Issues in the analysis of air quality modeling data in attainment demonstrations of the 8-hr ozone standard

Attainment demonstrations of the ozone standard serve as the basis for regional air quality planning, but do the recommended analytical techniques establish control strategies that will be effective? The relative interpretation of model data was introduced by the Environmental Protection Agency (EPA) to compensate for poor model performance, and use of 8-hr averages was introduced to mirror the standard. The sensitivity of the relative reduction factor (RRF) is tested against model bias, using a Houston, Texas regulatory model. There is an inverse linear correlation with model bias at each site (R2 = 0.47), changing future design values by 0-5 ppb. Ninety percent of cell selections for the RRF calculation in these episodes are 14-19 km from the monitor. The RRF approach is recommended only when modeled ozone response is linear. Model error statistics using 8-hr concentration averages overestimate performance compared to those using 1-hr values (11.7% versus 14.8%, median).