Integrated assessment of soil salinity dynamics on the dried Aral Seabed based on field observations, geochemical analysis, remote sensing, and machine learning
Year: 2026
Type: article
Abstract: Soil salinization is a major driver of land degradation in arid environments, particularly in desiccated inland basins such as the Aral Sea region. This study provides an integrated assessment of soil salinity dynamics on the dried Aral Seabed (Aralkum Desert, Kazakhstan) by combining field-based geochemical analysis, remote sensing, and machine learning. Field investigations conducted during 2020–2022 comprised 58 soil profiles and 204 soil samples analyzed for electrical conductivity (EC), total dissolved solids (TDS), major-ion composition, and vertical salinity distribution across different solonchak types. Geochemical analysis revealed pronounced spatial and genetic heterogeneity, with marine-derived solonchaks exhibiting hypersaline conditions (Σ > 250 mmolc/100 g −1 , or 16.1%) dominated by Na Cl systems, whereas typical and takyr-like solonchaks showed moderate salinity associated with capillary rise and continental processes. Principal Component Analysis identified a dominant salinity gradient controlled by Na + , Cl − , and SO₄ 2− , with clear vertical differentiation. EC and TDS showed a strong linear relationship ( r = 0.91, p < 0.001), supporting EC as an operational proxy for soil salinity. Remote sensing analysis using SAVI, MSAVI, NDSI, and SI4 showed strong spatial correspondence between salinity and vegetation patterns. Vegetation indices were negatively correlated with salinity indicators ( r = −0.54 to −0.72, p < 0.01), while high NDSI and SI4 values coincided with elevated EC–TDS and low SAVI/MSAVI values. Random Forest classification showed 75.86% overall agreement with the 58 field-reference observations (Cohen's Kappa = 0.691), with classification errors concentrated mainly in intermediate salinity classes. Soil salinity on the dried Aral Seabed represents a spatially continuous process controlled by groundwater dynamics, evaporative concentration, and inherited marine sediments. Integrating geochemical, remote sensing and machine-learning approaches provides an effective framework for regional assessment of salinity-driven land degradation in arid environments.
Source: CATENA
Institutions: Al-Farabi Kazakh National University, Ionosphere Institute
Language: English
Subfield: Aquatic Science
Domain: Life Sciences
SDGs: Life on land, Zero hunger
Open Access status: hybrid
Estimated APC paid: $3,680
OpenAlex ID: W7222815642