Delineating fragmented grassland patches in the tropical region using multi-seasonal SAR and Optical Satellite Images

• DOI: 10.32942/osf.io/rpwv9
• Publication Date: 2020-06-20T00:18:12Z
• Authors: Abhishek Samrat; M. S. Devy; Ganesh
• Keywords: 2. Zero hunger; bepress|Life Sciences; 13. Climate action; Life Sciences; 15. Life on land

Globally grasslands are declining and are in highly degraded conditions. In south Asia grasslands are neglected and treated as wastelands. They remain unprotected, highly fragmented, and poorly understood which has led to a loss of unique biodiversity and livelihoods. Mapping grasslands accurately is a challenge and current maps based on optical remote sensing often over- or underestimate grasslands in south Asia due to a prevalant complex landscape matrix, small patch sizes, and obscuring monsoonal clouds. Synthetic Aperture Radar (SAR) fused with moderate spatial resolution has been used to delineate grasslands but, high-resolution, freely available ESA’s sentinel-1(SAR) and -2(optical) provides an opportunity to map small and fragmented patches that were not possible earlier with the publically available moderate or medium spatial resolution remote sensing dataset. Further, high resolution imageries require high computing power which is often limited with stand alone machines. Here we demonstrate that using cloud computing and optimal use of multi-seasonal imagery one can obtain a highly accurate land cover/use classification for a complex habitat matrix. We used freely accessible cloud computing platforms like Google Earth Engine (GEE) and land cover/use classification of sentinel-1 and -2. We compared the accuracy of grassland delineation between 1) seasonal (pre, during, and post-monsoon) sentinel-1, 2) post-monsoon sentinel-2, and 3) combined sentinel-1 and -2. We tested this method at two sites in a highly fragmented habitat matrix in semi-arid areas of Western and Southern India. The classification result has shown the overall accuracy of for the combined image was higher than only sentinel-2 and sentinel-1 alone for both sites. Grasslands habitat accuracy was also consistent with combined image classification across the sites. Our results identified newer grassland areas that coarse landuse management maps used by the government did not. The computation was done on a basic laptop and processing completed very quick. We, therefore, suggest that this novel approach of using cloud computing and optimal use of resource-hungry (computation and storage) high-resolution ESA’s sentinel-1 and -2 data, can be used to identify major land classes and small patchy grassland in the semi-arid regions of Asia and has the potential to map at continent level.