Air quality models are frequently used to predict changes in air quality due to changes in emissions (e.g. potential emission control measures) or changes in meteorology (e.g. predicted climate change in future years). In these types of applications, it is more important for the model to accurately predict the relative change in pollutant levels rather than the absolute concentrations. 

A dynamic evaluation approach explicitly focuses on the model-predicted pollutant responses stemming from changes in emissions or meteorology. Dynamic evaluation approaches introduce new challenges compared to more traditional operational and diagnostic evaluation methods. In particular, retrospective case studies are needed that provide observable changes in air quality that can be closely related to known changes in emissions or meteorology.

EPA ’s Nitrogen Oxides (NOx) SIP Call required substantial reductions in NOx emissions from power plants in the eastern U.S. during summer ozone seasons with the emission controls being implemented during 2003 through May 31, 2004. Gégo et al. (2007) and USEPA (2007) show examples of how observed ozone levels decreased noticeably after the NOx SIP Call was implemented.  Gilliland et al. (2008) used this case study as a basis for dynamic evaluation of Community Multiscale Air Quality (CMAQ) model predictions of maximum 8-hour ozone for the summer of 2002 before the NOx emission reductions occurred compared to predictions for the summer of 2005 after the controls had been implemented. 

The figure below provides an example from this prototype evaluation study.  Spatial patterns of percentage decreases in ozone derived from observations and from the model exhibit strong similarities. However, these results also reveal model underestimation of ozone decreases as compared to observations, especially in the northeastern states at extended downwind distances from the Ohio River Valley source region.

Several hypotheses have been proposed to explain the muted model response in the CMAQ ozone predictions including:

• transport related issues,
• underestimation of NOx emissions reductions (specifically for mobile NOx in urban areas),
• muted chemical response in modeled NOx chemistry to emissions changes
• coarse vertical and horizontal grid resolution and
• unrealistic boundary conditions (e.g. lack of time-dependence, fixed profiles etc.). 

Evaluation of these different hypotheses led to many model improvements in the CMAQ version 5.0 modeling system released in 2012, including improvements to important model inputs such as the WRF meteorology model, source emissions and boundary conditions. The dynamic evaluation described in Gilliland et al. (2008) is currently being repeated with the latest modeling system to investigate the impact of these model updates on the model predicted response to the NOx SIP call. Additional model simulations will also be used to evaluate the model predicted change in pollutant levels due to recent controls on mobile emissions as well as the economic recession in 2008- and 2009. Findings from such dynamic evaluation case studies can ultimately lead to model improvements that are directly relevant to the way air quality models are used for regulatory decisions.

Example of dynamic evaluation showing a) observed and b) air quality model-predicted changes as percentages for differences between summer 2005 and summer 2002 ozone concentrations from Gilliland et al. (2008). The results illustrate the relative change in ozone when comparing the =95th percentage daily 8-hour maximum levels between the two summers.

Contacts: James M. Godowitch, Kristen Foley, Christian Hogrefe