Differences in the variability of measured and forecasted Tropospheric ozone mixing ratios over the Paso del Norte Region

Publication Date


Document Type



Stockwell, W.R., Fitzgerald, R.M., Lu, D. et al. J Atmos Chem (2013) 70: 91. https://doi.org/10.1007/s10874-013-9253-x


The objective of this study is to present differences in the variability of observed and ozone-mixing ratios simulated by a three-dimensional atmospheric chemical model using two chemical mechanisms. In this study the Comprehensive Air Quality Model with Extensions is used to make ozone simulations with the Carbon Bond mechanism, versions 4 and 5. The Paso del Norte region is used as a test-bed for these simulations. The shared variance between the simulations and measurements is typical for air quality models ranging from 0.51 to 0.86 for both mechanisms. The smallest mean normalized gross error is about 31 % with CB4 but the normalized bias is over 30 % as well. Boundary conditions, emissions and other factors affect the levels of ozone of the simulated mixing ratios and therefore error and bias but these factors have a much less affect on the simulated ozone variability. The differences in the ozone variability of the measurements and the simulations are very large and different for the two chemical mechanisms. There are many more instances of low ozone mixing ratios in the measurements than in the simulated ozone. One possible explanation is that these differences are due to problems associated with comparing point measurements with grid averages. A more disturbing possibility is that the bias could be due to the procedures used in the development and testing of air quality modeling systems. Air quality mechanisms are evaluated against environmental chamber data where the chemistry occurs at high concentrations and this may lead to a systematic positive bias in ozone simulations.