C-band dual polarization radar retrieval of rainfall: application of an iterative technique with embedded neural-network

Polarimetric radar measurements at C-band require correction for propagation effects before using them for rainfall retrieval. Since differential phase shift Kdp is affected by the spatial variation of the backscattering differential phase shift δ, a new neural-network estimation technique is applied to remove δ effects on Kdp estimate. In this study we also explore a new hybrid iterative technique with constraint to correct for attenuation and differential attenuation based on a neural-network scheme and a differential phase constraint. Numerical simulations are used to investigate the efficiency of this approach with respect to others profiling techniques. The simulator is based on a T-matrix solution technique, while the hydrometeor distribution has been characterized with respect to dielectric composition (water, ice and mixed phase), raindrop size distribution (normalized gamma distribution), shape (ellipsoid with parameterized aspect ratio) and angle orientation. A sensitivity analysis is performed in order to evaluate the expected errors of this method to system bias. The performance of these correction procedures and the effects of an error bias on radar measurements are evaluated by using mono-dimensional Gaussian raincell models. Numerical results are discussed in order to show the potential and robustness of the proposed technique.