Fine tuning of radar rainfall estimates based on bias and standard deviations adjustments

Abstract

This paper assesses the accuracy with which single radar 1.5 km CAPPI imagery and 2 by 2 km resolution can be used to estimate precipitation in the 9-10 June 2000 Montserrat flash flood episode in Catalonia, Spain (Llasat et al., 2002). Results using Z=A·R (Marshal and Palmer Z-R relationship, 1948) with coefficients for stratiform (A=200, B=1.6) and convective (A=800, B=1.6) rain are compared with those obtained using the Histogram Matching Technique (HMT) as perform by Crosson et al. (1996) and compared, also, with a Direct Calibration Method (DCM). 5 min. rainfall series of 126 automatic raingauges of the Agency Catalana of Water (ACA), well distributed over the damaged area by the flood, were used during the calibration process and to generate rain accumulations for the verification phase. Radar precipitation amounts are clearly underestimated for this flood case when using standard Marshal and Palmer Z-R relationships and coefficients. Short time calibration performed using the HMT or by the DCM can solve this problem. In addition, post-calibration fine tuning alternatives were explored for both, HMT and DCM, based on bias and standard deviations adjustments by keeping correlations and squared errors almost unchanged.