Ozone, a secondary pollutant, is a strong oxidant that can pose a risk to human health. It is formed from a complex set of photochemical reactions involving nitrogen oxides (NO_x) and volatile organic compounds (VOCs). Ambient measurements and air quality modeling of ozone and its precursors are important tools for support of regulatory decisions, and analyzing atmospheric chemical and physical processes. I worked on three methods to improve our understanding of photochemical ozone production in the Eastern U.S.: a new detector for NO₂, a numerical experiment to test the sensitivity to the timing to emissions, and comparison of modeled and observed vertical profiles of CO and ozone. A small, commercially available cavity ring-down spectroscopy (CRDS) NO₂ detector suitable for surface and aircraft

monitoring was modified and characterized. The CRDS detector was run in parallel to an ozone chemiluminescence device with photolytic conversion of NO₂ to NO. The two instruments measured ambient air in suburban Maryland. A linear least- squares fit to a direct comparison of the data resulted in a slope of 0.960±0.002 and R of 0.995, showing agreement between two measurement techniques within experimental uncertainty. The sensitivity of the Community Multiscale Air Quality (CMAQ) model to the temporal variation of four emissions sectors was investigated to understand the effect of emissions' daily variability on modeled ozone. Decreasing the variability of mobile source emissions changed the 8-hour maximum ozone concentration by ±7 parts per billion by volume (ppbv). Increasing the variability of point source emissions affected ozone concentrations by ±6 ppbv, but only in areas close to the source. CO is an ideal tracer for analyzing pollutant transport in AQMs because the atmospheric lifetime is longer than the timescale of bound- ary layer mixing. CO can be used as a tracer if model performance of CO is well understood. An evaluation of CO model performance in CMAQ was carried out using aircraft observations taken for the Regional Atmospheric Measurement, Mod- eling and Prediction Program (RAMMPP) in the summer of 2002. Comparison of modeled and observed CO total columns were generally in agreement within 5-10%. There is little evidence that the CO emissions inventory is grossly overestimated. CMAQ predicts the same vertical profile shape for all of the observations, i.e. CO is well mixed throughout the boundary layer. However, the majority of observations have poorly mixed air below 500 m, and well mixed air above. CMAQ appears to be transporting CO away from the surface more quickly than what is observed.

Turbulent mixing in the model is represented with K-theory. A minimum K_z that scales with fractional urban land use is imposed in order to account for subgrid scale obstacles in urban areas and the urban heat island effect. Micrometeorological observations suggest that the minimum K_z is somewhat high. A sensitivity case where the minimum K_z was reduced from 0.5 m2/s to 0.1 m2/s was carried out. Model performance of surface ozone observations at night increased significantly. The model better captures the observed ozone minimum with slower mixing, and increases ozone concentrations in the residual layer. Model performance of CO and ozone morning vertical profiles improves, but the effect is not large enough to bring the model and measurements into agreement. Comparison of modeled CO and O₃ vertical profiles shows that turbulent mixing (as represented by eddy diffusivity) appears to be too fast, while convective mixing may be too slow.