Prediction of Convective hazards across SpatioTemporal Scale

Severe thunderstorms cause over one billion of Euros worth in damage and dozens of fatalities across Europe each year. While their impact is large, thunderstorms and their hazards, such as large hail and damaging wind gusts, are very difficult to forecast. The fact that thunderstorms are small and relatively short-lived phenomena, poses a challenge to weather prediction models. PreCaST studies the predictability of convective hazards on the time scales of hours (short-range), out to ten days ahead (medium-range) using state-of-the-art weather prediction models. At medium-ranges, models are run with a resolution that is not high enough that severe convective storms themselves are simulated. Instead, the risk of severe weather is assessed using the simulated large-scale conditions that are required for thunderstorm development, such as an unstable atmosphere or a strong vertical wind shear. The European Severe Storms Laboratory has developed a statistical method called AR-CHaMo that can do this with climate model data. In PreCaST, ESSL will adapt AR-CHaMo to be used with the weather prediction system of the European Centre for Medium Range Weather Forecasting (ECMWF). ECMWF itself has developed an Extreme Forecast Index, which tells forecasters how extreme a particular weather situation is compared to what is normal for a given location and time of year. In PreCaST, both approaches will be validated using actual observations of severe weather. Subsequently, AR-CHaMo will be improved by investigating which additional conditions should be incorporated into the model. In order to find this out, PreCaST calls on the help of forecasters of European weather services who participate in the annual ESSL Testbed programme. For short-range forecasts of severe weather, the high-resolution convection-allowing numerical modelling system of the Zentralanstalt für Meteorologie und Geodynamik (ZAMG) will be used in PreCaST, called C-LAEF (Convection Permitting-Limited Area Ensemble Forecasting System). While the resolution of this system is fine enough to simulate some of the processes occurring in thunderstorms, others are approximated. An example of such microphysical processes are those that predict the phase of precipitation particles (i.e., ice or water) and their sizes, which have a profound influence on hail and wind gusts. By performing multiple simulations, called an ensemble, in which slight variations are made to the details of these approximations, the forecasts will reflect a range of possible outcomes. This forms the basis of a probabilistic short-range forecast of hail and severe wind occurrence that will be pioneered in PreCaST. In a last step, the predictions on the medium and short-range that will have been developed will be combined, or smoothly blended, on the basis of a verification of how well they perform for a range of lead times. The result will be a forecast system that does not yield abrupt changes to the forecast as a potential severe weather episode approaches, but instead smoothly transitions from one system to another. The blended system will be evaluated by forecasters from all over Europe who will investigate how well the PreCaST system fares.

PreCAST aimed to improve forecasts of severe thunderstorms hours to days in advance. We concentrated on forecasting large hail with a diameter of 2 cm or more, which can cause injuries to people caught outside and lead to costly damage. Optimizing predictions hours in advance requires different tools for predictions several days ahead, called the medium range. To enable medium-range forecasts, the European Severe Storms Laboratory adapted a statistical model, initially developed for climate research, to the medium-range predictions of the European Centre for Medium-range Weather Forecasts, ECMWF. The model uses a set of parameters that best represent atmospheric conditions resulting in severe hailstorms. The model gives a probability that lightning and large hail will occur as a function of the values these parameters take. Within PreCAST, the statistical model was improved and applied to the ECMWF ensemble weather model system, which, on a twice-daily basis, consists of 51 forecasts representing different pathways the atmosphere can take. We have learnt that the improved large hail model is skilful at least 7 days in advance, and occasionally longer. A variant of the model that uses publicly available forecasts as a basis can be found at the website www.stormforecast.eu, which includes an archive of forecasts going back to July 2022. To improve forecasts on the shorter timescale of hours to two days ahead, GeoSphere Austria used their high-resolution weather forecasting model C-LAEF. C-LAEF is operated on a 2.5 km grid and can explicitly simulate the initiation and development of thunderstorms. Still, the exact location and timing of thunderstorms are unknown. Thus, C-LAEF is also run in ensemble mode: each prediction consists of 17 members with slightly modified settings. Each of the members represents a possible outcome from which probabilities can be calculated. Within PreCAST, the C-LAEF system has been extended with new storm diagnostics, i.e., parameters that represent thunderstorm properties such as hail, lightning, rotation, and others, giving users more direct information on the type of storm and risk of severe weather. Furthermore, an improved microphysics parametrization scheme was introduced, which is the part of the model that predicts how ice and water particles in clouds, such as raindrops, hail, and graupel, interact in the simulated storms. These improvements resulted in better probabilistic information on the occurrence of large hail and wind gusts. The final step of the project aimed to merge both the medium- and short-range forecasting methods to create seamless, optimized forecasts from a few hours to several days before a potential severe weather event. Although this integration was not fully completed by the end of the project, verification of past severe weather cases indicated that the two approaches effectively complement each other.