The problem with AQICN

2 min read
Reading Time: 2 minutes

Singapore Air Quality Index

The website is very popular when haze levels increase (as explained here). But there is one problem: their risk interpretation is wrong.

They calculate the Air Quality Index based on 2 inputs:

The conversion behind that calculator can be roughly summarised as follows:


So far, so good.

Only one problem: those health scores are based on a different PM2.5 input.

EPA’s bands for Good, Moderate, etc are based on 24 hour numbers. If you are exposed to a 24hr average PM2.5 of 100, it is unhealthy. AQICN, however, uses the 1hr number. The health implications for exposure to a 1-hr average number has not been determined. This means that the risk is almost certainly overstated.

Singapore’s calculation of PSI takes into consideration 6 different pollutants. While it is a “3 hour average” number, the calculation is affected by some 24 hour readings.

PSI Calculation

EPA’s own recommended calculation of AQI is in fact far more complex that conversion of PM2.5. It looks at all available pollutant readings and calculates AQI based on a maximum hazard value of pollutants measured. The measurements are the same as those used by NEA in its calculation of PSI. It does not use 1 hour PM2.5.

As a measure of health risk, both the NEA and the EPA use the same measurements and take a similar approach.

AQICN, on the other hand, has misused the EPA health bands to estimate risk.

So it is wrong to say AQICN is more accurate – certainly not as a measurement of health risk. There is no basis for the cutoff points for each band.

It will, however, fluctuate more widely. It is based on 1 hour numbers and will therefore more closely approximate people’s perceptions of haze volatility on an hour to hour basis.

So people will like it, but the number will certainly be misunderstood because it uses a different risk scale.

Note perfect but still useful

Saying that, this 1hr number is far more useful for making on the spot decisions about outdoor activities than the 3hr PSI which looks at longer trends. Yes, use AQICN, but do not expect the government to announce a state of emergency based on their numbers.