Landscape Ecological Risk Assessment and Driving Factors During 1995–2024 in the Dianzhong Five Lakes Region of Yunnan Province, China Using the XGBoost-SHAP and Random Forest Models

The assessment of landscape ecological risk and the exploration of its driving factors is a critical approach to alleviating the conflict between the growing demand of human activities and ecological environment conservation, and the Five Lakes Area in Central Yunnan serves as a typical representative of landscape ecological risk issues in plateau lake regions. Therefore, this study, based on the land use transfer change characteristics of the Five Lakes Area in Central Yunnan across four periods (1995–2024), employed the landscape pattern index method to calculate the spatiotemporal variation characteristics of the landscape ecological risk index; additionally, 10 driving factors (including natural and socio-economic factors) were selected, and the XGBoost-SHAP model and Random Forest model were applied to explore the driving factors, with the results showing that: (1) In terms of land use transfer, farmland, forest, and Grass land were transferred among each other, the inflow of Construction land increased, and Grass land had the largest outflow area; (2) regarding landscape ecological risk, the landscape pattern was unstable, the loss degree increased, and the moderate and moderately high-risk areas expanded; and (3) for driving factors, the dominance shifted from natural factors to socio-economic factors; among these, Precipitation, NDVI (Normalized Difference Vegetation Index), Land use intensity, and Night-time light index were significant influencing factors. Based on the above results, a zoning management and control strategy for landscape ecological risk was proposed, aiming to provide a scientific reference for policy formulation to reduce risks and alleviate human–land conflicts.