Accuracy correction formulae for PurpleAir sensors

[nilaykumar]

I've stumbled upon a few papers/projects that compare the PM2.5 numbers reported by PurpleAir sensors and (typically colocated) FRMs/FEMs (monitors used by the EPA or similar). It seems to be widely understood that the PA sensors are very consistent but tend to report much higher numbers. To this end there are various correction formulae (for example, here's a 2020 EPA formula) for comparing to numbers reported by e.g. the EPA. PA's online map can compute these for you (and briefly explains them as well).

The data downloaded from the API does not seem to incorporate these corrections, so it may be worth redoing our previous analyses with some of these corrections taken into account (they typically rely on relative humidity and/or temperature, which are available in the datasets already downloaded from the API), as we are comparing PA monitors to the EPA's AirNow monitors.

A quick estimate (assuming 60% constant relative humidity) of these corrections shows that they may make a significant difference in the conclusions of the analysis:

[nilaykumar]

I've started working on this here: https://github.com/nilaykumar/Air-Quality/tree/accuracy-correction