Abstract: Landscape ecological risk assessment plays a vital role in identifying vulnerable ecosystem areas for targeted management. While current methods primarily rely on land-use change data for ecological risk analysis, they often lack a comprehensive evaluation of multiple factors, especially the prediction of landscape ecological risk dynamics under climate change scenarios integrating climate variations and socio-economic trends. To tackle this issue, we constructed a predictive model for ecological landscape risk influenced by diverse factors by integrating traditional landscape ecological risk assessment models with deep learning technique, and further applied this model to simulating the change in landscape ecological risks of Hanjiang River Basin. Findings reveal that: 1) during the baseline period (2000-2015), higher ecological risk levels predominantly clustered in the downstream of Danjiangkou reservoir; 2) both SSP370 and SSP585 scenarios exhibited elevated ecological risk levels, particularly concentrated in the downstream of Danjiangkou; 3) the high ecological risk area in Hanjiang River basin significantly expanded under the 2042 scenario for SSP370 and SSP585, with an average increase of 14.58% per decade under the SSP370 scenario. The proposed landscape ecological risk prediction approach in consideration of multiple factors serves as a valuable reference for ecological risk assessment in the basin under changing climatic conditions and the formulation of ecological compensation policies.

Key words: climate change, landscape ecological risk, deep learning, SSP-RCP, Hanjiang River Basin