TITLE:
Spatiotemporal Analysis of Forest Cover Change Using Sentinel-2 Imagery and Maximum Likelihood Classification: A Case Study of Shimla, India
AUTHORS:
Shivam Gupta, Shivai Gupta
KEYWORDS:
Forest Cover Monitoring, Land Cover Change Detection, Maximum Likelihood Classification, Sentinel-2 Satellite Imagery, Normalized Difference Vegetation Index (NDVI)
JOURNAL NAME:
Advances in Remote Sensing,
Vol.15 No.3,
August
28,
2026
ABSTRACT: Forests represent a critical component of land-based ecosystems, making accurate forest cover assessment essential for sustainable landscape management. This study presents a two-part methodology combining Maximum Likelihood Estimation (MLE) and Sentinel-2 satellite imagery to support forest monitoring and spatial mapping. Unlike conventional forest mapping studies that primarily report land cover statistics, our work develops an integrated hybrid classification framework combining NDVI transformation and ISODATA spectral clustering with Maximum Likelihood Classification to quantify forest canopy transitions in a complex Himalayan landscape. Remote sensing and GIS techniques were applied using NDVI classification and pixel-based extraction, with unsupervised ISODATA clustering and ERDAS Imagine software employed in the classification workflow. Forest cover was categorized into five canopy density classes: Very Dense Forest (VDF), Moderately Dense Forest (MDF), Open Forest (OF), Scrub, and Non-Forest (NF). The study area encompassed Shimla Municipal Forests in the Western Himalayas, with change detection conducted over a four-year interval from 2015 to 2019. Results revealed a net loss of 113 ha in open forest and 48 ha in non-forest areas. The classification achieved an overall accuracy of 89.2% with a Kappa coefficient of 0.86, reflecting strong agreement between classified outputs and reference data. These findings demonstrate the effectiveness of MLE paired with Sentinel-2 imagery in accurately detecting forest cover change, while also offering a cost-effective solution and beyond estimating forest extent, the proposed framework provides a reproducible workflow for analyzing canopy density transitions that can support long-term environmental monitoring and evidence-based forest management in mountainous regions on a large scale.