Soil moisture estimation based on integration of optical and SAR images

In the past, different soil moisture estimation models were developed using remotely sensed data. Some of these models are based on optical images (i.e., optical models), some are based on synthetic aperture radar (SAR) images (i.e., SAR models), and a few models were developed based on integration of optical and SAR images (i.e., hybrid models). In this study, these three different types of soil moisture estimation models are compared. Results show that generally the SAR models are more accurate than the optical models, and by using the hybrid models the accuracies improve. Using a calibrated Dubois model, a root mean square surface roughness parameter was estimated for all the in situ data. This parameter was then considered in the SAR models to correct the surface roughness effects on radar backscatter coefficients. Consideration of this parameter improved the SAR models. Vegetation water content (VWC) is an important parameter in the SAR models to correct vegetation effects on radar backscattering coefficients. However, normally it is collected by extensive field work. It was shown that there is high correlation between VWC and Normalized Difference Vegetation Index by using a linear regression model. Therefore, it may be possible to estimate VWC using optical images. A new hybrid model has been developed based on integration of optical and multipolarization SAR images, and its accuracies have been assessed using ground check points. The results show that this model is the most accurate and can be used as a suitable model to estimate soil moisture. However, because limitations existed in the SAR, optical images, and the ground data used in this research, to verify the validity of obtained results for a global scale it is necessary to do more experiments using more aerial or satellite image data and also to test the models for different geophysical conditions.