ANNOUNCEMENTS
Soil salinity, which is induced by climate change and rising sea levels, is one of the most serious environmental dangers, affecting agricultural activity in coastal areas in most tropical climates. In the Uttar Pradesh districts of Firozabad and Etah, as well as Bhavnagar, Gujarat, this problem has become more serious. The key purpose of this study is to map soil salinity encroachment in Firozabad and Etah Districts in Uttar Pradesh, as well as Bhavnagar in Gujarat, using Sentinel-1 SAR C-band data, Sentinel-2 Optical data, and Landsat-8 data, as well as five machine learning models, Random Forest (RF), Support Vector Regression (SVR), Gaussian Processes (GP), K Nearest Neighbor (DT). 115 soil samples were gathered for this purpose during a field survey done from June 16 to June 20, 2021, according to Sentinel-1 SAR, Sentinel-2 images, and 25 soil samples were obtained during a field survey conducted in last week of April 2013, corresponding to Landsat-8 data. The root-mean-square error and the correlation coefficient were used to evaluate and compare the performance of the five models (r). The results showed that the GP model (PH) (RMSE = 0.77 and R = 0.74) and RF model (ESP) (RMSE = 9.71 and R = 0.64) had the best prediction performance in SAR imagery, the SVM model (ESP) (RMSE = 12.42 and R = 0.67) had the best prediction performance in Sentinel-2 imagery, and the GP model (PH) (RMSE = 0.29 and R = 0.79) had the best prediction performance in Landsat imagery. We can summarize that advanced machine learning techniques can be used to map soil salinity in arid and semi-arid regions, delivering a valuable product for farmers and policymakers to aid in the selection of improved crop varieties in the face of climate change.
