Prediction and Selection of Appropriate Landscape Metrics and Optimal Scale Ranges Based on Multi-Scale Interaction Analysis

Landscape metrics are widely used in landscape planning and land use management. Understanding how respond with scales can provide more accurate prediction information; however, ignoring the interference of multi-scale interaction may lead to a severe systemic bias. In this study, we quantitatively analyzed scaling sensitivity based on predict their optimal scale ranges. Using big data method, multivariate adaptive regression splines model (MARS), partial dependence (PHP), studied relationships changing scales. The results show that commonly exists most metric responses, making significant contribution. general, effects three (i.e., spatial extent, resolution, classification use) often different direction, resolution is primary driving isolation. findings only few highly sensitive throughout whole spectrum, while other limited within certain threshold range. This study confirms scaling-sensitive scalograms be as an application guideline for selecting appropriate