Document Type : Research Paper

Authors

• Omid Rahmati ¹
• Naser Tahmasebipour ²
• Ali Haghizadeh ²
• Hamidreza Pourghasemi ³
• Bakhtiar Feizizadeh ⁴

¹ PhD Student, Agriculture Faculty, Lorestan University, Khormabad, Iran

² Assistant Professor, Agriculture Faculty, Lorestan University, Khoramabad, Iran

³ Assistant Professor, Faculty of Agriculture, Shiraz University, Iran

⁴ Assistant Professor, Department of Physical Geography, University of Tabriz, Iran

Abstract

Gully erosion is an important challenge in natural resource management and sustainable development that often has severe environmental, economic, and social consequences. Thus, the objective of the present study is to assess the capability of maximum entropy (ME) model for spatial prediction of gully erosion susceptibility at Kashkan-Poldokhtar Watershed, between Lorestan and Ilam provinces, Iran. At first, a gully erosion inventory map was produced using GPS in field surveys. The gully conditioning factors including lithology, soil texture, land use, drainage density, distance to streams, topographic wetness index, altitude, slope percent, slope aspect, plan curvature, and distance from road were selected, and their maps were prepared in geographical information system (GIS). A total of 65 gully locations were divided into two groups (1) training of the model (45 gully occurrences), and (2) validation of the model (20 gully occurrences). The prediction of gully susceptibility and variables importance analysis were carried out based on maximum entropy model using MAXENT software. Finally, the validation of the prediction results was conducted based on the receiver operating characteristic (ROC) curve method, and the area under the curve (AUC) was calculated using MedCalc software. Results indicated that highest gully erosion susceptibility is located on the center parts of the study area. According to validation results, the resulting map of areas susceptible to gully erosion obtained by ME model has a prediction accuracy of 90.7%. In addition, the results demonstrated soil texture, drainage density, lithology, and distance to streams are most important factors and their variable importance index (VII) 23, 18, 15.2, and 15.1 were obtained, respectively. However, altitude, distance from road, slope aspect, land use, topographic wetness index, and plan curvature have a less influence on gully erosion occurrence. Therefore, it was established in current study that the ME is promising of make accurate prediction in gully erosion susceptibility.

Keywords

• Erosion susceptibility
• Machine learning model
• Spatial prediction
• Conditioning factors