Machine learning-based temporal change detection of land use and land cover changes in the Tunçbilek open pit coal mine region using Planetscope imagery

This study investigates land use and land cover (LULC) changes in the Tunçbilek open-pit coal mine and its surroundings, a region experiencing intense mining activity in western Türkiye. Understanding LULC dynamics is crucial for assessing the long-term environmental impacts of surface mining operations and supporting sustainable land management. High-resolution PlanetScope imagery from 2016 and 2021 was used in conjunction with two supervised machine learning algorithms Maximum Likelihood Classification (MLC) and Support Vector Machine (SVM) to detect temporal changes in six land cover classes. The results show that SVM outperformed MLC in classification accuracy. The kappa values for MLC were 0.73 (2016) and 0.72 (2021), whereas SVM achieved 0.87 and 0.84, respectively. SVM also provided higher user and producer accuracy rates, particularly for the forest and planted classes. Between 2016 and 2021, notable land cover transitions were observed, including a 6.83% increase in cultivated lands and a 7.9% decrease in barren land. The mining area itself expanded by approximately 1.39%. These results highlight the effectiveness of machine learning-based remote sensing methods in monitoring LULC changes and contribute to a better understanding of the environmental impacts of mining activities in complex and sensitive landscapes.