Abstract:
Due to population growth and the increasing demands of human activities, land use across the world has become highly diverse. Within a limited land area, land utilisation continues to expand intensively. In this context, Remote Sensing and Geographic Information Systems enable accurate monitoring and assessment of Land-Use and Land-Cover (LULC) changes. By applying different analytical techniques, geospatial technologies are widely used to assess land-use and landcover dynamics. However, a considerable number of researchers still rely heavily on conventional classification approaches. With the advancement of modern technology, this study examines how LULC patterns can be identified and enhanced through deep learning-based techniques. The Galle District was selected as the study area, and Landsat imagery from the years 2003, 2014, and 2024 was analysed. The results reveal a rapid increase in built-up areas, which expanded from 1,206.7 km² in 2003 to 1,481.1 km² by 2024. In contrast, forest cover experienced a significant decline, decreasing from 368.04 km² in 2003 to 106.43 km² in 2024. Water bodies and barren lands also show a reduction in 2024 when compared to 2003. Accuracy assessment of the classified data indicates that Producer’s Accuracy remained above 85% throughout the entire study period, while User’s Accuracy demonstrated comparable values. Over the study period, the conversion of land cover into land-use has intensified considerably. As a result of these transformations, numerous environmental and urban challenges have emerged within the Galle District. These include increased flooding, land degradation, coastal erosion, and rising lower-atmospheric temperatures. Consequently, the findings emphasise the importance of sustainable urban development and the implementation of systematic land-use planning strategies.
