Estimating the nonlinear response of landscape patterns to ecological resilience using a random forest algorithm: Evidence from the Yangtze River Delta

Ecological resilience (ER) is considered a key factor in resolving complex human and natural systems conflicts and managing risk during the period of rapid development. However, the importance of landscape patterns for ER is not given enough attention in previous studies. Therefore, we selected the Yangtze River Delta region as the study area, incorporated landscape patterns into the assessment of ER and proposed the way for identifying ecological sources based on the combination of ecosystem services and landscape patterns. Combining the multidimensional ecological resilience assessment indicators system, we calculated the ER for the Yangtze River Delta region and explored the importance and nonlinear effects of landscape patterns on the ER using a random forest regression model. The results showed that the comparative analysis of ecosystem services and landscape patterns enhanced the scientificity of ecological source identification. We identified 94 ecological sources mainly in the southern of the Yangtze River Delta region, including important mountain ranges and water bodies. In addition, the ER was significantly strengthened by considering the landscape patterns. The random forest regression results indicated the nonlinear relationship between the landscape patterns and the other elements in terms of ER. This study can contribute to a comprehensive and integrated approach to the identification of ecological sources and the evaluation of ER, which can promote the conservation of regional ecological functions.