Spatial configuration is critical to mitigating carbon-water trade-offs for sustainable vegetation restoration in drylands

Vegetation restoration in drylands can enhance carbon sequestration but also intensify the regional carbon-water trade-off (CWT) by increasing water consumption. Optimizing restoration strategies to mitigate this trade-off is important for the long-term sustainability of dryland ecosystems. Therefore, this study presents a coupled assessment framework integrating Gross Primary Productivity (GPP) and Water Availability (WA) to quantify CWT, using the Loess Plateau (LP) as a representative case of large-scale dryland restoration. By integrating remote sensing and multi-source datasets with trend analyses and time-series diagnostics, this study quantified the spatiotemporal dynamics of CWT and used Random Forest models to identify key drivers and their thresholds, ultimately proposing targeted adaptive management strategies. The results showed that areas with a significant increase in GPP accounted for 90.42% of the LP, whereas areas with a significant decrease in WA accounted for 42.56%. 8.65% of the region was classified as intensified CWT zones, indicating potential hotspots of ecological degradation. Furthermore, interpretable machine learning revealed that the dominant drivers of CWT shifted from water limitation in high trade-off areas to energy limitation in low trade-off areas. These results suggest that current rapid, high-density restoration may constrain the long-term sustainability of vegetation growth, highlighting that adjusting spatial configuration is crucial for optimizing regional carbon-water relationships. Our findings characterize the spatial patterns of CWT and identify targeted mitigation strategies, providing critical insights into reconciling carbon sequestration with water consumption, which underpins sustainable vegetation restoration in the LP and other dryland ecosystems.