Explanation of the confidence values

I see the confidence properties listed in the API documentation but there is no explanation of them. Specifically, what is the difference between confidence, confidence_auto, and confidence_manual?

When calculating AQI I’m trying to decide when to average the pm cf values for A and B, and when to just use either A or B based on the confidence of each sensor. Is that possible? If so, what confidence threshold is recommended?



1 Like

As an alternative, you can use the EPA’s data removal criteria for their NowCast calculations of PA sensors:

Points removed if 24-hr averaged A & B PM2.5 differ by

  • ≥ ± 5 μg*m-3 AND
  • ≥ ± 2*sd(% error)
    • 56% for 24-hr data
    • 70% for 1-hr data

Source: PurpleAir PM2.5 U.S. Correction and Performance During Smoke Events 4/2020 | Science Inventory | US EPA

1 Like