干旱区生态环境知识资源中心(Arid): Evaluation of the gully erosion susceptibility by using UAV and hybrid models based on machine learning

Arid

DOI	10.1016/j.still.2024.106218
	Evaluation of the gully erosion susceptibility by using UAV and hybrid models based on machine learning
	Wang, Qian; Tang, Bohui; Wang, Kailin; Shi, Jiannan; Li, Meiling
通讯作者	Tang, BH
来源期刊	SOIL & TILLAGE RESEARCH
ISSN	0167-1987
EISSN	1879-3444
出版年	2024
卷号	244
英文摘要	Gully erosion removes the top layer of fertile soil, which is essential for agriculture and plant growth. Gullies can alter or destroy natural habitats for plants and animals. This disruption can lead to a loss of biodiversity and negatively impact wildlife populations. Thus, it was crucial to study its susceptibility for gully erosion control and prevention. Three hybrid models were built based on Multi-Objective Optimization by Ratio Analysis (MOORA). The input of the MOORA was determined by frequency ratio (FR), and the criteria of these input was determined by using three classification algorithms, including random forest (RF), LightGBM (LG) and Catboost (CB). The objective of the hybrid models was to evaluate the gully erosion susceptibility (GES) in a small watershed in Ordos city, China. Total of 488 gully heads with 15 conditioning factors were extracted based on digital ortho map (DOM) and digital elevation map (DEM) conducted by UAV, which merged the separated photos captured by UAV and Sentinel-images. These gully heads were utilized to build the gully inventory data set. Furthermore, frequency ratio (FR) was utilized to study the spatial correlation between the conditioning factors and the gully presence pixels, whereas three clarification algorithms were used to classify the gully presence and absence pixels, and also determine the relative importance of the 15 conditioning factors. Three hybrid models, named MOORA-FR-RF, MOORA-FR-LG, and MOORA-FR-CB were utilized to establish the gully erosion susceptibility mapping (GESM). The receiver operating characteristic curve (ROC) and the area under curve (AUC) as well as the Kappa coefficient were utilized to evaluate the accuracy of three hybrid models. The result showed that LG processed the highest accuracy in gully pixels classification, and the slope steepness, distance to road, and drainage density were significantly contributed to the occurrence of gully. The MOORA-FRLG had the highest performance with an AUC of 0.997, followed by MOORA-FR-LG (AUC = 0.994) and MOORAFR-RF (AUC = 0.980). Therefore, it was concluded that MOORA-FR-LG is the most efficient approach to predict gully erosion susceptibility in the studied watershed. The results are helpful to gully erosion prevention measures in the arid and semi-arid regions.
英文关键词	Gully erosion susceptibility mapping Hybrid model MOORA Machine learning Ordos
类型	Article
语种	英语
收录类别	SCI-E
WOS记录号	WOS:001266395300001
WOS关键词	HILLY LOESS PLATEAU ; CERTAINTY FACTOR ; WATER EROSION ; RESOLUTION ; IMAGERY ; REGION ; FOREST
WOS类目	Soil Science
WOS研究方向	Agriculture
资源类型	期刊论文
条目标识符	http://119.78.100.177/qdio/handle/2XILL650/405657
推荐引用方式 GB/T 7714	Wang, Qian,Tang, Bohui,Wang, Kailin,et al. Evaluation of the gully erosion susceptibility by using UAV and hybrid models based on machine learning[J],2024,244.
APA	Wang, Qian,Tang, Bohui,Wang, Kailin,Shi, Jiannan,&Li, Meiling.(2024).Evaluation of the gully erosion susceptibility by using UAV and hybrid models based on machine learning.SOIL & TILLAGE RESEARCH,244.
MLA	Wang, Qian,et al."Evaluation of the gully erosion susceptibility by using UAV and hybrid models based on machine learning".SOIL & TILLAGE RESEARCH 244(2024).