Capabilities and Expectations for Flash Flood Nowcasting

This talk will focus on the nowcasting of convective rainfall using automated systems. Nowcasting refers to time and place specific 0-6 hr forecasts that are heavily based on observations. In the U.S. most operational methods for flash flood warnings do not even involve a forecast but rely on monitoring rainfall accumulation using radar and rain gages. Particularly helpful have been time lapse displays of radar reflectivity and rainfall accumulation based on the WSR-88D. Nowcast systems that do make rainfall forecasts make heavy use of extrapolation. Some of these systems are able to initiate new storms and grow or dissipate existing storms. These systems can produce products completely automatically or can be used interactively by forecasters. Typically better results are produced when a human is part of the system.

Many flash floods have been analyzed in detail. From these studies it is apparent there is no unique type of storm that produces flash floods (Hjelmfelt 1999). Particularly dangerous are quasi-stationary mesoscale convective systems, supercells and "training echoes" where a series of cells passes over the same location. From these studies it is apparent that heavy rainfall that produces flash floods typically occurs on scales of 10's of kilometers. Factors that influence or control the evolution and motion of these storms are typically of similar scale: for example the initiation of "training cells" by a stationary outflow boundary. Numerical models initialized with synoptic sale data or even mesoscale data have almost no skill in forecasting convective scale flash flood events (Fritsch et al. 1998). Numerical methods initialized with storm scale data from radar show promise (Warner et al. 2000), however this work is in its infancy. Therefore it is anticipated that improvements in flash flood forecasting in the next 10 years still remains with the improvement of nowcasting systems.

Three of the more advanced systems for nowcasting rainfall are NIMROD (Golding 1998) and GANDOLF (Hand 1996) from Great Britain and the Auto-nowcaster from NCAR. These three systems participated in a three-month field project in Sydney Australia from 2 September to 22 November 2000. This project called Sydney 2000 was organized by the World Weather Research Program.

Features of these three systems will be described and preliminary results from the Sydney 2000 Field Demonstration Project will be provided. Also to be discussed are expected improvements in nowcasting systems during the next 10 years and how the forecaster is likely to use them.