Extended Abstract (496K)P1.3
Statistical analysis of numerical model output used for forecasting during Joint Urban 2003
Peter K. Hall Jr., Oklahoma Climatological Society and University of Oklahoma, Norman, OK; and J. B. Basara
During July of 2003, Oklahoma City was the location of the largest urban dispersion study conducted in the United States. The overall success of Joint Urban 2003 (JU2003) depended heavily on the weather, particularly wind speed and direction. To aid the scientists in identifying when favorable conditions for tracer releases were present, a forecasting team provided daily weather briefings. The ultimate challenge for the forecasters was to predict specific wind speed and direction values twenty-four to thirty-six hours in advance. Numerous tools were used by the forecast team (e.g. climatology, numerical forecast models, experience, and intuition) to aid in compiling forecasts for the experiment. However, due to the length of the forecasts considerable attention was placed on the output from numerical forecast models.
Leading up to the July experiment, a preliminary assessment was made to understand initial forecast model error. During and after the completion of the experiment, analyses of model output statistics (MOS) for Oklahoma City were conducted to document the degree of error in wind forecasts. Results of MOS and tabular data wind forecasts for Will Rogers World Airport (KOKC), compared to surface station verification located at KOKC, showed as much as forty-five degrees error in wind direction. Conversely, the error in wind speed was approximately three knots. Both errors were in the forecast range of twenty-four to thirty-six hours. It was found that because no one model out-performed another during JU2003, and significant errors were common in the model output related to the surface winds, forecasters were faced with many dilemmas when trying to predict wind speed and direction in such a narrow band during JU2003. The results demonstrate that techniques used during JU2003 have applications for further experiments and public use, and further exposes the limits of using current regional models for urban forecasting.
Poster Session 1, Urban Zone Posters (Hall 4AB)
Monday, 12 January 2004, 2:30 PM-4:00 PM, Hall 4AB
Previous paper Next paper