Although
meteorological data assimilation has been successfully applied to
improve forecasts the community
is in the early stages of developing a system that couples the
assimilation of weather elements
(wind, pressure, temperature and water vapor) with chemical species
(e.g., nitrogen oxides, volatile
organic compounds, ozone and PM). The
strategy includes the testing of the viability of both dynamic/deterministic
schemes (e.g., 3DVAR) and stochastic/dynamic schemes (e.g., ensemble
Kalman Filter).
These approaches will rely on available National Centers for
Environmental Prediction (NCEP) prediction
models presently available (or soon to be available) such as the
Weather Research and Forecast
Model (WRF)/CMAQ (Community Multi-scale Air Quality model). We will use
the 3DVAR scheme
developed at NCEP, the so-called Grid-Point Statistical Interpolation
system (GSI). In essence, the
assimilation scheme(s) will rely on a prediction model that is faithful
to the evolution of the meteorological/chemical
state of the atmosphere. With output from this model serving as
background, the observations
are combined with this output to produce the estimate of the system
state. The estimate is found
optimally, i.e., the error in the estimate is minimized by properly
accounting for the relative goodness
of forecast compared to observation – statistics contained in the error
covariance matrix (error variances
and covariances associated with the meteorology and chemistry). The
construction of this matrix
for the combined problem is a major component of the proposed research. In parallel with the work
on the operational system, we will perform data assimilation
experiments that take advantage of data
collected in field programs. We are especially interested in comparing
air quality forecasts in geographically
different regions – regions that exhibit differences in terrain and
ocean-atmosphere-land interactions.