The United States and Western Europe have seen great improvements in air quality, presumably in response to various regulations curtailing emissions from the broad range of sources that have contributed to local, regional, and global pollution. Such regulations, and the ensuing controls, however, have not come without costs, which are estimated at tens of billions of dollars per year. These costs motivate accountability-related questions such as, to what extent do regulations lead to emissions changes? More important, to what degree have the regulations provided the expected human health benefits?
Here, the impacts of specific regulations on both electricity generating unit (EGU) and on-road mobile sources are examined through the classical accountability process laid out in the 2003 Health Effects Institute report linking regulations to emissions to air quality to health effects, with a focus on the 1999-2013 period. This analysis centers on regulatory actions in the southeastern United States and their effects on health outcomes in the 5-county Atlanta metropolitan area. The regulations examined are largely driven by the 1990 Clean Air Act Amendments (C). This work investigates regulatory actions and controls promulgated on EGUs: the Acid Rain Program (ARP), the NOx Budget Trading Program (NBP), and the Clean Air Interstate Rule (CAIR) - and mobile sources: Tier 2 Gasoline Vehicle Standards and the 2007 Heavy Duty Diesel Rule.
Each step in the classic accountability process was addressed using one or more methods. Linking regulations to emissions was accomplished by identifying major federal regulations and the associated state regulations, along with analysis of individual facility emissions and control technologies and emissions modeling (e.g., using the U.S. Environmental Protection Agency's [U.S. EPA's] MOtor Vehicle Emissions Simulator [MOVES] mobile-source model). Regulators, including those from state environmental and transportation agencies, along with the public service commissions, play an important role in implementing federal rules and were involved along with other regional stakeholders in the study. We used trend analysis, air quality modeling, satellite data, and a ratio-of-ratios technique to investigate a critical current issue, a potential large bias in mobile-source oxides of nitrogen (NOx) emissions estimates.
The second link, emissions-air quality relationships, was addressed using both empirical analyses as well as chemical transport modeling employing the Community Multiscale Air Quality (CMAQ) model. Kolmogorov-Zurbenko filtering accounting for day of the year was used to separate the air quality signal into long-term, seasonal, weekday-holiday, and short-term meteorological signals. Regression modeling was then used to link emissions and meteorology to ambient concentrations for each of the species examined (ozone [O3], particulate matter ≤ 2.5 μm in aerodynamic diameter [PM2.5], nitrogen dioxide [NO2], sulfur dioxide [SO2], carbon monoxide [CO], sulfate [SO4-2], nitrate [NO3-], ammonium [NH4+], organic carbon [OC], and elemental carbon [EC]). CMAQ modeling was likewise used to link emissions changes to air quality changes, as well as to further establish the relative roles of meteorology versus emissions change impacts on air quality trends. CMAQ and empirical modeling were used to investigate aerosol acidity trends, employing the ISORROPIA II thermodynamic equilibrium model to calculate pH based on aerosol composition. The relationships between emissions and meteorology were then used to construct estimated counterfactual air quality time series of daily pollutant concentrations that would have occurred in the absence of the regulations. Uncertainties in counterfactual air quality were captured by the construction of 5,000 pollutant time series using a Monte Carlo sampling technique, accounting for uncertainties in emissions and model parameters.
Health impacts of the regulatory actions were assessed using data on cardiorespiratory emergency department (ED) visits, using patient-level data in the Atlanta area for the 1999-2013 period. Four outcome groups were chosen based on previous studies identifying associations with ambient air pollution: a combined respiratory disease (RD) category; the subgroup of RD presenting with asthma; a combined cardiovascular disease (CVD) category; and the subgroup of CVD presenting with congestive heart failure (CHF).
