Glacier boundary extraction and spatiotemporal variation analysis in Geladandong region

Glaciers, as sensitive indicators of global climate change, play a crucial role in influencing the global water cycle, sea level rise, and ecosystem dynamics. Understanding the interactive mechanisms between glacier boundary changes and multidimensional factors such as climate and topography is essential for revealing the complex relationships underlying the ecological functions supported by glacier systems. This study proposes a Remote Sensing Index-Random Forest fusion method (NDI-RF) to map glacial extent. The NDI-RF approach combines remote sensing index techniques with Random Forest modelling, ensuring the extraction accuracy of the Random Forest model while effectively enhancing boundary extraction precision in mosaic pixel scenarios. Then, the spatiotemporal changes in glacier extent and their response to climate change were analyzed. The experimental results indicate that the NDI-RF method can reduce spectral confusion on ice lakes and thin ice surfaces to a certain extent. The extraction results show Kappa coefficient, OA, F1-score, recall, and Precision values of 0.92, 0.94, 0.92, 0.88, and 0.93, respectively, all of which outperform the Normalized Difference Snow Index (NDSI) extraction method and the Random Forest model. From 2000 to 2024, glaciers in the Geladandong region have mainly been shrinking in area, with a total reduction in glacier area of 110.29 km² and an average annual area change rate of 0.47%. Among these years, the period from 2010 to 2015 was marked by the most significant glacier retreat, with a reduction of 36.87 km², and also saw the highest glacier area change rate. Analysis based on different terrain conditions showed that glaciers retreated more notably at altitudes below 5,250 m, with slopes greater than 45°, and on north-west facing slopes. Over the past 25 years, the average annual temperature and total precipitation have shown a fluctuating upward trend. The glacier area shows a negative correlation with the average annual temperature.