Assessing vegetation–terrain relationships based on the fusion of high-resolution ULS and SfM data for a topographically complex catchment in the semi-arid area, China postprint

Abstract: Topography plays a crucial role in regulating vegetation distribution through its control on solar radiation, soil moisture, and erosion-deposition dynamics. However, fine-scale vegetation–terrain relationships remain insufficiently understood because previous studies have relied on coarse-or single-sourced data and lacked an integrated analytical framework. This study integrated unmanned aerial vehicle laser scanning (ULS) and structure-from-motion (SfM) photogrammetry by mapping SfM orthophoto texture onto the ULS point cloud, to generate a unified high-resolution vegetation–terrain dataset for the Qiaogou catchment in the hilly and gully Loess Plateau, China. Nine topographic factors were derived from ULS digital elevation models (DEMs), while texture and vegetation indices were extracted from SfM imagery. The vegetation–terrain relationship was quantified using a hierarchical framework that sequentially evaluated the linear effects using ordinary least squares (OLS), nonlinear effects using generalized additive models (GAM), random forest (RF), and extreme gradient boosting (XGB), spatial non-stationarity using geographically weighted regression (GWR), and interaction among explanatory factors using Geodetector. Results showed that vegetation cover exhibited pronounced spatial heterogeneity, with dense vegetation primarily located on moist valley floors and sparse cover on steep and rugged slopes. Elevation and slope imposed the primary topographic constraints on vegetation cover, while terrain roughness index (TRI), topographic position index (TPI), topographic wetness index (TWI), and slope length factor (LS) provided complementary descriptors of local terrain variability and moisture conditions. Nonlinear models found that slope (45°) and wetness thresholds determined ecological stability limits, while spatial models revealed a significant non-stationarity (Moran's I = 0.352) in topographic influence. Geodetector interaction analysis indicated that terrain effects were expressed mainly through factor combinations, with several slope-related pairs providing additional explanatory power beyond individual factors, while the single-factor explanatory power for overall vegetation cover remained modest. By fusing ULS and SfM data and integrating multi-level analytical models, this study provided a process-based and scale-consistent understanding of how geomorphic structure and hydrological connectivity co-regulated vegetation heterogeneity. The proposed framework offered a robust reference for high-resolution vegetation monitoring, eco-geomorphic modeling, and restoration planning in complex loess landscapes.