Operational and phenomenological evaluation of a 2001 continental United States 36/12km MM5 application

In order to solve for the change in pollutant concentrations over time and space, air quality models require certain meteorological inputs that, in part, determine the eventual formation, transport, and destruction of pollutant material. Over the past decade, the usual procedure for acquiring the needed meteorological data has been to apply a meteorological model over a relevant historical time period. When executed retrospectively, these models are able to continually blend ambient data with model predictions via data assimilation, thereby yielding temporally and spatially complete data sets that are somewhat grounded by actual observations. However, since it will never be possible to characterize the actual meteorological fields at all points in space and time with complete accuracy, a successful meteorological evaluation should have two objectives:

1) to identify and quantify the existing biases and errors in the meteorological data in order to allow for a downstream assessment of how the air quality modeling results will be affected by any meteorological data issues, and

2) assess whether the meteorological modeling output fields represent a reasonable approximation of the key meteorological events that most greatly influence air quality over a given area.

EPA has recently completed development of a large meteorological data set for the entire year of 2001 with the Pennsylvania State University / National Center for Atmospheric Research mesoscale model (MM5). This simulation was conducted at a resolution of 36km over the entire continental United States, but also included a 12km grid over the eastern 75% of the country. This meteorological data set is expected to serve as the backbone for several upcoming EPA regulatory analyses and, as such, has been evaluated as thoroughly as possible. This paper will summarize the findings of the operational and phenomenological model performance evaluation of this MM5 data set. From a statistical standpoint we will summarize biases and errors in temperature, water vapor mixing ratios, wind speed (aloft/surface), wind direction (aloft/surface), clouds/radiation, precipitation, and the planetary boundary layer, (height, initiation, rate of rise, duration). Multiple spatial and temporal breakdowns of the data will be completed to assess whether certain periods are prone toward inaccurate model predictions. From an event-based perspective, we will summarize how well the model data approximates those specific meteorological phenomena that are thought to strongly affect regional air pollution. At a minimum, features such as sea breezes, low-level jets, and the amount of convective precipitation will be considered. The event-oriented portion of the evaluation will summarize model performance in terms of probability of detection and false alarm rate.