Soil salinity prediction models constructed by different remote sensors

As a significant environmental problem, soil salinity should be timely and accurately mapped and monitored. In recent years, remote sensing data and techniques have been widely used for soil salinity estimation. However, with the difference in the sensors' characteristics, satellite-based prediction of soil salinity remains highly uncertain. This study investigates and compares soil salinity models from remote sensing sensors with different spectral, and most importantly, spatial resolution. Thus, data from two middle (Landsat - 8 and Sentinel - 2) and one high spatial resolution (PlanetScope) sensors have been used for salinity indices have been used for developing salinity prediction models from in-situ data. For this purpose, data from random points in different agricultural fields have been collected in the study area. The developed statistical models were validated using 20% of the dataset using accuracy indices. The results showed that the higher spatial resolution tent to give a better model prediction. However, it also means that the higher the spatial resolution of the imagery, the more complex the prediction model will be developed. The results also showed that simpler models give a higher correlation between the observed and the predicted values. For future studies, we recommend a similar investigation using different sensors and more in-situ data over similar agricultural fields.