A new method for estimation of bare surface soil moisture using time-series radar observations

This paper describes a new algorithm for the retrieval of bare surface soil moisture using dual-polarization time-series radar data. The roughness index is used to describe the soil surface roughness condition. The new algorithm assumes the roughness condition is constant over a shot time, so that the roughness index retrieval accuracy can be improved by using temporal data to minimizing the effect of radar speckle noise. The uncertainty of the roughness index is predicted by using an error propagation theory. By applying the retrieved roughness index and corresponding uncertainty as a constraint, a Bayesian approach, which takes account the uncertainties of radar observation, is implemented. The algorithm is validated with a field ground dataset at 1.25 Ghz and 40° incidence angle. The result shows an rms error (RMSE) of 0.06 cm³/cm³ for soil moisture. The correlation coefficient between retrieved soil moisture and in situ data is 0.81. Surface rms height estimates are found with RMSE of 0.41 cm and correlation coefficient of 0.99. It is shown that the new algorithm using time-series data outperforms the Bayesian approach without using temporal information and snapshot method.