Soil salinity estimating from satellite dataset using multiple regression analysis over semi-arid region, UAE

Abdul Latif Jameel Water and Food Systems Lab (J-WAFS) visiting scholar Eihab Fathelrahman co-authored a research paper published in 'The arab world geographer,' titled, 'Soil salinity estimating from dataset using multiple regression analysis over semi-arid region, UAE.'

The study's researchers aimed to estimate the spatial variation of soil salinty over semi-arid land using satellite imagery and soil samples. In the paper's abstract, the authors write, 'Saline soil presents a real threat to the infrastructure and surface/subsurface resources of arid and semi-arid regions. Satellite-based datasets provide a promising source for monitoring the Earth’s surface, along with the development of computer programs and algorithms over the past decades. This study attempted to estimate the spatial variation of soil salinity over the semi-arid climate of the United Arab Emirates using the statistical technique of multiple linear regression analysis with satellite images from the Landsat 8 OLI optical sensor and Sentinel-1 SAR radar sensor. To develop a model to estimate soil salinity, 396 soil samples were collected from the study site, and soil salinity was measured. Raw OLI and SAR images were converted to reflectance values and backscatter coefficients, respectively. Multiple regression models were generated between the measured soil salinity and extracted satellite dataset, and the best performance models were chosen for validation. The validation analysis showed that the best performance model utilized reflectance values from the near-infrared, shortwave infrared, and backscatter coefficient of the VV channel. This model showed a coefficient of determination of R² = 0.5868, which makes the model unsuitable for future applications. This low model performance was expected because of the complex relationship between soil spectral reflectance and soil physical and chemical properties.'