Can operational polarization radars provide improved rainfall rates with error estimates?

Polarimetric radars are being introduced into operational networks in Europe and the USA but because of dwell time considerations the random noise on differential reflectivity, ZDR, can be as high as 0.5dB. This means that for light to moderate rainfalls where differential phase changes will be too small to measure, the ZDR will be very noisy and can even take on unphysical negative values, so that empirical relationships giving R as f(Z,ZDR) will fail. We explore a new technique. Rather than use the data at each gate, we analyse the statistical variation of Z and ZDR over a region of about 5km square, and hence derive the value of Nw, the normalized drop concentration, and its error over this region. The value of Nw over the region fixes the 'a' (and its error ) which should be used over that region in the empirical expression Z=aR**1.5 so that the rainfall rate and its error can be derived. In this talk we will evaluate this technique. 1) Firstly, we will compare the values of Nw obtained simultaneously by three independent methods: a) the integrated Z/ZDR technique described above, b) From the Doppler spectrum obtained by a vertically pointing radar c) From a disdrometer. Remarkable consistency is obtained, indicating that Nw can be derived from Z and ZDR to better than 3dB, so that the rainfall rate, R, should be accurate to within 1dB. 2) Secondly, we will describe comparisons of radar observations of Z and ZDR just 150m above a ground based disdrometer to justify our choice of drop shape model. 3) Thirdly, we will present results on the variability of Nw in different types of rainfall inferred from an operational radar, report on the derived rainfall rates and their errors, compare them with rapid response rain gauges, and discuss how the technique may be implemented operationally.