Enviroment Canada health monitoring

Next month I will begin a post doc at the university of Dalhousie, Halifax. I've been looking forward to starting a new project, so here we go.

Environment Canada is facing some serious cuts this year. It's no surprise that the government is tithing its departments equally after the budget was announced, yet some environmental studies have been in place for decades, such as the Experimental Lake Area (studies include the spread of mercury pollution in a marsh environment, METAALICUS).

The National Air Pollution Surveillance Program (NAPS) is a subsidiary of EC, though it looks like some projects are facing cuts. The team collectively monitors air quality in most provinces. This includes Alberta. It seems the federal government might not be all that curious how polluting the tar sands are. Hmmm... Fortunately there are many sub-programs withing the NAPS, some of which monitor general air quality in major cities, like what I'll be doing. The air quality health index (AQHI) checks on levels of O₃, NO₂, and (particulate matter) PM₁₀ / PM_2.5 within Canada and reported on a scale of 1-10. The last of these quantities, PM_2.5, is my future specialty. I have to say browsing the EC website, it's not easy to find the precise formula they use. For instance in the FAQ section, the answer to "How is the AQHI calculated?" they reply

In the development of the Air Quality Health Index, a formula that combined these three pollutants were [sic] found to be the best indicator of the health risk of the combined impact of the mix of pollutants in the air.

But what's the formula? I found it in the publication titled A New Multipollutant, No-Threshold Air Quality Health Index Based on Short-Term Associations Observed in Daily Time-Series Analyses:

I can see why this information isn't listed prominently on EC's website, though the paper should have been linked via the FAQ. What this formula attempts is take an incredibly complex concept -surplus urban deaths from three major pollutant sources- and translate the information into a scale between 1 and 10(+). A huge task; naturally the devil is in the details. Herein we find the interesting and contentious data: they lie in the regression coefficients that link pollutants to mortality. You have to dig further to find what the excess mortality is for a given pollutant. The authors state

the percent increase in mortality associated with the mean concentration of each pollutant was: CO (1.8 ppm) = 0.83 +/- 0.45; NO₂ (33.6 ppb) = 2.08 +/- 1.19; O₃ (29.9 ppb) = 1.82 +/- 0.61; PM₁₀ (25 ug/m³) = 2.20 +/- 1.10; PM_2.5 (12.8 ug/m³) = 1.69 +/- 0.79 and SO₂ (13.7 ppb) = 0.55 +/- 0.33

Hence for the given concentrations of pollutants listed above, that amount of additional pollution will cause such-and-such percent excess deaths. This is a complex idea that isn't discussed directly in the above-quoted paper, only reported (there would be no room to discuss them). Regardless, what the numbers mean to state is the relation between pollutant of concentration X and their associated danger. For low concentrations this is roughly linear: e^bx-1 = bx for bx <<1. For a doubling of carbon monoxide in the air (say 1.8 ppm to 3.6 ppm) there is a doubling of danger of sudden (?) death from 0.83 % to 1.66% for the deaths already present in a city's population as a whole (defined as m in the paper). Likely cause of the 'excess' deaths? Cardiac arrest [NB: Continual breathing of CO above 1,000 ppm leads to death within hours; no more than 9 ppm should ever be breathed for any significant period of time].

These numbers are not without controversy. For starters we must accept there is even potentially a single attributable risk number for each separate pollutant, nearly as difficult as associating an IQ with someone's intelligence (but with more scientific grounding). Also there can be debate as to where (elevation, which cities) and when (time of day/year) the data is measured, and who exactly declares if such deaths occur (post-mortem by a team of doctors/other specialists?). These are all difficult topics to agree upon. Many co-factors must be eliminated (death from smoking, other diseases etc) and I see this reflected in the very broad 2-sigma (95%) confidence intervals; pollutants' pre-factors have a 30-60% uncertainty.

The rest of the formulaic conversion (from pollution levels to the AQHI) is accounting; these values are converted into percents while keeping the averages representative of large cities and establishing confidence limits, sensitivity tests, etc. Recall that CO and SO₂ have not been included in the AQHI; this is because they were found too well correlated with the other pollutants (hence did not provide an independent measure). But all this detail is the least of the confusions for curious readers, as the EC FAQ informs us that every place region has a different formula to consider:

The American index is standards-based and emphasizes the impact of a single pollutant, much in the same way as provincial  air quality indices have done up until now, The different jurisdictions have different standards and in addition the US uses different scales and categories.

See for instance, the EPA's Air Quality System (AQS).