Spatial modelling of gully headcuts using UAV data and four best-first decision classifier ensembles (BFTree, Bag-BFTree, RS-BFTree, and RF-BFTree)

摘要: Abstract Despite the importance of delineating spatial modelling gully headcuts (GHs) in erosion-prone environments, assessments factors that control occurrence is lacking. To fill this gap research, we identified 129 GHS field surveys. These cases were randomly divided into two groups: 90 GHs (70%) for model training and 39 (30%) validation. Subsequently, new unmanned aerial vehicle (UAV) imagery used to develop predict location at sites prone soil erosion Golestan Province, Iran. Mapping enables evaluation 4 machine-learning techniques (or ensembles) – best-first decision tree (BFTree), bagging (Bag-BFTree), random-subspace (RS-BFTree), rotation-forest (RF-BFTree) GHs. We use information-gain ratio method analyze relationships between 22 GH conditioning factors. The ensemble outputs are validated using a receiver operating characteristic (ROC) curve. areas under curves (AUCs) prediction rates methods applied group BFTree 88.3%, Bag-BFTree 92.7%, RS-BFTree 95.7%, RF-BFTree 93.2%. AUCs model-validation cases, however, 84.9%, 94.1%, 97.4%, 9.18%, respectively. Therefore, is, statistically, most effective accurate Variable-importance analyses indicate out GH-influential factors, land use, slope degree, slope-length more developing occurrence. Finally, address need detailed observations highly data field, UAV image-acquisition technologies demonstrated.