Operational polarimetric variables calibration at Météo France: where do we stand?

Calibration of reflectivity (Z_h) and differential reflectivity (Z_dr) is key to obtain good quality radar products such as hydrometeor classification and rainfall rate estimation. A significant effort has been dedicated lately at Météo France to developing, testing and implementing approaches to monitor and calibrate operationally the reflectivity and differential reflectivity of the 11 polarimetric radars (9 C-band, 1-S band and 1 X-band out of a total of 25) currently available.

Bias in the differential reflectivity (Z_dr) may be introduced by 3 different elements: 1) differences in the transmitted power between the vertical and the horizontal channels, 2) differences in the receiver gain of each channel and 3) impact of the radome structure or near-range obstacles. This last element introduces and spatial variability in the Z_dr bias and consequently azimuth- and elevation-dependent Z_dr bias curves have to be generated. Such curves are obtained by observing the daily Z_dr average of rainfall with a reflectivity between 20 and 22 dBZ, which, if the radar were well calibrated, should be 0.2 dB. This technique, along with the one using the Z_dr sun signature and the one exploiting data collected at 90° (a tilt that is routinely revisited every 15'), have been operational for more than one year in the radar network and long-term statistics are now available which provide a robust overview of the stability and precision that can reasonably be achieved currently with the French polarimetric radars. In a word, with the current systems and calibration procedures, the Z_dr stability is about +/- 0.3 dB, close to, but still below, the +/- 0.2 dB requirement.

Absolute calibration of horizontal reflectivity (Z_h) is a much more complex issue. The technique currently under development is based on the self-consistency method described in Gourley et al. (2009). This method is based on the fact that rainfall is relatively well represented by two parameters and polarimetric radars provide more than two variables. Z_h, Z_dr and specific differential phase K_dp can thus be considered as redundant in rain. Knowing the value of two of them allows estimating the 3rd one. If Z_dr is well calibrated (see previous paragraph), then its combination with the measured Z_h leads to a K_dp, which after range integration, provides a theoretical value of the differential phase Φ_dp that can be compared with the observed Φ_dp. In case of mismatch, then all the steps are repeated after adding a bias on Z_h. The final estimated bias on Z_h is the one that leads to the best agreement with theoretical and measured Φ_dp.

In practice, there are several factors that affect the performance of such technique and therefore robust constraints on the data used must be applied. In particular, the self-consistency only holds for rainfall and therefore other types of echoes (such as hail or ground clutter) must be eliminated from the dataset, the measurements may be affected by attenuation (even more so at higher frequency bands) and there is an intrinsic variability due to noise. An extensive sensitivity study has been carried out to determine the optimal constraints that should be imposed on the data for its operational implementation. The method is currently being validated by 1) looking at the results in known partially blocked azimuth (assuming that Z_dr is not impacted by partial beam blocking), 2) looking at the results in known wet-radome attenuation cases, 3) looking at cases of known changes in the radar receiving chain, 4) using the temporal series of reflectivity bias measurements with sun power measurements (which provide the bias introduced by the receiver). The impact of the miscalibration correction on the rainfall rate estimation is also being analysed.