Geospatial Analysis of Agroforest Landscape in the Central Western Ghats, India, using Google Earth Engine and Machine Learning

Abstract Assessing land cover land use change (LCLU) in the Central Western Ghats, India is crucial for evaluating ecosystem degradation and environmental processes influenced by climatic and anthropogenic pressures. Despite widespread awareness of the links between biodiversity loss and human wellbeing, the species continue to disappear at an alarming rate. This study investigates the LCLUC dynamics over the past two decades (2002–2023) in Central Western Ghats, spanning an area of 10,419 km² in Karnataka, India. The Landsat 5 and Landsat 8 satellite imagery served as primary data sources, with Machine Learning Algorithm employed in the Google Earth Engine, a cloud-based platform, to classify images using a Random Forest classifier validated with 17,639 ground training points. The analysis revealed that, dense forest covered largest area (6926.31 ha) followed by open forest (1257.92 ha) and cropland (1195.73 ha). Agroforest and waterbody occupied 249.65 ha and 265.18 ha, respectively, while built-up areas, barren land, and plantation forest constituted 131.38 ha, 83.45 ha and 127.29 ha, respectively. The results indicate a 7.14% reduction in dense forest area, contrasted with a 1.09% increase in agroforestry. The classification achieved an overall accuracy of 94.08% and Kappa index of 0.95. These findings highlight the importance of integrating measures and policies with investment in research and satellite monitoring to curb deforestation in the Central Western Ghats and ensure continuity of Western Ghats in halting degradation and climate change.