The 3 May 2013 tornadic supercell in the Po River Valley, Italy: a case study for evaluating WRF-ARW at convection-allowing scale in the 24h forecasting range

The Po River Valley in northeastern Italy is known for being one of the most active regions in Europe for thunderstorm development (e.g., Cacciamani et al., 1995), with some storms occasionally becoming severe (Alberoni et al., 2000). With more powerful computing systems becoming available in recent years an opportunity exists to evaluate the performance of operational NWP for that region with convection-allowing models. Because the Po River Valley is surrounded by complex terrain capable of producing important mesoscale circulations, numerical solutions for that region may display strong sensitivities to the horizontal resolution (and physical parameterizations) with potential impact on aspects that are particularly relevant to deep moist convection, such as convective initiation, for instance.

The Center of Excellence for the Integration of Remote Sensing and Numerical Modeling for the Forecast of Severe Weather Phenomena (CETEMPS) from the University of LxAquila, Italy, runs an operational suite for the Advanced Research WRF (WRF-ARW) with 12km and 3km horizontal grid-spacing. The geographic domains cover, respectively, south-central Europe and the Italian Peninsula.

The objective of this study is twofold: first, to describe briefly the atmospheric environment that led to the formation of a right-moving supercell storm in the Po River Valley that produced large hail and one (well-documented) significant tornado in the afternoon of 3 May 2013. Data sources include scans from San Pietro Capofiume weather radar, visible and infrared geostationary satellite imagery, surface observations, and soundings from northern Italy. Second, the CETEMPS WRF-ARW outputs at 12km and 3km are analyzed to assess the model performance in representing the mesoscale pre-convective environment (at 12km) and the convective initiation, mode and evolution (at 3km) for an operational run that started approximately 24h prior the tornadic storm. Radar data available for this event are used to subjectively evaluate the model solution at 3km grid spacing, while for the 12km run satellite imagery, soundings and surface data are employed to evaluate the model forecast. Emphasis is given in assessing whether the model solution at 3km grid spacing showed any skill in indicating the correct timing, position and/or convective mode for this high impact event.

Model sensitivity to the choice of the PBL scheme is also investigated by re-running the 3km WRF-ARW simulation with the Bougeault-Lacarrere scheme (BouLac; Bougeault and Lacarrere, 1989) rather than the Mellor-Yamada-Janjic scheme (MYJ; Janjic, 1994) which is the option used in the operational runs.

Overall, the main findings are that the simulations at 3km grid spacing did produce strong discrete convective cells over the Po Valley during the afternoon of 3 May that underwent storm-splitting process with the right-movers becoming the dominant ones. Thus, the convective mode was correctly predicted, which is an encouraging result given that the simulation started 24h before the convective event. Regarding sensitivity to the PBL scheme, the simulation with the BouLac scheme was clearly superior in aspects other than the convective mode: it displayed better location and timing of the convective initiation for the discrete cell that would eventually evolve into the tornadic supercell; and better representation of the storm structure, including a hook-echo-like appendage and a well defined mesocyclone, in fairly good agreement with the observations. In the BouLac simulation the supercell was shorter lived and slightly displaced to the north as compared to the observed one, though.

The early stages of convective initiation in the simulation with MYJ scheme were good in both timing and location, but this first convective activity failed to evolve into a supercell and dissipated quickly. Later in the simulation new storms that formed farther north and west became long-lived and eventually underwent splitting and displayed mid-level rotation. However, such evolution only occurred with significant delay compared to observations. In addition, cells generated in the MYJ simulation were placed much farther north (and, thus, more displaced away from the real cells) than in the BouLac solution. The reason why the first cells in the MYJ simulation were short-lived is still being analyzed; this is an important aspect because had these cells developed into supercells the solution would have been better than the one produced with the BouLac PBL scheme in terms of timing, and possibly better also in terms of location.

In parallel to the ongoing analysis, additional simulations at 1km grid spacing are also being carried out and results will be shown at the Conference.

Alberoni, P. P., V. Levizzani, R. J. Watson, A. R. Holt, S. Costa, P. Mezzasalma, and S. Nanni, 2000: the 18 June 1997 companion supercells: multiparametric Doppler radar analysis. Met. Atmos. Phys., 75, 101-120.

Bougeault, P., and P. Lacarrère, 1989: Parameterization of orography-induced turbulence in a mesobeta-scale model. Mon. Wea. Rev., 117, 1872-1890.

Cacciamani, C., F. Battaglia, P. Patruno, L. Pomi, A. Selvini, and S. Tibaldi, 1995: a climatological study of thunderstorm activity in the Po Valley. Theoret. Appl. Climatology, 50, 185-203.

Janjic, Z. I., 1994: The step-mountain eta coordinate model: further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927-945.