Mapping rodent-induced disturbance of biological soil crusts using PlanetScope imagery and deep-learning-informed graph cut

This dataset contains annual maps of burrow-like surface patterns detected from PlanetScope imagery across the Gurbantunggut Desert, China, for 2017–2025, as generated in our study on rodent-induced disturbance of biological soil crusts (biocrusts). Two versions of the mapping results are provided. The **original products** contain all burrow-like patterns detected by the Swin-UPerNet model and subsequently refined through morphology-guided probability adjustment and graph-cut-based boundary optimization, without restricting them to biocrust-covered surfaces. These patterns therefore represent remotely detected features with the characteristic brightness and morphology of rodent burrow systems. The **biocrust-filtered products** are the final datasets used in the manuscript. They retain only detected patches occurring in areas with sufficient surrounding biocrust cover, following the biocrust fractional cover screening procedure described in the study, and therefore represent the mapped surface disturbance of biocrusts associated with rodent excavation.