Land use and land cover (LULC) change represents a major driver of environmental degradation in forested landscapes undergoing increasing anthropogenic pressure. This study investigates multi-temporal land cover dynamics in the Shafarood watershed, northern Iran, over a 20-year period (2000–2020) using Landsat satellite imagery and Support Vector Machine (SVM) classification. Landsat 7 ETM+ (2000) and Landsat 8 OLI (2020) Level-1 terrain-corrected images were geometrically verified and analyzed following optimal band selection using the optimum index factor (OIF). Four land cover classes were identified: forest, agriculture, dense rangeland, and semi-dense rangeland. Classification accuracy was high for both years, with overall accuracy values of 96.75% (Kappa = 0.9472) for 2000 and 98.96% (Kappa = 0.9307) for 2020. Results reveal a net loss of 413 ha of forest and 577 ha of dense rangeland over two decades. In contrast, agricultural land expanded by 191 ha and semi-dense rangeland increased by approximately 800 ha, indicating vegetation degradation and land use conversion trends. The findings highlight ongoing ecological pressure in the Hyrcanian forest region and emphasize the importance of continuous satellite-based monitoring for sustainable land management. This case study demonstrates that medium-resolution satellite imagery combined with machine learning classification provides a reliable, cost-effective, and reliable framework for long-term environmental monitoring and sustainable land management.
Hashemi et al. (Thu,) studied this question.