Models were fit to estimate the joint effects of multiple pollutants on ED visits in a time-series framework, using Poisson generalized linear models accounting for overdispersion, with a priori model formulations for temporal and meteorological covariates and lag structures. Several parameterizations were considered for the joint-effects models, including different sets of pollutants and models with nonlinear pollutant terms and first-order interactions among pollutants. Use of different periods for parameter estimates was assessed, as associations between pollutant levels and ED visits varied over the study period. A 7-pollutant, nonlinear model with pollutant interaction terms was chosen as the baseline model and fitted using pollutant and outcome data from 1999-2005 before regulations might have substantially changed the toxicity of pollutant mixtures. In separate analyses, these models were fitted using pollutant and outcome data from the entire 1999-2013 study period. Daily counterfactual time series of pollutant concentrations were then input into the health models, and the differences between the observed and counterfactual concentrations were used to estimate the impacts of the regulations on daily counts of ED visits. To account for the uncertainty in both the estimation of the counterfactual time series of ambient pollutant levels and the estimation of the health model parameters, we simulated 5,000 sets of parameter estimates using a multivariate normal distribution based on the observed variance-covariance matrix, allowing for uncertainty at each step of the chain of accountability. Sensitivity tests were conducted to assess the robustness of the results.
EGU NOx and SO2 emissions in the Southeast decreased by 82% and 83%, respectively, between 1999 and 2013, while mobile-source emissions controls led to estimated decreases in Atlanta-area pollutant emissions of between 61% and 93%, depending on pollutant. While EGU emissions were measured, mobile-source emissions were modeled. Our results are supportive of a potential high bias in mobile-source NOx and CO emissions estimates. Air quality benefits from regulatory actions have increased as programs have been fully implemented and have had varying impacts over different seasons. In a scenario that accounted for all emissions reductions across the period, observed Atlanta central monitoring site maximum daily 8-hour (MDA8h) O3 was estimated to have been reduced by controls in the summertime and increased in the wintertime, with a change in mean annual MDA8h O3 from 39.7 ppb (counterfactual) to 38.4 ppb (observed). PM2.5 reductions were observed year-round, with average 2013 values at 8.9 μg/m3 (observed) versus 19.1 μg/m3 (counterfactual). Empirical and CMAQ analyses found that long-term meteorological trends across the Southeast over the period examined played little role in the distribution of species concentrations, while emissions changes explained the decreases observed. Aerosol pH, which plays a key role in aerosol formation and dynamics and may have health implications, was typically very low (on the order of 1-2, but sometimes much lower), with little trend over time despite the stringent SO2 controls and SO42- reductions.
Using health models fit from 1999-2005, emissions reductions from all selected pollution-control policies led to an estimated 55,794 cardiorespiratory disease ED visits prevented (i.e., fewer observed ED visits than would have been expected under counterfactual scenarios) - 52,717 RD visits, of which 38,038 were for asthma, and 3,057 CVD visits, of which 2,104 were for CHF - among the residents of the 5-county area over the 1999-2013 period, an area with approximately 3.5 million people in 2013. During the final two years of the study (2012-2013), when pollution-control policies were most fully implemented and the associated benefits realized, these policies were estimated to prevent 5.9% of the RD ED visits that would have occurred in the absence of the policies (95% interval estimate: -0.4% to 12.3%); 16.5% of the asthma ED visits (95% interval estimate: 7.5% to 25.1%); 2.3% of the CVD ED visits (95% interval estimate: -1.8% to 6.2%); and -.6% of the CHF ED visits (95% interval estimate: 26.3% to 10.4%). Estimates of ED visits prevented were generally lower when using health models fit for the entire 1999-2013 study period.
Sensitivity analyses were conducted to show the impact of the choice of parameterization of the health models and to assess alternative definitions of the study area. When impacts were assessed for separate policy interventions, policies affecting emissions from EGUs, especially the ARP and the NBP, appeared to have had the greatest effect on prevention of RD and asthma ED visits.
This study demonstrates the effectiveness of regulations on improving air quality and health in the southeastern United States. It also demonstrates the complexities of accountability assessments as uncertainties are introduced in each step of the classic accountability process. While accounting for uncertainties in emissions, air quality-emissions relationships, and health models does lead to relatively large uncertainties in the estimated outcomes due to specific regulations, overall the benefits of regulations have been substantial.