The previous calculation of the baseline and target erroneously included an additional undocumented step in the EH-1 calculation. The previous calculation divided the sum of the three-year average of AQI days over 100 (weighted by population and AQI) by the total population across all counties with at least one AQI day greater than 100. The calculation simplified the units (i.e., days, instead of AQI-weighted people-days) and yielded more digestible values; however, unintended consequences made it impossible to look at yearly trends in the metric.
This metric is designed to track the number of people exposed to unhealthy air quality days with some weighting for severity of that exposure. The revised calculation (i.e., following the original measure technical approach calculation and not including the additional step) applies adjustments so that poorer air quality days in areas with larger populations are given larger weights. When the yearly value for 2009 was calculated, EPA determined that the number of unhealthy AQI-weighted people-days was considerably lower than 2008. However, because many areas had no unhealthy days in 2009, the denominator in the erroneous calculation step was also smaller. The net effect was a 2009 value that was higher than the previous 10.85 baseline. This would imply, in the context of EH-1, that air quality deteriorated in 2009 relative to 2008, when it had not. The proposed revision of baseline and target values brings the baseline and target calculations in line with the existing definition of the measure and allows for meaningful interpretation of trends.
Steps in EH-1 Calculation
- For each year in the base period (e.g., for 2008 use 2006, 2007 and 2008), determine the daily AQI values by county.
- For days with AQI values over 100, divide the actual AQI value by 100.
- Sum these AQI-weighted days by year for each county.
- For each county, multiply the sum of yearly AQI-weighted days by the county’s population for a reference year (e.g., 2008).
- Average the population-weighted and AQI-weighted value for each county across the three years in the period.
- Sum the three-year average population-weighted and AQI-weighted values across all counties to determine the national metric value.
Simplified conceptual model showing why the revision is needed:
For simplicity, assume that only two counties have AQI days > 100 and that a single year can be representative of a three-year average. County A is a large population county with frequent episodes of poor air quality (e.g., 30 days with an average AQI of 120 on those days, population 2 million). County B is a smaller population county with less-frequent episodes of poor air quality (e.g., 6 days with an average AQI of 110, population 1 million). In the original calculation, the metric would have been: 72 million people days (County A) + 6.6 million people days (County B) / 3 million people = 26.2 days.
Now assume there’s a 10% improvement in ozone levels for the next year that affects all days equally. Assume there are still 30 days in County A but the average AQI is only 108. The improvement leads to no days with an AQI greater than 100 in County B. The original calculation would now be (30 * 1.08 * 2) million people days / 2 million people days, or 32.4 days. So despite the improvement in air quality, the original calculation leads to a counterintuitive increase in the number of days. This is fixed with the revised approach.