Abstract
The protection of agricultural lands is crucial for maintaining food security and environmental sustainability, especially in regions undergoing rapid urbanization. This study examines temporal land use changes in the Tahtalı Neighborhood of Derince District, Kocaeli Province, between 2014 and 2024. Three zones along the lakeshore, identified as highly attractive for human activity, were selected as study plots. High-resolution satellite images, orthophotos, and drone data were processed using the MeanShift segmentation method and classified with the Support Vector Machine algorithm to differentiate built-up areas from agricultural lands. Training samples were prepared based on spectral, textural, shape, and contextual features to ensure high classification accuracy. The classification performance was evaluated using confusion matrices, yielding overall accuracies ranging from 0.90 to 0.97 and Kappa coefficients between 0.72 and 0.93 across the study zones and years. The results indicate a substantial decline in agricultural lands due to urban expansion, temporary structures, and stone wall constructions, reflecting the increasing pressure of urbanization on rural landscapes. The integration of highresolution drone data with machine learning classification effectively captured both fine-scale and broader land use patterns, offering a robust framework for monitoring spatial and temporal land use dynamics. These findings support informed land management decisions and emphasize the importance of sustainable land use policies in rapidly developing peri-urban regions.