Automatic Detection and Quantification of Erosional Badland Levelling in Central India Using LandTrendr with PlanetScope Imagery in Google Earth Engine

The highly dissected morphology of the Chambal badlands characterizes the Lower Chambal Valley in Madhya Pradesh, Central India. It is considered the heaviest form of land degradation through gully erosion in the entire country and ranges among the largest badland zones in the world. In order to combat the loss of land and decline in agricultural productivity through badland formation, land levelling by local communities and farmers as well as in governmental reclamation projects has become widespread. The relief in the Chambal badlands is anthropogenically altered by infilling valley bottoms or smoothing shallow badlands. While this may help to increase agricultural area and productivity, there is evidence that soil quality decreases and erosive processes increase after land levelling. This study aims to identify and quantify the anthropogenic reshaping of relief in the Chambal badlands using the cloud-computing platform Google Earth Engine (GEE) and imagery data archives. Our method is based on the GEE implementation of the time ­series analysis algorithm LandTrendr (Landsat-based Detection of Trends in Disturbance and Recovery). While originally developed to identify disturbances in forested regions, LandTrendr can be applied to various landscapes and land cover changes. Since land levelling patterns in the Chambal badlands occur at various spatial scales, we have adapted the GEE algorithm to process data from the high-resolution PlanetScope archive as opposed to the originally implemented medium-resolution Landsat data. Land levelling is accompanied by a removal of the badland vegetation cover of shrubs, trees, and occasional patches of moderately dense forest. Thus, annual time series of vegetation indices are used to detect newly levelled areas at pixel-level. The high temporal resolution of PlanetScope allows to calculate vegetation index values from cloud-free scenes from approximately the same date every year. The algorithm is tested in a small study area within the Chambal badlands; upon successful implementation it may be extended to a large-area analysis of anthropogenic relief reshaping in the entire Chambal Valley. Furthermore, our LandTrendr implementation of PlanetScope imagery in Google Earth Engine will allow to monitor future land levelling and agricultural reclamation activities in the unique geomorphological and ecological environment of the Chambal badlands.