Monitoring the salinization of agricultural land and assessing its drivers in the Altay region

Abstract

Soil salinity in the Altay region presents a formidable challenge that greatly jeopardizes the sustainability of human agriculture. Thus, it is necessary to monitor the salinization of agricultural land in the region and assess the corresponding drivers. A random forest regression model was constructed using the Google Earth Engine platform and the accuracy of different image inversion models was compared to achieve precise large-scale spatial inversion and the dynamic monitoring of topsoil salinity. Furthermore, augmented regression tree models were utilized for the quantitative monitoring and analysis of salinity drivers. The feature importance ranking method identified six vegetation and salinity spectral indices (brightness index, enhanced normalized difference vegetation index, salinity index 2, the improved salinity monitoring index, and salinity index) to exert the most significant impact on salinity. A comparison of the random forest models constructed from Landsat and Sentinel-2 data revealed that the degree of soil salinization in the study area was more accurately predicted using the former, achieving an R2 value of 0.905 and RMSE value of 2.8 g/kg. A general decline was observed in the salinization of agricultural land in the Altay region between 2000 and 2010, with 52.36 % experiencing reduced salinization. However, mild and moderate salinization areas in the Fuhai and Altay regions expanded during this period, increasing by 267 km2 and 496.59 km2, respectively. Irrigation water gradually emerged as a pivotal factor influencing the spatial distribution of topsoil salinity in agricultural fields within the study area. Furthermore, following the transition from traditional water-saving to information-based efficient water-saving, the volume of irrigation water emerged as a pivotal factor impacting topsoil salinity in the study area, with a contribution rising from 11.4 % to 18.8 %, thus placing it as the most important factor influencing salinity. This work enhances the understanding of the trends and drivers of agricultural land salinization, which is of paramount significance for addressing national food security, efficiently utilizing saline land, and mitigating the escalation and spread of salinization.

Keywords

• Altay region
• Driving factors
• Random forests regression
• Salinity index
• Soil salinization
• Vegetation index