Polarimetric Rainfall Retrieval with a C-Band Radar in Minas Gerais, Brazil

With the implementation of polarimetric methods rainfall estimates improved significantly, thus providing a better radar contribution, e.g., to flood forecasting. In 2011 a dual polarization C-band radar was deployed in a tropical environment at Morro do Elefante (19º 56.7'S, 44º26.1' W) in the state of Minas Gerais , as an operational support to hydropower generation activities within its coverage area. In a previous paper (Calheiros et al, 2014) rainfall estimates with this radar were derived, based on an available limited data set , through both a R(Z) and a R(Z,ZDR) approach. CDF curves then obtained have shown a clear stratification both in range and daily interval; particular the latter has proved important in outflow prediction as verified in a study carried out in central São Paulo with the Bauru radar (Calheiros and Gomes, 2012). The work in the present paper is carried out in continuation of the study developed in the before initially paper. Radar data are from the three-month period December-January-February (DJF) 2013/2014 corresponding to the peak rainfall in the radar coverage area. Gage data are from a set of 22 stations spread under the radar umbrella, from a eight-year period. Homogeneity of rainfall within radar range is verified as well as the representativeness of the period of radar data utilized as referred to the average rainfall for the series of seven DJF periods, and the required adjustments in the derivation of the radar rainfall relations are applied accordingly. R(Z) and R(Z,ZDR) are then derived based on a probability distribution matching approach. For R(Z) two methodologies are used, the first is based on obtaining pairs of R and Z from the same probability levels in the respective CDFs (cumulative distribution function) (Calheiros and Zawadzki,1987) and the second relies on the minimization of the mean square error between the CDFs from radar and gage by iteratively varying the parameters of the R(Z) relationship (Gorgucci et al,1995). For R(Z,ZDR) the latter methodology applies and is utilized. Data from the gage network are interpolated in using the IDW technique, and impact of both time and area integration on R(Z) and R(Z,ZDR) is verified. CDFs are stratified by daily intervals as derived from the hourly rainfall evolution along the day derived from both gage and radar measurements. Radar mean area rainfall from both relations is computed over the Alto Rios das Velhas basin, a drainage area of approximately 542 km2 within radar range, and compared to the corresponding gage mean area rainfall. Next in this work, R(KDP) and R(Z,KDP) will be derived, based on recent work carried out in a tropical area based on polarimetric C-band measurements (Crisologo et al, 2014). A seasonal stratification of the relations will also be explored.