Three Decades of Land Cover Dynamics in a Boreal Coastal Basin: A Multisensor Spectral Index and Machine Learning Approach Using Landsat Data and GB-SAR Data

Boreal coastal habitats, with their complex hydrological, ecological, and geological functions, play a pivotal role in regulating regional climate, supporting biodiversity, and sustaining local communities. This study presents a comprehensive, multidecadal assessment of land cover dynamics, a large and environmentally sensitive coastal-fluvial watershed. Using a 30-year Landsat satellite data archive (1990–2020) from Landsat 4, 5, 7, 8, and 9 sensors, we analyzed long-term changes in land cover patterns, focusing on vegetation health, surface water extent, and urban expansion. A suite of spectral indices, including the normalized difference vegetation index, normalized difference built-up index, normalized difference water index, and modified normalized difference water index, was calculated for each time slice to capture key environmental variables related to vegetation productivity, urbanization, and hydrological dynamics. To ensure methodological consistency and computational efficiency, a cloud-based processing workflow was implemented using the Google Earth Engine platform. This approach enabled automated preprocessing, atmospheric correction, cloud masking, and multidecadal time-series analysis across the study area. Landsat surface reflectance and spectral index data were fused and subsequently classified using a random forest machine learning algorithm, allowing for enhanced land cover classification accuracy across seven time intervals spanning three decades. Our findings reveal distinct spatiotemporal trends in the Moose River Basin’s landscape. A marked decline in vegetation cover was observed during the early 2000s, likely driven by a combination of climate variability, hydrological fluctuations, and local disturbances. Urban and built-up areas, primarily concentrated near river corridors and accessible regions, exhibited gradual but steady expansion over the study period, signaling increasing human settlement pressures. © 2025 Elsevier B.V., All rights reserved.