Forest dynamics and above-ground forest biomass changes utilizing Google Earth Engine, machine learning, and field-based observations in the Kashmir Himalaya, India

Kashmir Himalaya, with its distinct topography and diverse vegetation, presents both the significance and challenges associated with forest monitoring and above-ground carbon stocks. This study employed extensive field-based observations, remote sensing techniques including temporal image fusion, Google Earth Engine, and machine learning algorithms to analyze forest changes and their impact on the above-ground biomass and carbon stocks in the forests of Kashmir Himalaya from 1978 to 2021. The integration of emerging image processing and machine learning techniques utilized in this study aids in addressing complexities, such as cloud and shadow interference in optical remote sensing, providing an effective substitute for complex radar-based methods. The analysis revealed a 6% reduction in forest cover, primarily in the areas proximal to human settlements. Substantial variations in above-ground biomass (1.8-1391 Mg ha -1 ) were linked to tree species, size, and site characteristics. The loss of 7.44 Mt of carbon due to a 6% reduction in forest area highlights the critical need for further enhancement of conservation efforts. Importantly, the allometric equations developed specifically for this region yielded more accurate estimates of the above-ground biomass. Their accuracy improved when species-specific normalized difference vegetation index responses were augmented with topographic information. The complexities of classifying forest cover in the Himalaya, evidenced by discrepancies in official forest surveys, underscore the importance of high spatiotemporal estimation of forest cover for precise carbon stock quantification. This study highlights the significance of remote sensing techniques complemented by site-specific observations to illustrate forest dynamics and associated changes in above-ground biomass in topographically rugged areas. • Temporal image fusion and AI/ML help overcome cloud-shadow interference. • Advanced remote sensing methods in cloud platforms offer a complementary approach to SAR. • Topographically augmented above-ground biomass models perform better in the Himalaya. • Global models used in various studies underestimated AGB in the Kashmir Himalaya. • A loss of 6% of forest and 15.83 Mt of above-ground biomass was observed in 43 years.