Spatial Representation Effects on Ecosystem Service Trade-Offs, Bundles, and Associated Factors: Implications for Sustainable Land Management on the Loess Plateau, China

Reliable ecosystem service (ES) assessments are essential for sustainable land management and ecological restoration; yet, their implications for decision-making may vary with the spatial units used for analysis. This study examined how spatial representation influences ES relationships, ecosystem service bundles (ESBs), and associated environmental and socioeconomic factors across the Loess Plateau, China, from 1990 to 2020. Five ESs, including water yield, carbon storage, soil conservation, habitat quality, and food production, were quantified using the InVEST models and the Revised Universal Soil Loss Equation based on multi-source remote sensing and environmental datasets. Five spatial representations, comprising 5 km and 15 km grids, county-level and prefecture-level administrative units, and hydrological basins, were compared using Pearson correlation analysis, geographically weighted regression, self-organizing maps, and random forest modeling. The results showed that ES relationships, bundle configurations, and the relative importance of associated factors varied substantially across spatial representations. Fine-resolution grids retained local ecological heterogeneity and revealed more spatially fragmented ES patterns, whereas administrative and basin units emphasized broader regional associations and aggregated ecosystem-function patterns. These findings indicate that assessments based on a single spatial representation may overlook localized ecological conditions or misrepresent patterns relevant to management at broader scales. From a sustainability perspective, integrating complementary spatial representations can provide a more comprehensive evidence base for identifying localized ecological restoration priorities, coordinating land-management responsibilities, and addressing hydrologically connected processes. This study demonstrates that spatial representation is not merely a technical choice but an important consideration in translating ES assessments into spatially appropriate management strategies. The findings provide methodological and practical insights for improving the reliability of ES assessments and supporting sustainable land management and ecological restoration in environmentally vulnerable regions. Spatial representation is an important source of uncertainty in ecosystem service (ES) assessments because different aggregation schemes may alter the observed relationships among ecosystem services, their bundles, and associated factors. This study investigated how spatial representation affects ES interactions, ecosystem service bundles (ESBs), and associated factors across the Loess Plateau, China, from 1990 to 2020. Five ecosystem services, including water yield, carbon storage, soil conservation, habitat quality, and food production, were quantified using the InVEST models and the Revised Universal Soil Loss Equation based on multi-source remote sensing and environmental datasets. Five spatial representations, including 5 km and 15 km grids, the county scale, the prefecture-level city scale, and the basin scale, were compared using Pearson correlation analysis, geographically weighted regression, self-organizing maps, and random forest modeling. The results showed that ES relationships were highly sensitive to spatial representation. Fine-scale grids preserved local heterogeneity and revealed more fragmented trade-offs and ESBs, whereas broader spatial units emphasized stronger regional synergies and aggregated ecological patterns. Spatial aggregation also changed the relative importance of environmental and socioeconomic factors associated with ES variations. These findings highlight that spatial representation should be considered a critical component of remote sensing-based ES assessments and ecological management.