Air quality forecasts are typically initialized using chemical species concentrations obtained from the previous day's forecasts with no regard to the observations. This approach generally leads to better forecasts compared to those initialized with chemical species concentrations obtained from climatology. Data assimilation is an essential part of weather forecasting in all major meteorological centers. However, few attempts have been made to assimilate chemical species for air quality forecasting. Chemical data assimilation in air quality modeling is both a result of problem complexity (the number of chemical species varies in the model from tens to hundreds and is the multiple of the number of atmospheric state variables) and the scarcity of observations (especially with respect to vertical profiles). This is likely to change in the near future with the proliferation of satellites and unmanned observing platforms. The presentation will concentrate on the development of background error covariances for fine aerosols, the implementation of these species in the Grid Statistical Interpolation (GSI) and evaluation of forecasts with assimilation of surface measurements of these constituents. Preliminary results show forecast improvement when data assimilation is used.
This publication was presented at the following: