Earlier in radar rainfall estimations by dual-wavelength method attenuation-rain rate (Att-R) predictor was commonly used, but it has not revealed advantages in comparison with the conventional Z-R predictor because of unjustified assumption on constant DSD was used.

At the 30 AMS Radar Conference we have presented a new version of dual-wavelength radar rainfall measurement method based on comparison of attenuation effects in radial paths: i) measured directly by the S,X-band reflectivity differences, and ii) calculated by our self-consistency model. The input model parameter is A coefficient in Z-R relation. The coefficient b is supposed fixed (b=1.6). The novelty of this method consists in rejection of the DSD constancy assumption, and use of variable A coefficient according to the rain microphysics. In this sense the method provides the attenuation-tuned Z-R relation.

Numerical simulation of the method implementation to radar rainfall measurements has revealed dependence of the method error from noise in Z(S,X-band) datasets. We apply a spatial averaging of attenuation data and fuzzy-logic multicriteria filtration of calculated A coefficients for error reduction.

For evaluation of radar rainfall estimations data from about twenty gauges within 100 km from radar site was used. Comparison with gauges was carried out on 12-hour precipitation amount as well as time-series of 10-minutes rain rates that corresponds to period of radar observations in Moscow.

In our operational practice of radar observations in various rainfall conditions many events occur with synchronous variations of 12-hour accumulation error signs for the most of gauges. In our opinion this synchronism is evidence of existence of large-scale spatial homogeneous patterns in clouds with similar DSD and Z-R relation. Therefore the application of time-variable and constant for the whole radar zone Z-R relation which depends on current mean DSD is substantiated.

Application of attenuation-tuned Z-R relation in most cases gives significant improvement of radar rainfall measurement accuracy in heavy rain events with essential attenuation in X-band. On the other hand we observe peak errors of rainfall estimations based on fixed Marshall-Palmer coefficients A=200 b=1.6 in heavy rain events as well.

In the paper we will present statistical results for application of the method for rainfall estimation in 2000 in Moscow. The retrieved difficulties and limitations of the method will also be discussed.