Integrated approach to land degradation risk assessment in arid and semi-arid Ecosystems: Applying SVM and eDPSIR/ANP methods

Abstract

• The land degradation hazard and vulnerability maps were combined to identify high-risk areas. • Socio-environmental vulnerability to land degradation was assessed using the Analytic Network Process (ANP). • Low-altitude areas with low rainfall and high temperatures are most susceptible to land degradation hazard. • Land use, vegetation cover, and population density were identified as key factors contributing to vulnerability. • The final risk map shows that approximately 7.56% of the region is classified as high-risk. The degradation of land (LD) is a major concern for the health and sustainability of natural resources. It is primarily caused by the deterioration of vegetation and soil. In this study, we focus on assessing the risk of LD in the Bakhtegan basin in Iran, one of the arid and semi-arid ecosystems. Our objective is to predict LD hazard and vulnerability maps, and then combine them to identify areas at high risk. To predict LD hazard, the Support Vector Machine (SVM) algorithm was used with 179 LD locations and twelve variables, including land use, lithology, rainfall, temperature, distance to the stream, elevation, aspect, slope, curvature, distance to the road, Normalized Difference Moisture Index (NDMI), and population density. The LD hazard map was evaluated using five error statistics extracted from the contingency table. For LD vulnerability mapping, eight criteria were deployed, including land use, population density, Normalized Difference Vegetation Index (NDVI), livestock density, groundwater quantity and quality, Salinity Index (SI), and migration. These criteria were weighted using the integrating eDPSIR framework and Analytic Network Process (ANP) methods. The results show that low-altitude areas which have low rainfall and high temperatures face the highest LD hazard. Additionally, the western and northwestern regions of the basin are more vulnerable compared to other areas due to factors such as land use and vegetation cover. Lastly, the LD risk map reveals that some 7.56 % of the region falls into the high-risk classification, totaling 2413.37 km 2 . Notably, salt lands emerge as the most at-risk land use, with 77 % under high risk, with rain-fed agricultural land following as the second-highest risk class. These findings underscore the importance of considering LD risk in land management strategies.

Topics & Concepts