Filter out or weight "low confidence" PurpleAir stations

[deanlyoung]

If sensors are faulty or reporting abnormally for some other reason (PurpleAir's algorithm), they would ideally be discarded or weighted lower in averaging.

[ianhoffman]

Per this comment, I think that this needs to be added to both the build-side and application-side logic. We already filter out sensors reporting pm2.5 <= 0 or > 500 (but I think 500 is too low — I have seen readings up to 3000 though, which can't be right). What fields do you look at in the Purpleair API response to determine the confidence interval? I've been using their JSON API (example: https://www.purpleair.com/json?show=37124) because I didn't learn about their REST API until recently, but maybe it's worth switching if the REST API exposes confidence intervals.

[deanlyoung]

Ah, but there are two fields in the json that do this for you.

[deanlyoung]

https://docs.google.com/document/d/15ijz94dXJ-YAZLi9iZ_RaBwrZ4KtYeCy08goGBwnbCU/edit

[deanlyoung]

Interesting... the example Station ID you posted seems to have far fewer fields than quite a few stations near me 🤔

https://www.purpleair.com/json?show=54351

[ianhoffman]

I think it may be that the sensor you linked to has a "child" sensor (I'm still not sure what that means). I don't see A_H or Flag fields for either sensor.

[deanlyoung]

They seem to only appear when the conditions are met (again, PurpleAir's algorithm 🤷🏻‍♂️). For example:

https://www.purpleair.com/json?show=17781

[deanlyoung]

And the ParentID is for 2 channel (i.e. 2 particle sensors) stations so that you can poll them separately if one channel is faulty (afaik).