Reliable Assessment of Long-Term Land Use Change to Support Sustainable Watershed Management Using Multi-Sensor Landsat Imagery

Land use change analysis is widely applied in watershed studies, yet its reliability is often limited when multi-temporal classification accuracy is not systematically evaluated. This study aims to assess long-term land use change while emphasizing the accuracy and reliability of change detection in the Petung Watershed, East Java, Indonesia. Multi-temporal Landsat imagery from Landsat 7, 8, and 9 (2005–2024) was analyzed using a supervised classification approach implemented in ArcGIS. Classification accuracy was evaluated using a confusion matrix and the Kappa coefficient to ensure temporal consistency and methodological reliability. The results indicate a clear shift in land use patterns, characterized by a continuous decline in vegetation cover and a substantial expansion of agricultural and built-up areas. Classification reliability is supported by Kappa coefficient values ranging from 0.701 to 0.829, corresponding to substantial to almost perfect agreement, with the highest accuracy achieved in the later observation years. These findings demonstrate that the proposed supervised, GIS-based framework provides a reliable and replicable approach for accurately measuring long-term land use change at the watershed scale. While the methodology can be applied to other regions, the observed land use dynamics are specific to the Petung Watershed and reflect local environmental and anthropogenic conditions.