The French operational dual-polarization processing chain

Météo France investment on operational radar polarimetry started 7 years ago with the installation, in 2004, of the first French polarimetric radar (C-band) in Trappes, near Paris. Since then, a significant amount of work has been carried out and the positive results obtained have led to the extension of dual-polarization to other radars of the network, which now counts 11 operational polarimetric elements (9 at C-band, 1 at S-band and 1 at X-band).

Given the high sensitivity of dual-polarization algorithms to biases on Z_DR, Φ_DP, ρ_HV and, to a lesser extent Z_H, the operational introduction of dual-polarization prompted the definition and production of monitoring indicators on a daily basis. Indeed, the analysis is that polarimetry does improve the quality of all radar products compared to conventional technology if and only if the system is well calibrated. If not, then the usage of polarimetry may lead to a degradation of the quality of the conventional products. The idea was then to detect as early as possible a failure in the radar system (rotary joint failure, wave guide losses, TR tube failure, …) that would cause problems on subsequent products. Several examples have been identified during the last years of dual-polarization operations at Météo France. The monitoring indicators that were designed and coded are the following (many of them are described in Gourley et al. (2006)): Mean Z_DR at 90°, Mean Z_DR for Z_H between 20 and 22 dBZ in close-range, high-SNR and rain gates, Z_DR in sun spikes on sunrise and sunset, Φ_DP offsets, upper 80% quantile of all ρ_HV values in close-range, high-SNR pixels in rain and mean Z_DR at lowest elevation at several close ranges. Alerts are automatically triggered at the end of each day if one of the monitoring indicators falls outside predefined thresholds. In that case, the radar processing chain is turned back to conventional and the problem is analyzed and fixed by radar experts.

The polarimetric processing chain performs successively calibration of the polarimetric variables (Z_H and Z_DR), non meteorological echoes identification, bright band identification, path-integrated attenuation and differential attenuation correction, Φ_DP offset removal and filtering (using a running 6 km median filter), K_DP estimation and hydrometeor classification. In its first version, the Φ_DP offset removal is done dynamically for each ray and attenuation correction (both on Z_H and Z_DR) is done using Φ_DP (with constant proportionality coefficients γ_H and γ_DP). Decision was made not to use Z_DR in this first version because the experience of more than one year of monitoring indicator production on the 11 radars has revealed that the current calibration / stability of Z_DR was close to but still below the requirement (+/- 0.2 dB) for robust and accurate quantitative exploitation. The chain provides Cartesian outputs of the Path Integrated Attenuation (PIA) of horizontal reflectivity (Z_H) and of echo types that are subsequently used in all operational products (QPE, reflectivity pseudo-CAPPI image, …).

In the Summer of 2010, the dual-polarization chain has been implemented for testing on the C and S-band polarimetric radars (10 overall). From that moment on, all operational products (incl. QPE, base reflectivity, radar products for data assimilation, ..) have been generated continuously in parallel with and without dual-polarization. This allowed assessing the benefits of dual-polarization over a long time period and also training the operational maintenance team to getting accustomed to polarimetric radars. Operational procedures have been designed to 1) switch from single-pol to dual-pol in case of a detected problem on the dual-pol variables, 2) update the Z_DR calibration curves used in the polarimetric processing chain, 3) switch back from single-pol to dual-pol after the identified problem has been fixed. The objective evaluation shows a clear improvement of all operational products. For instance, objective comparison with gauges shows that the polarimetric QPE (including polarimetric attenuation correction) performs better than the conventional one, especially at high rainfall accumulations. Interestingly, the benefits are sometimes lost after the operational rain gauge adjustment scheme is activated, which raises the question of the cohabitation between polarimetry and rain gauge adjustment schemes, which are used operationally by many of not all radar services. Clear air pixels are also much better identified and removed on the products that have been using polarimetry. The current plans are to switch gradually on all polarimetric radars from the conventional processing chain to the polarimetric processing chain between July 2011 and December 2011. A second version of the polarimetric processing chain (including a combined Z_H-K_DP estimator) is planned for 2012.