Analyzing spatial non-stationarity effects of driving factors on landslides: a multiscale geographically weighted regression approach based on slope units

Landslides pose a significant threat to the safety of people and their property. Previous landslide susceptibility assessment efforts have often overlooked the impact of spatial variations in the distribution of driving factors on disaster occurrences. Geographically Weighted Regression (GWR) is the most commonly used method for spatially heterogeneous modeling. However, it uses a single bandwidth and cannot explain the spatially varying scaling parameters of each factor. This study focuses on Luhe as the research area and introduces the Multiscale Geographically Weighted Regression (MGWR) model. By considering different spatial scales, the spatial non-stationarity of the relationship between landslides and their driving factors was explored. The final results indicate that the MGWR model outperforms the global regression Ordinary Least Squares (OLS) model and the traditional GWR model. Within the study area, the influence of various factors on landslides is complex, exhibiting a multipolar pattern in space. Elevation emerges as the dominant driving factor in the research area, showing a negative correlation with landslides. By employing the concept of multiscale spatial local regression, one can better analyze the interaction patterns between factors and landslides, providing improved insights for disaster prevention and mitigation.