ABSTRACT Understanding the complex linkages between landscape patterns and ecosystem service multifunctionality (ESMF) and ecosystem service (ES) trade‐offs is crucial for sustainable land management. However, few studies have examined how these relationships vary non‐linearly and spatially under alternative future development pathways. This study developed an integrated framework that combined land‐use simulation, ES assessment, machine learning, and spatial analysis. It was employed to examine how landscape patterns influence ESMF and ES trade‐offs in 2040 under the business‐as‐usual (BAU), cropland protection (CP), and ecological priority (EP) scenarios in the Xiamen–Zhangzhou–Quanzhou metropolitan area. The results show that the EP scenario generally provided the favorable basis for maintaining multifunctionality and avoiding severe trade‐offs. Landscape composition dominated variation in ESMF and ES trade‐off intensity, whereas landscape configuration moderated ES trade‐offs across scenarios and sub‐basins. The combined partial dependence plot (PDP) and geographically weighted regression (GWR) analysis further show that landscape‐ES relationships were governed by scenario‐dependent non‐linear response ranges together with strong local spatial heterogeneity. Forest cover generally promoted ESMF, with the most effective response range occurring at approximately 40%–50%. By contrast, built‐up land consistently suppressed ESMF and shifted several trade‐off relationships toward low‐level equilibrium state. Cropland became more influential under the CP and EP scenario, and the ecological meaning of landscape diversity varied across local landscape contexts. These findings show that similar landscape changes may generate contrasting ES outcomes across scenarios and sub‐basins, and they provide a spatially explicit, scenario‐dependent basis for differentiated land management.
Zhu et al. (Sun,) studied this question.