An analysis of meteorological services under extreme weather conditions based on a Bayesian decision-support model: a case study of the thunderstorms in Beijing on July 21, 2012

摘要

The decision-making procedures of the meteorological service concerning the extreme thunderstorm in Beijing on July 21, 2012, were simulated and analyzed in a scenario using a Bayesian decision-support model. A thorough analysis of the decision-making process during that terrible thunderstorm demonstrated that a decision-support model can be used to make optimal decisions regarding uncertainty problems in the meteorological service supported by current meteorological technology and data resources, e.g., the mesoscale numeric weather prediction (NWP) system and observational data. Using NWP grid data, we assessed the flooding and debris flow risks on that day, and the high risks were clearly apparent. Consulting the historical flooding records, we also recognized the high thunderstorm risk that day even though the predicted precipitation was reported as 100-200 mm in most areas. Because of the low probability of extreme precipitation indicated by climate data, the posteriori probability estimated by the Bayesian model was only 23.1 %. For the differences between expected losses in a disaster and a non-disaster state, issuing a prediction for a non-disaster state could obviously lead to greater expected losses than predicting a disaster state. Therefore, it would be advisable to provide a disaster state prediction and take a correspondingly worst case scenario outlook in the meteorological service, which was the optimal decision-making strategy at that time. This study reveals that (1) the objective promotion of an emergency response level corresponding to a severe weather warning is recommended to realize the advantages of a worst case scenario prediction, even if the forecasters underestimate the devastating impact of the weather, and thus, it can obviously relieve unnecessary pressure on forecasters, and (2) the public should be provided with uncertainty information along with severe weather forecasts and warnings so, as the end users of meteorological services, they can make better informed decisions.