Review on Leveraging Advanced Remote Sensing and Artificial Intelligence‐Based Technologies for Mapping and Managing Land Soil Salinity

ABSTRACT

Soil salinity is a persistent environmental constraint that threatens agricultural productivity, food security and ecosystem sustainability. Advances in remote sensing (RS) and artificial intelligence (AI) are increasingly expanding the capacity to detect, map, monitor, and manage salt‐affected soils across spatial and temporal scales. This review synthesizes 61 studies published between 2010 and 2025 and identified through Scopus and Web of Science, of which 18 (29.5%) were published during 2010–2020 and 43 (70.5%) during 2021–2025. The analysis examines the principal satellite and sensor datasets, spectral indices, machine‐learning (ML) and deep‐learning (DL) methods, multi‐sensor integration strategies, and management‐oriented applications reported in the selected literature. Landsat was the most frequently used individual platform (21.3%), followed by Sentinel‐2 (14.8%), multi‐satellite approaches (13.1%), while combined Landsat–Sentinel‐2 and Landsat–MODIS applications accounted for 9.8% and 8.2% of the studies, respectively. Overall, optical RS remained dominant because of its accessibility, long‐term data availability, spectral sensitivity to soil and vegetation responses associated with salinity, and compatibility with data‐driven modelling. ML methods were more frequently applied than DL approaches and were particularly common in studies relying on relatively limited field datasets. By contrast, DL, multi‐sensor fusion, hyperspectral observations, UAV imagery, and cloud‐computing environments emerged as promising directions for increasing spatial detail, strengthening predictive performance, and improving temporal monitoring. Cross‐study comparison nevertheless reveals persistent constraints related to limited and spatially uneven field observations, inconsistent validation strategies, weak geographic transferability, inadequate uncertainty assessment, and incomplete harmonization across sensors and acquisition conditions. A further limitation is the still limited translation of salinity maps and predictive outputs into operational land‐management decisions. The evidence therefore points to a need to move beyond site‐specific accuracy optimization toward transferable, uncertainty‐aware and management‐oriented frameworks. Future research should prioritize multi‐sensor data fusion, harmonized soil‐salinity datasets, robust and spatially explicit field validation, transparent uncertainty assessment, and operational frameworks linking RS‐ and AI‐derived information with practical salinity mitigation and land‐management strategies.