Unveiling nonlinear driving mechanisms and ecological thresholds of natural and tourism cultural ecosystem services in oasis cities: Insights from an XGBoost-SHAP framework

In oasis tourism cities with limited ecological carrying capacity, rapid urbanization intensifies the competition between economic development and ecological conservation. Understanding the nonlinear trade-offs between Tourism Cultural Ecosystem Services (TCES) and natural Ecosystem Services (ESs) is critical for sustainable management. Taking Jiuquan City—a typical Silk Road oasis—as a case study, this research integrates the InVEST model with Kernel Density Estimation to quantify the spatiotemporal evolution of Water Yield (WY), Soil Retention (SR), Carbon Storage (CS), Habitat Quality (HQ), and TCES. We further construct a diagnostic framework using the Geographical Detector and XGBoost-SHAP models to identify nonlinear driving mechanisms and ecological thresholds. The results indicate that: (1) ESs exhibit significant spatial heterogeneity, characterized by a stable “high-in-the-south, low-in-the-north” pattern. TCES show continuous growth and agglomeration, spatially segregating from regulating services, revealing intense competition between tourism and conservation. (2) Trade-offs remain structurally stable, with WY generally trading off against other services while HQ, CS, and SR exhibit strong synergies; however, Geographically Weighted Regression (GWR) reveals marked spatial heterogeneity in the intensity and direction of these interactions, broadly corresponding to the mountain–oasis topographic gradient. (3) The “Geographical detector—XGBoost-SHAP” framework reveals that natural factors dominate regulating services, while tourism facilities drive TCES. Crucially, key factors exhibit significant nonlinear threshold effects: rainfall impacts on water yield show saturation points, while tourism facilities demonstrate exponential growth characteristics for TCES upon breaching specific thresholds. This study confirms the importance of threshold-based nonlinear management. It provides a scientific basis for zoning management and threshold early warning systems to resolve the “resource dependence vs. ecological constraints” conflict in arid oasis tourism cities.