基于决策树的天山冰湖提取方法研究

he Tianshan region, located in the middle of the Eurasian continent, is a major distribution area of modern glaciers, and its glacial meltwater has formed a large number of and widespread glacial lakes. Glacial lakes are vital indicators of climate change and an important source of surface and groundwater supply in arid and semi-arid regions of Northwest China. The impact of topographic factors and spectral characteristics of ground objects makes it difficult to distinguish between glacial lakes, mountain shadow, and snow cover when ex-tracting glacial lakes from remote sensing images using a single water index. In this study, based on the Google Earth Engine platform and Landsat 8 remote sensing images as the data source, a decision tree extraction method of the glacial lake was constructed according to the topographic characteristics (slope, elevation, and buffer analysis) and spectral characteristics of the glacial lake. This method was compared with the NDWI and MNDWI threshold methods. Experimental results demonstrated that the decision tree method can effectively reduce the im-pact of mountain shadow and snow cover and accurately extract glacial lake information. The overall accuracy of the extraction results was 89.14%, the kappa coefficient was 0.783, and the F1 score was 87.85%. The decision tree method, which combines spatial topographic features and spectral features, is a relatively efficient extraction method for dynamic monitoring and research analysis of glacial lakes.