Modelling spatio-temporal soil salinity of farmlands in the hyperarid environment of Kuwait using remote sensing data

Abstract

Soil salinity in hyperarid regions is a significant environmental problem resulting from high temperatures, high evaporation rates, and low precipitation, negatively impacting agricultural development and food security. Kuwait's desert features limited arable land, where soil salinity poses a significant challenge for farming, resulting in soil degradation and crop loss. This paper aimed to map the expansion of surface soil salinity during the last 25 years in Kuwait's Al-Wafra and Al-Abdaly farmlands through utilising spectral indices extracted from multi-seasonal imageries and regression-based modelling. The seasonal variation influence spectral indices behavior demonstrated statistically significant supported the inclusion of a seasonal term in regression modelling and improved model performance. Among the models evaluated, the linear configuration with ASTER_SI, Normalised Difference Salinity Index (NDSI), and Season achieved the best performance (R² = 0.651, RMSE = 0.182), offering an optimal balance between predictive accuracy and interpretability. Temporal generalisation is tested during 2000 and 2024, and ecological consistency is assessed using NDVI. The results revealed that in conjunction with farm expansion, the non-saline areas in Al-Abdaly decreased from 62.9% in 2000 to 9.5% in 2024, and in Al-Wafra from 85% to 14.2%. The highly to strongly saline surfaces increased by 3.3% in Al-Abdaly and by 2.3% in Al-Wafra in the same period. This study confirms that soil salinity in Kuwait's farmland has increased over the last 25 years.