Forecasting and Verifying in a Field Research Project: DOPLIGHT '87

Abstract

Verification of forecasts during research field experiments is discussed and exemplified using the DOPLIGHT '87 experiment. We stress the importance of forecast verification if forecasting is to be a serious component of the research. A direct comparison and contrast is done between forecasting for field research and forecasting in the operational sense, highlighting the differences between them. Ale verification of field research program forecasting is also different from that done in operations, as a result of those forecasting differences. DOPLIGHT '87 was a field project conducted jointly by the National Severe Storms Laboratory and the Oklahoma City National Weather Service Forecast Office, and is described in detail. During the experimental design, special attention was given to forecast design, to ensure that verification would be unambiguous and that the data collected would be appropriate for validating the forecasts. This a priori design of the forecasts to consider proper objective verification is, we believe, unique among research field programs. The forecast evaluation focuses on the contingency table and summary statistics derived from it, as treated in a companion paper by Flueck (1989; hereafter referred to as F1u89). Results are interpreted in terms of their implications for future field research experiments and for operational forecasting. For example, it is noted that DOPLIGHT '87 forecasts of convective potential were nearly constant from the evening before an anticipated operational day to about local noon on that day. This suggests that convective storm field research operational decisions could be made as early as the evening before an anticipated operational day with negligible loss of skill. Summary measures of the forecast verification suggest that the DOPLIGHT '87 forecasters demonstrated skill roughly comparable to the forecasters at the National Severe Storms Forecast Center in issuing outlooks of convective potential. The requirement for time to assimilate the most recent data is noted both for field experiments and for operations, and some discussion of the potential impact of new data acquisition and processing systems is offered.