Out-scaling apparent conductivity from ground acquired electromagnetic induction using satellite reflectance data to assess regional salinity and waterlogging

The salt-affected coastal zone of the Ganges-Brahmaputra Delta (CZGBD, ∼37,400 km 2 ), is home to approximately 54 million people, many of whom face livelihood challenges. Primarily an agricultural economy, soil salinity and waterlogging during the dry season (Rabi, from November to March), plus climate induced natural disasters and inadequate infrastructure, significantly decrease food security. The spatial delineation of a salinity/waterlogging proxy, Apparent Conductivity (ECa), can indicate locations of impaired agricultural productivity, and thereby improve the efficiency of limited supplies of irrigation water. To assist in this, a novel technique was developed to integrate and out-scale 19 ground-based electromagnetic induction ECa surveys conducted in agricultural sites with Sentinel-2 10 m reflectance data. These data are used to predict 5 ECa classes (from low to severe) regionally from 2019 to 2023 using a Random Forest model, with a cross-validation accuracy of 66 %. The observed regional ECa dynamics are qualitatively linked to antecedent monsoon rainfall, discharge and the hydromorphology of the CZGBD. Relatively higher monsoon rainfall, particularly late in the season, results in higher ECa classes that are related to waterlogging in areas to the north of the CZGBD. Higher ECa classes are also related to soil salinity in areas of decreased fluvial activity over time, particularly in the southwest of the CZGBD. These maps can be used to guide future investments in waterlogging mitigation (drainage systems), prioritize irrigation supply network upgrades, support improved farming systems choices including reducing the risks of Rabi season crop failure and inform supply chain investment or management decisions. • Maps show soil salinity and waterlogging risks in the Ganges Delta from 2019 to 2023. • ECa data from ground surveys and Sentinel-2 used to classify salinity into 5 classes. • Random Forest model predicts salinity with 66 % accuracy using satellite reflectance. • Salinity and waterlogging patterns are linked to regional hydromorphology and climate. • Results support better irrigation, drainage, crop planning, and supply chain decisions.