Frederic Fabry

Short-term (0-3 hr) prediction of disrupting atmospheric phenomena is one of the rare areas of meteorology that has not seen any significant progress in the past few decades. As a result, the most publicly visible forecast busts in recent years occurred from missed warnings of severe weather. Short-term prediction is also a task where we do not yet largely rely on numerical forecasting. There are many reasons for this: severe atmospheric phenomena evolve quickly; their numerical prediction requires considerable processing as well as proper initial conditions at high resolution; these initial conditions, coming largely from remote sensing, remain currently incomplete especially on properties that drive storm evolution such as temperature, humidity, and winds.

To overcome these problems, progress must occur on many fronts. We must improve the quantity and quality of the data to be assimilated in models. Radar data assimilation is also more complex than that of other data, and radar-specific ideas should be explored. It is also possible that direct detection and non-numerical forecasting may remain the only practical prediction approach in the foreseeable future, and we need to develop new ideas on that front too.