Land cover and forest health indicator datasets for central India using very-high resolution satellite data

Despite increased availability and accessibility of satellite data, approaches for detecting forest degradation have been limited. We produce a very-high resolution 3-meter (m) land cover dataset and develop a normalized index, the Bare Ground Index (BGI), to detect and map exposed bare ground within forests at 90 m resolution in central India.

Tree cover and bare ground was identified from Planet Labs Very High-Resolution satellite data using a Random Forest classifier, resulting in a thematic land cover map with 83.00% overall accuracy (95% confidence interval (CI): 61.25% - 90.29%). The BGI is a ratio of bare ground to tree cover and was derived by aggregating the land cover. Results from field data indicate that the BGI serves as a proxy for intensity of forest use although open areas occur naturally.