. Within the ESA Fire CCI Project the Remote Sensing Solutions GmbH in Munich mapped the burned area across Sumatra, Kalimantan and West Papua for 2015.
For their study they used Trimble’s eCognition software in combination with Sentinel-1 Synthetic Aperture Radar (SAR) data promising to provide improved detection of land use and land cover changes in the tropics as compared to methodologies dependent upon optical images. First they used eCognition Developer for the creation of an automated rule set to be applied across vast areas (1,158,481 km² (Sumatra, Kalimantan and West-Papua)). Therefore, the rule set needed to be very robust and transferable. Moreover, to process the large amount of data (10 m spatial resolution and multiple time-steps (several TB)) they used the eCognition Server for batch and parallel processing. For the classification, eCognition’s object-based image analysis (OBIA) techniques such as, statistical features of objects, contextual features, and object refinement were used to create a meaningful and highly accurate classification of burned area. Moreover, creating objects also reduces the impact of speckle effect greatly which makes live more easy working with this kind of data.
Their study presents the first spatially explicit estimates of burned area across Sumatra, Kalimantan, and West Papua based on high-resolution Sentinel-1A SAR imagery. The study revealed that huge amounts of areas burned (4,604,569 hectares (ha)) during the 2015 fire season (overall accuracy 84%), and compared this with other existing operational burned area products (MCD64, GFED4.0, GFED4.1s).
This research nicely demonstrates that eCognition is an effective tool for the classification of vast amounts of SAR data.