Abstract: The prediction of nature reserve distribution and identification of vacant areas have become a key research focus in ecology, particularly in macro-ecology. We established an ecological resilience evaluation system involving three dimensions “stress risk-ecosystem connectivity-recovery potential”, and then analyzed the spatio-temporal evolution characteristics of ecological resilience in the Poyang Lake Basin. Based on the ecological resilience evaluation results, we used the random forest model to predict the distribution probability of nature reserves. By comparing these predictions with the current status of nature reserves, we identified vacant areas for potential reserves, and proposed suggestions for optimization. The results showed that the overall average ecological resilience index of the Poyang Lake Basin increased steadily from 2000 to 2020, but low-value ecological resilience areas exhibited an expanding trend. Using 3 km×3 km grid as the evaluation unit, the random forest model identified 3599 suitable units for nature reserves, accounting for 20.0% of the total grid area. The existing nature reserves covered 62.7% of the suitable units, with significant vacancies. The vacant protection areas were mainly concentrated in the following five regions: the Jiujiang-Yichun vacant area, the Shangrao-Jingdezhen vacant area, the Shangrao-Yingtan-Fuzhou vacant area, the Ji’an-Ganzhou-Fuzhou vacant area, and the southern Ganzhou vacant area. This study would provide theoretical and practical insights for improving and optimizing the spatial distribution of nature reserves.

Key words: ecological resilience, random forest, nature reserve, vacancy area identification