Potential of WorldView-3 for soil salinity modeling and mapping in an arid environment

Soil salinity is gaining international consideration due to its effects on crop production and subsequent food security and ecosystem services. Therefore, salinity assessment and mapping is essential to understand the nature and extent. In this study, a semi-empirical model is developed and validated for soil salinity prediction and mapping in an arid environment especially for WorldView-3 (WV-3). The methodology is based on spectroradiometric measurements and the electrical conductivity, determined in the laboratory (EC- Lab ) for model-calibration process, considering 160 soil samples with various degrees of salinity. The model was then implemented on a WV-3 image, which was radiometrically calibrated and atmospherically corrected. The derived large-scale WV-3 predictive salinity map (EC-Predicted) was validated with respect to the ground truth (EC- Lab ). Results showed that only the last four SWIR bands (5 to 8) correlate significantly with EC- Lab (R 2 > 0.71), and the proposed model discriminated significantly among all the considered salinity classes (R 2 of 0.83) using second order polynomial regression (p Predicted ) based on the developed model show a good correlation (R 2 of 0.91) with the ground truth (EC- Lab ) data and an acceptable RSME of 2.43 dS/m.