Raindrop size distribution estimation from S-band polarimetric radar using a regularized neural network

Radar rainrate estimates are prone to a high degree of uncertainty due to several error sources, among which the space-time variability of the raindrop size distribution (RSD). Polarimetric radar measurements enable the use of combined algorithms which reduce the sensitivity to the variability of the RSD. The aim of this work is to develop a new procedure to retrieve the RSD parameters which can be used to estimate the corresponding rainfall rates. The polarimetric variables (Zhh, Zdr, and Kdp) are used to retrieve the RSD parameters by means of an ad-hoc neural network (NN) technique. The reason for this choice is the ambition to exploit the capability of NNs to approximate strongly nonlinear functions such as those describing the relationships between radar observables and RSD parameters. A stochastic model, based on disdrometer measurements, is used to generate realistic range profiles of raindrop size distribution parameters while a T-matrix solution technique is adopted to compute the corresponding polarimetric variables at S band.