Chemical transport models (CTMs) are widely used in scientific studies and air quality management. A particular application is to assess how pollutant concentrations, particularly ozone and PM, respond to emission controls. As part of a dynamic evaluation, two widely used CTMs, CMAQ and CAMx, were evaluated for how well they captured trends in observed surface ozone, as well as the changes in associated concentrations, between 2001 to 2011 and 2011 to 2016. Those periods were chosen as additional effort was used to improve emissions estimates for all three years. Prior studies found that both emissions inventories and observations indicate reductions in both ozone and NO2 during these periods, but studies have found that NO2 observations have not agreed as well with estimated emissions. In general, given the efforts to harmonize model inputs, the two models performed very similarly and captured the ozone declines for both the 2001–2011 and 2011–2016 periods at the national level, though moderate biases were found in some locations. At the national scale, ozone trends for the 2011–2016 period were better captured by the models than those for the 2001–2011 period. Both models overestimated the observed decrease in ground level NO2 at the higher end of the concentration spectrum. This is linked to, and can cause, the detected tendency of the models to have larger than observed increases in ozone at the lower end over time. At the metropolitan statistical area (MSA) level, the performance of capturing the ozone trends varied. For some MSA's, the models estimated the wrong direction in the trends; e.g., in Denver, where maximum 8-hr average ozone levels increased from 2001 to 2011, the models predicted a decrease. The US EPA recommends location specific Relative Response Factors (RRFs) to adjust results when applying the models for air quality management purposes. At the regional level, the simulated ozone RRFs were up to 8% larger than the observed RRFs for the 2001–2011 period, suggesting that the models underestimated the ozone decrease. For the 2011–2016 period, with a couple of exceptions, the simulated RRFs were up to 9% smaller than observed RRFs, indicating a negative bias in the simulated trend. For individual MSAs, the ratio of the simulated to observed RRFs ranged between 0.86 (Knoxville for 2001–2011) and 1.40 (Los Angeles for 2001–2011). This is possibly linked to potential biases in the estimated emissions trends. Such high biases could lead to similar biases in the estimated emissions controls required for an area to attain the air quality standards. The results indicate that there are both spatial and temporal biases in how well the two models have captured the observed ozone and NO2 trends. Air quality planning agencies should aim to diagnose and reduce such biases before using model results for air quality management.
Read full abstract