Evaluation of spatial distribution and sensitivity of erosion gullies based on random forests in low-hill areas of Northeast China

Abstract

This study explored in depth the problem of erosion gully in farmland in the black soil zone of Northeast China. As a major food production area in China, the northeastern black soil region has been facing the challenge of increasing erosion gully problems in recent years, which not only affects soil quality and ecological environment, but also poses a threat to food security. By using high-resolution remote sensing images, the spatial and temporal distribution and morphological characteristics of erosion gullies in the region were comprehensively analysed. In terms of spatial and temporal distribution, comparison of data from 1968 to 2020 reveals that erosion gullies have increased significantly in number, area and length, reflecting the severity of the soil erosion problem. In terms of spatial distribution, the erosion gullies showed a significant aggregation effect, and this aggregation was enhanced over time. In addition, through the cold hotspot analysis, the hotspot areas of erosion gully activities were successfully identified, which provided a scientific basis for the development of prevention and control measures. In the analysis of influencing factors for the formation and development of erosion gullies, natural and anthropogenic factors, including topography, precipitation, soil and human activities, were considered comprehensively, and a comprehensive database of influencing factors was constructed. Logistic regression model and random forest model were introduced for assessment. Comparison of model performance revealed that the random forest model demonstrated higher accuracy in prediction. Therefore, a sensitivity map was produced based on the random forest model and the predicted probabilities were classified into four levels: low, medium, high, and very high sensitivity zones. The results show that the proportion of each sensitive area is relatively balanced, which provides strong support for the subsequent prediction of erosion gully risk and the development of prevention and control measures.