Forest disturbance detection in Garhwal Himalayas using MODIS NDVI time-series and BFAST model

Abstract

In this study, the BFAST model was used to decompose MODIS NDVI time-series and detect negative abrupt changes within trend during the study period (2005 to 2018). These negative abrupt changes were filtered and assigned as disturbances in the selected forest classes. The results showed that on average 15.01% of the forests were disturbed annually and 86.84% area was affected during the whole study period. The selected forests were affected the most in year 2008, 2011, and 2016. The model validation established a strong agreement between measured and estimated disturbance with 92.63% overall accuracy and kappa coefficient value of 0.86. The information about the time, extent, and impact of disturbances derived from remote sensing and GIS might be a critical tool for the degradation evaluation of forest ecosystem and also help in the development of forest management policies for sustainable development of the Himalayan region.