Quantifying local creation and regional transport using a hierarchical space-time model of ozone as a function of observed nitrogen oxides, a latent space-time VOC process, emissions, and meteorology

We explore the ability of a space-time model to decompose the 8-hour ozone concentration on a given day at a given site into the parts attributable to local emissions and regional transport, and ultimately to assess the efficacy of past and future emission control programs. We model ozone as created plus transported ozone plus an error term that has a seasonally varying spatial covariance. The created component uses atmospheric chemistry results to express ozone created on a given day at a given site as a function of the observed NOx concentration, the latent VOC concentration, and temperature. The ozone transported to a given day at a given site we model as a weighted average of the ozone observed at all sites on the previous day, where the weights are a function of wind speed and direction. The latent VOC process model has a mean trend that includes emissions, temperature, a workday indicator, and an error term with a seasonally varying spatial covariance. Additionally, we specify two hierarchical-Bayesian analogues to this model, one of which has the same seasonally varying exponential spatial covariance matrices as the likelihood model, while the other has seasonally varying unstructured covariance matrices. We outline an MCMC fitting algorithm for each Bayesian model, and we derive the exact and approximate full conditional distributions required to implement the algorithm. Using likelihood methods, we fit the original model and obtain space-time predictions for comparison with a withheld dataset and with predictions from CMAQ, the deterministic model used by EPA to assess emission control programs. Our model produces one set of predictions based on the mean trend and spatial correlations and another based on the mean trend alone. The predictions based on the mean trend and spatial correlations have a lower root mean squared error (RMSE) when compared to point observations than do than do the 36 km gridcell averages from CMAQ; predictions based on the mean trend alone have the same RMSE as CMAQ but systematically underpredict high ozone values.