The viability of improving ozone predictions of air quality models by incorporating satellite data to specify initial and boundary conditions is investigated, with focus on the U.S./Mexico border cities. This border is rapidly growing, devoid of a comprehensive surface monitoring network, has experienced ozone-related health problems and in need of a predictions-based health warning system. The Models-3 (MM5-SMOKE-CMAQ) air quality modeling system, with best available emission inventories were used. Surface meteorology and ozone/NO2 data from the EPA-AIRS network and four satellite instruments (SCIAMACHY and OMI for tropospheric NO2 vertical column densities and TES and AIRS for tropospheric ozone profiles) were employed for model evaluation. Simulations were conducted with default boundary/initial conditions as well those derived using TES. MM5 showed excellent performance for temperature and moderate performance for wind speed. Models-3 showed moderate performance for ground level ozone and captured its diurnal variation well, but ground level NO2 predictions were less satisfactory. The modeled NO2 vertical column densities, however, correlated well with SCIAMACHY measurements and to a lesser extent with OMI. Modeled 3D ozone concentration correlated well with AIRS measurements in the middle troposphere.

Although model predictions with TES data ingestion for initial/boundary conditions did not improve surface ozone predictions significantly, improvements could be noted for upper-tropospheric ozone (above 500 mbar) when evaluated against satellite data such as TES samples from the following day and 3D grids of the middle and upper troposphere from AIRS. The results point to the dominance of local ozone chemistry at lower levels vis-à-vis advection and diffusion across boundaries prevalent at higher levels.