Integrating Machine Learning and Scenario Simulation to Decouple Multi‐Scale Ecosystem Service Trade‐Offs/Synergy in Shaanxi Province

ABSTRACT The complex terrain and significant spatial heterogeneity of ecosystem services (ESs) in Shaanxi Province (SXP) make it crucial to analyze their multi‐scale trade‐offs/synergies and driving mechanisms for regional ecological management. This study integrates machine learning (SRF, SVM, etc.) with scenario simulation (PLUS‐ InVEST) to evaluate the spatiotemporal differentiation and interaction effects of water production (WY), carbon storage (CS), habitat quality (HQ), soil conservation (SC), and nitrogen and phosphorus storage (NS/PS) under urban priority development (CPD), ecological priority (EPD), and inertial development (ID) scenarios from 2000 to 2040. The multi factor driving mechanism was revealed through geographic detectors, structural equation models (SEM), and ecosystem service bundles (ESB). The results show that: (1) In the future scenario, the urban expansion in Guanzhong (CS) is significant, with an overall increase in WY (EPD scenario+19.58 mm), a decrease in CS in CS (−1.5 t/ha), a decrease in SC in northern Shaanxi (NPS), and an increase in southern Shaanxi (SPS); (2) In 2020, WY‐CS and WY‐HQ showed a significant trade‐off in NPS/CS, while WY‐NS/PS showed synergy in NPS/SPS. By 2040, the explanatory power of human activities on ESs has increased ( q value increased by 72.4%); (3) The driving factors are ranked as climate > vegetation > terrain > humanities (pre > gpp > ndvi > slp). SEM shows that the direct effect of terrain factors on SC decreased from 0.812 (2000) to 0.296 (2020); (4) ESB identification indicates that CS needs to optimize land use to restore ecology, SPS should increase forest coverage, and NPS needs to strengthen degraded land restoration. This study provides data support and decision‐making basis for multi‐scale ecological collaborative governance.