WRFDA-4DVAR radar data assimilation for operational very short-range precipitation forecasts in Catalonia: Initialisation strategies and preliminary results

Nowcasting is a crucial tool in every meteorological office: it provides steadily updated forecasting guidance and is essential in cases of extreme weather when forecasters need to know the evolution of the situation in great detail and issue early warnings if necessary. 

It is well known that forecasting products based on radar echo extrapolation have the best skill for the first one or two hours. Beyond this range, numerical weather prediction (NWP) models are needed to obtain reliable precipitation forecasts. The Meteorological Service of Catalonia (SMC), like many other operational centres around the globe, blends extrapolation techniques with NWP to produce seamless forecasts (Casellas et al., 2022). 

The NWP component of the nowcasting system at the SMC is based on the WRF model. To build a time-lagged ensemble, several simulations run every 3 hours that assimilate radar and conventional data with the WRF-3DVAR and 3DEnVAR methods and the vLAPS system (Jiang et al., 2015). 

However, WRFDA also incorporates the 4DVAR technique, which takes into account the time tendency of the observations and uses an NWP model as a dynamical constraint, thus linking microphysics with the dynamical and thermodynamical fields (Sun and Wang, 2013). Several studies conclude that this initialisation method yields better results when compared to 3DVAR for precipitation forecasts (Mazzarella et al., 2021), especially at longer lead times. 

The main challenge for using 4DVAR for operational purposes is its computational cost. Fortunately, the high frequency of SMC’s composite radar data, available every 6 minutes, allows us to use small assimilation windows and, at the same time, keep a large number of observations that can be assimilated at their appropriate time. 

In this work, we introduce our strategy to obtain timely analysis so that the WRFDA-4DVAR method can be used to initialise some members in our nowcasting system in an operational framework. In addition, the performance of the forecasts compared to those initialised with WRF-3DVAR and 3DEnVAR are shown. 

References 

Casellas, E., Atencia, A., Mercader, J., Moré, J., Rigo, T., Sairouni, A., & Segalà, S. (2022): Blending of precipitation probability forecasts: weather radar advection and NWP models. 11th European Conference on Radar in Meteorology and Hydrology, Locarno (Switzerland), 29 August-2 September 2022. 

Mazzarella, V., Ferretti, R., Picciotti, E., and Marzano, F. S.: Investigating 3D and 4D variational rapid-update-cycling assimilation of weather radar reflectivity for a heavy rain event in central Italy, Nat. Hazards Earth Syst. Sci., 21, 2849–2865, https://doi.org/10.5194/nhess-21-2849-2021, 2021.    

Sun, J., and H. Wang, 2013: Radar Data Assimilation with WRF 4D-Var. Part II: Comparison with 3D-Var for a Squall Line over the U.S. Great Plains. Mon. Wea. Rev., 141, 2245–2264, https://doi.org/10.1175/MWR-D-12-00169.1.