Combined Use of Radar Volume Scans and High-Resolution NWP Model Output for Better Quantitative Precipitation Estimations over Plain and Mountainous Areas

Quantitative Precipitation Estimate (QPE) by radar is complementary to the one mesured by the rain gauge network. The combination of 1h radar QPEs and rain gauge accumulations provides what is considered for now as the best QPE product (ANTILOPE) available at Météo France.

However, there are still limitations which lead to unreliable rainfall estimates, especially over mountainous areas. Indeed, beam radar elevation and blocking create blind zones, in particular far away from the radar or in complex terrain.

In the current French operational QPE algorithm, this issue is mitigated by correcting the measured rainfall rates based on the Vertical Profile of Reflectivity (VPR). This VPR correction relies on an idealized VPR established from four parameters: the freezing level and the decreasing rate of reflectivity above, the bright band thickness and its amplitude (Tabary, 2007). An important limitation of this method is that only one VPR is defined for the whole radar domain, ignoring the spatial variability of precipitation.

The new approach studied to improve the VPR correction takes advantage of the performances of the French high-resolution NWP model AROME (Brousseau et al., 2016). Previous studies have shown that synthetic, yet realistic radar observations can be obtained by applying a radar simulator to model outputs (e.g., Caumont et al. 2006). In our new VPR correction algorithm, polarimetric data (Zhh, Zdr, Kdp) are first simulated by the polarimetric radar forward operator of Augros et al, 2016). Then, through a Bayesian method (Kummerow et al., 1996, 2001; Caumont et al. 2010), the closest simulated VPR profile to the apparent one (observation) in terms of reflectivity is searched in the vicinity of the radar pixel considered. The determined simulated VPR is finally applied to estimate ground rain rate.

To evaluate the relevance of this new approach, operational radar and closest simulated polarimetric profiles from AROME ensemble and deterministic models are thoroughly compared during several precipitating events (rain or snow, convective and stratiform). First results show a good consistency between both datasets. The results presented at the conference will show to which extent high-resolution numerical weather forecasts may prove useful for improving radar-based QPEs.