Gully erosion is the most severe type of water erosion and is a major land degradation process. Gully erosion susceptibility mapping (GESM)’s efficiency and interpretability remains a challenge, especially in complex terrain areas. In this study, a WoE-MLC model was used to solve the above problem, which combines machine learning classification algorithms and the statistical weight of evidence (WoE) model in the Loess Plateau. The three machine learning (ML) algorithms utilized in this research were random forest (RF), gradient boosted decision trees (GBDT), and extreme gradient boosting (XGBoost). The results showed that: (1) GESM were well predicted by combining both machine learning regression models and WoE-MLC models, with the area under the curve (AUC) values both greater than 0.92, and the latter was more computationally efficient and interpretable; (2) The XGBoost algorithm was more efficient in GESM than the other two algorithms, with the strongest generalization ability and best performance in avoiding overfitting (averaged AUC = 0.947), followed by the RF algorithm (averaged AUC = 0.944), and GBDT algorithm (averaged AUC = 0.938); and (3) slope gradient, land use, and altitude were the main factors for GESM. This study may provide a possible method for gully erosion susceptibility mapping at large scale.

Gully erosion susceptibility (GES) maps are essential for managing land resources and erosion control. Choosing the optimal horizontal resolution in GES mapping is a challenge. In this study, the optimal resolution for GES mapping in a complex loess hilly area on the Chinese Loess Plateau was tested using two machine learning algorithms. Unmanned aerial vehicle (UAV) images with a 9 cm resolution and GNSS RTK field-measured data were employed as base datasets, and 11 factors were used in the machine learning models. A series of horizontal resolutions, from 0.5–30 m, was used to determine which was the optimal level and how the resolution influenced the GES mapping. The results showed that the optimal resolution for GES mapping was 2.5–5 m in the loess hilly area, for both the random forest (RF) and extreme gradient-boosting (XGBoost) machine learning algorithms employed in this study. High resolutions overestimated the probability of gully erosion in stable regions, and it became difficult to identify gully and non-gully regions at too-coarse resolutions. The variable importance for GES mapping changed with the resolution and varied among variables. Slope gradient, land use, and contributing area were, in general, the three most critical factors. Land use remained an important factor at all the tested resolution levels. The importance of the slope gradient was underestimated at coarse resolutions (10–30 m), and the importance of the contributing area was underestimated at resolutions that were comparatively fine (0.5–1 m). This study provides an essential reference for selecting the optimal resolution for gully mapping, and thus, offers support for approaches attempting to map gullies using UAV.

Gully erosion is the most severe type of water erosion and is a major land degradation process. Predicting gully erosion susceptibility (GES) map efficiently and interpretably remains a challenge, especially in complex terrain areas. In this study, a new method called WoE-MLC model was used to solve the above problem, which combined machine learning classification algorithms and the weight of evidence (WoE) model in the Loess Plateau. The three machine learning algorithms taken into account included random forest (RF), gradient boosted decision trees (GBDT), and extreme gradient boosting (XGBoost). And the performance of the models was evaluated by the receiver operating characteristic (ROC) curve. The results showed that: (1) GES maps were well predicted by machine learning regression and WoE-MLC models, with the area under the curve (AUC) values both greater than 0.92, and the latter was more computationally efficient and interpretable; (2) The XGBoost algorithm was more efficient in GES map than the other two algorithms, with the stronger generalization ability and best performance in avoiding overfitting (averaged AUC = 0.947), followed by the RF (averaged AUC = 0.944), and GBDT algorithm (averaged AUC = 0.938); (3) Slope gradient, land use, and altitude were the main factors for GES mapping. This study may provide a possible method for gully erosion susceptibility mapping at large scale.