Coherent Model of L-Band Radar Scattering by Soybean Plants: Model Development, Evaluation, and Retrieval

An improved coherent branching model for L-band radar remote sensing of soybean is proposed by taking into account the correlated scattering among scatterers. The novel feature of the analytical coherent model consists of conditional probability functions to eliminate the overlapping effects of branches in the former branching models. Backscattering coefficients are considered for a variety of scenarios over the full growth cycle for vegetation water content (VWC) and the complete drydown conditions for soil moisture. The results of the coherent model show that HH scattering has a significant difference up to 3 dB from that of the independent scattering when VWC is low, e.g., 0.2 kg/m2. Forward model calculations are performed for the scattering from the soybean field for the full range of three axes of root-mean-square (RMS) height of bare soil, VWC, and soil moisture using the coherent model. The soybean volume scattering including the double-bounce term is combined with the back scattering of bare soil from the numerical Maxwell solutions that incorporates RMS height, soil permittivity, and correlation length, to form the forward model lookup table for the vegetated soil. The results are compared with data from 13 soybean fields collected as part of the soil moisture active passive validation experiment 2012 (SMAPVEX12). Time-series retrieval of soil moisture is also applied to the soybean fields by inverting the forward model lookup table. During the retrieval, the VWC is optimized with physical constraints obtained from ground measurements. The retrieval performances are significantly improved using the proposed coherent model: the root-mean-squared error (RMSE) of the soil moisture retrieval is decreased from 0.09 to 0.05 cm3/cm3 and the correlation coefficient is increased from 0.66 to 0.92.