Assessing urban resilience in Northwest China using a resistance-recovery-adaptability framework and explainable machine learning approach.

Against climate change and increasing human activities, cities face severe external shocks. Urban resilience under the "Resistance-Recovery-Adaptability" framework offers an effective approach to risk governance, especially for Northwest China's ecologically fragile and socioeconomically underdeveloped cities. However, systematic resilience assessments and nonlinear dynamic analyses remain insufficient. This study built an RRA-based index system covering economic, social, and ecological subsystems to evaluate the resilience of 33 prefectural-level cities in Northwest China from 2000 to 2020, and applied XGBoost-SHAP to identify influencing factors. Results show that urban resilience increased on average, but the growth was unevenly distributed with provincial capitals (e.g., Xi'an, Lanzhou, Urumqi) increasing much faster than peripheral prefectures. GDP contributed most positively. The population showed strong spatial heterogeneity and turning points. Water factors displayed nonlinear effects. Major obstacles included water infrastructure, R&D investment, and consumption vitality. This research supports arid-region resilience research and high-quality urban development in Northwest China.