Collection of Time-Series Datasets from Sentinel and Landsat

Abstract

Satellite time series data plays a critical role in the monitoring of forest and vegetation changes in India's diverse forest landscape. This review is based on the processing of Sentinel, Landsat and Global Forest Watch-derived datasets and their availability for forest monitoring in India, specifically for Maharashtra and central Indian landscapes. The process of data set collection is described through the methodological steps of sensor selection, temporal coverage, data pre-processing, harmonisation, data validation, metadata documentation and integration of the collected data with derived forest monitoring indicators. Landsat offers long-term consistency for monitoring forest-cover and tree-cover changes, whereas Sentinel supplies recent fine resolution observations for detecting fine resolution vegetation change, disturbance, canopy condition and fire. The review also leverages the Maharashtra case dataset, which includes tree-cover extent, annual tree cover loss, primary forest loss, VIIRS fire alerts, biomass, carbon flux, plantations and dominant drivers of loss. These indicators help in the interpretation at the state and district level, but are subject to nuanced distinction between tree cover, natural forest and derived analytical products. The biggest challenges are the cloud cover in the monsoons, the deciduous nature of the phenology, the variation in the sensors, mixed pixels, the lack of ground validation, and the uncertainty in distinguishing between seasonal vegetation changes and degradation. The review underscores the need for harmonized, metadata-rich, machine learning enabled forest datasets at various regional scales in India to support monitoring; fire-risk assessment; biomass estimation; conservation planning and predictive vegetation-change modelling.