Spatial heterogeneity of gully erosion and its dominants in multiple spatial scales in the rolling hilly region of Northeastern China

Gully erosion is a common global degradation process. Gully erosion shows a high spatial heterogeneity because its influencing factors varies greatly in different spatial scales. However, gully erosion and its attribution analyses at different scales are lacking. Given this, gully interpretation was conducted in 5005 grids (1 km * 1km) sampled from an area of 17.75×10 4 km 2 by stratified random sampling method to obtain gully area density in grid, geomorphic zone and whole region scales in the black soil region of Northeast China. Climate, topography, land cover and human activity factors including 18 potential parameters were selected to explore the causes of spatial differentiation in gully erosion by using interpretable machine learning models. The research results indicated that the correlation coefficient and Nash’s efficiency coefficient from the interpretable machine learning models are greater than 0.77 in all zones, and its simulation accuracy is satisfactory. Regionally, the annual average temperature significantly drove the spatial differentiation in gullies, increasing it above 3°C and inhibiting it below this threshold. The dominant factors for plains, platforms, hills, and mountains were snow cover days, annual average temperature, annual average temperature, and annual minimum temperature, respectively. At the grid scale, the annual average temperature and annual maximum temperature were identified as dominant drivers of gully densities in 23.31% and 22.42% of the study area, respectively. The main controlling factors had a significant spatial scale effect on gully erosion. Thus, it is recommended to formulate prevention and control measures based on the dominant factors at multiple scales to decelerate gully erosion, protecting the ecology and land resources.