Meteorological Modeling for Air Quality Applications: A Brief History and Future Prospects

Meteorological fields are required inputs for air-quality models, along with estimates of emissions and concentrations of atmospheric chemical species. Each of these inputs, however, can contain significant errors that contribute to inaccurate model simulations of trace gases and particulates. The challenges and historic development of meteorological techniques used to support air-quality applications are traced from the 1950s to the present, with a brief look at future directions.

Meteorological states can be diagnosed from observations or simulated by dynamical models (with or without four-dimensional data assimilation, FDDA). Scales of interest range from the plume scale (~100 m to 10 km) to regional, synoptic and global scales, and from seconds to seasons or years. In general, diagnostic models are straightforward to operate and are most suitable perhaps for the plume scale. But obtaining sufficient observations to analyze regional-scale features is costly and may omit key information such as vertical velocity and divergence. Moreover, the data often lack sufficient spatial or temporal density to resolve important mesoscale processes. Unfortunately, dynamical models of the 1960s and 1970s(and the early air-quality models they supported) were too coarse and lacked sufficient accuracy for the emerging needs of regulatory policy makers. By the mid-1980s finer grids and practical FDDA techniques had improved dynamical model accuracy to the point that they became practical for use in complex regional air-quality cases and they soon became accepted tools for regulatory applications. Introduction of high-performance computing in the 1990s soon allowed extended simulations for episodes lasting days or weeks, even with resolutions of 1-5 km or less. In cases requiring regional-scale simulations lasting seasons or longer, a series of these shorter episodes is generally used. In recent years dynamic weather prediction models have been crucial to the development of routine operational air-quality forecast systems, such as the CMAQ-Eta system supported by the ongoing NOAA and EPA partnership.

Future developments in meteorological modeling for air-quality applications will include the use of ensemble modeling techniques to quantify the uncertainty in air-quality states due to errors in the meteorological inputs. Characterization of model uncertainty and defining the likely range of forecast solutions already are becoming critical factors in the prediction of plume transport and diffusion needed for homeland security. For all types of air-quality applications, advancements will be needed in model physics for fine-scale grids, better data assimilation techniques, and full (“on line”) coupling between meteorological and chemical models. These ongoing developments, including the introduction of the new Weather Research and Forecast (WRF) system for meteorological predictions, will gradually allow air quality forecasts to be issued for ozone and particulates on national domains out to 3-5 days in advance.