Abstract
<title>Abstract</title> <p>Understanding the spatially heterogeneous drivers of water yield is essential for effective environmental monitoring and landscape-oriented water-resource management. Taking the Liaodong Green Economic Zone as the study area, this study integrated the InVEST model, landscape pattern indices, and multiscale geographically weighted regression (MGWR) to investigate the spatiotemporal dynamics of water yield and its landscape-pattern drivers from 2003 to 2023. The results show that water yield exhibited a fluctuating but overall increasing trend during the study period, with pronounced interannual variability. Spatially, water yield displayed a clear gradient decreasing from southeast to northwest, reflecting strong spatial heterogeneity. Landscape pattern analysis indicated a general tendency toward reduced fragmentation, enhanced connectivity, and increased compositional diversity at the landscape level. MGWR results revealed significant spatial heterogeneity in the relationships between landscape pattern indices and water yield. Mean patch area and contagion index were predominantly positively associated with water yield, whereas patch density, aggregation index, and Shannon’s diversity index mainly exhibited negative effects. At the class level, farmland-related landscape indices were predominantly negatively associated with water yield across most areas, while forest proportion showed predominantly positive effects. Overall, the integrated InVEST–MGWR framework effectively captured the spatially heterogeneous influences of landscape pattern on water yield, providing a practical methodological reference for environmental monitoring and landscape-oriented water-resource management in ecologically heterogeneous regions.</p>