Thursday, 15 January 2004
Opportunities for human forecasters to improve upon model forecasts now and in the future
It has been quantitatively shown (for example, in HPC verification of QPF and surface low tracks) that human forecasters can consistently outperform NWP models in the present forecaster/ model mix. How will the current role of the human forecaster in the NWS evolve, now that gridded forecast products are being issued for boxes 5 km or 2.5 km on a side? The authors have been developing and providing training on NWP models for four years to help forecasters both maximize model usefulness and minimize the impact of model flaws in the forecast process. That experience has reinforced our view that forecasters can intelligently assess and improve upon NWP model forecasts by
· Understanding the capabilities and limitations of the NWP model components, such as the physics packages, the analysis method, and the data assimilation system,
· Knowing the behavior of NWP models in previous scenarios similar to the current forecast situation, and
· Paying attention to the observations.
This provides a basis for determining which parts of model solution are most likely to be correct, and qualitatively adjusting NWP models for some types of error. This adjusted deterministic forecast addresses biases and specific errors that also will appear in ensemble members, and thus it compliments, rather than duplicates, forecasts of uncertainty provided by ensemble prediction systems.
We will present case examples showing opportunities to correct for various aspects of NWP models, including their synoptic and mesoscale analyses, specific model artifacts and peculiarities, convective parameterizations, and downscaling issues. While improvements in future models will eliminate some of the specific problems we will show, many of these types of problems will remain and will be a difficult challenge for modelers to address. We believe that even as model skill continues to improve, humans will still have opportunities to make major operational forecast corrections for specific events, some of which will have significant societal impact.