Challenges facing the operational quantitative precipitation forecaster when predicting heavy to extreme events (formerly paper P1.35)

There are a number of challenges facing of quantitative precipitation (QP) forecaster. One important challenge is that a forecaster needs to know the performance characteristics of constantly changing numerical models. A forecaster also needs a basic understanding of how simplifications in model physics may negatively impact a models precipitation forecasts. Another challenge is understanding the physical processes that determine the amount of precipitation that will fall over a given point and applying this knowledge to the forecast problem. This requires the forecaster to assimilate a large array of different types of data. A third challenge is understanding the needs of the users of quantitative precipitation forecasts (QPF) and tailoring the forecasts to those needs. Still another challenge is how best to verify and calibrate the precipitaton forecasts.

The Hydrometeorological Prediction Center (HPC) has been tasked by the NWS Office of Meteorology to prepare 6 hourly precipitation forecasts from 00-84 hours. These forecasts are meant to represent mean areal precipitation which are then converted to individual basin averages for input into NWS hydrological river forecast models. The timing and intensity of rainfall is critical to correctly predicting the stream flow and crest of the river that results from any precipitation event. HPC also issues excessive rainfall outlooks when forecasters think rainfall rates will exceed the flash flood guidance values from NWS River Forecast Centers. These outlook areas are primarily for local NWS forecast offices to help them determine when there is a reasonably high chance of flash flooding.

One of the toughest challenges is trying to forecast extreme rainfall events, the ones that typically produce most of the floods and flash floods. Unfortunately, the physical processes that lead to extreme rainfall rates and accumulations act on the meso, cloud and sub-cloud scales which are not well resolved by current operational numerical models. Often, the best the models can do is identify the synoptic and meso-á scale environment conducive to extreme rainfall events. Because the scale of heavy to extreme rainfall events is so small, correctly predicting the timing and distribution of the rainfall over a river basin or sub-basin is very difficult. The accuracy of QPF falls of rapidly with time, especially for higher thresholds. It can be argued that 6 hourly forecasts for precipitation thresholds of greater than 1.00" (25.4 mm)have little skill or accuracy beyond the first 24 hours. Yet forecasters have been tasked with trying to forecast qauntitative precipitation at time ranges that are unrealistic when forecast in a deterministic fashion. The small scale of extreme rainfall events therefore argues for a more probabilistic approach to the problem. How to best quantify the probabilities is still unresolved, especially for higher precipitation thresholds. The small scale of these events suggests that high resolution models that explicitly predict convection are needed. Therefore, any ensemble approach will need incorporate models run at sufficient horizontal and vertical grid spacing to explicitly predict convection.