Assessment of land degradability and water quality in a semiarid region: Machine learning approach and multi scale analysis

A significant environmental challenge is to examine the relationship between agricultural land abandonment (ALA), land degradation, and soil properties, as well as their effects on water quality. This study developed a geo-environmental modeling approach to assess the connections between ALA, land degradability, and water quality in the Khuzestan Plain of southwestern Iran. The Random Forest (RF) was employed to map land degradability, and spatial correlation analysis was utilized to assess its relationship with water quality at various spatial scales during both dry and wet periods. The RF-based land degradability map revealed the presence of low to moderate land degradation in the region. The eastern and southeastern parts of the region exhibited the highest levels of degradation, while the central parts, as well as those in the north and west, demonstrated the least degradation. Severely degraded regions corresponded to abandoned agricultural lands. The Water Quality Index (WQI) results showed that 54% of water samples were classified as poor (50 ≤ WQI <75) during the dry period, while 96.5% were deemed unsuitable (WQI ≥100) during the wet period. The correlation between land degradability and WQI was slightly stronger in the dry period, peaking at the 500 m buffer. This finding highlights the importance of collecting data from various sources for land degradability mapping, exploring different spatial and temporal scales of water quality, and comprehending the impact of buffer zones on land degradation. These insights are essential for effective water and land protection as well as the development of sustainable, nature-based policies. • Significant spatial correlations between land degradability and WQI were found in the 500 m buffer. • The highest proportion of land degradability was associated with abandoned agricultural areas. • Water quality was poorer during the wet period compared to the dry period. • Random Forest demonstrated strong performance and efficiency in generating land degradability maps.