I’m looking at historical purple air data from the past year for a particular city (the city contains around 35 sensors total), and I have noticed several sensors showing a strange pattern:
This is a plot of the uncorrected PM2.5 data for one of the suspicious sensors. There are 7 sensors in my dataset which have an unexpected jump in PM2.5 in the 1700 range, and which stay at the value for up to several months. Should I assume the sensor is faulty? What could be causing such high readings? I appreciate any insight on what might be going on, and what kind of quality control I can do on this data. Thank you!
Very likely, this one sensor is faulty: NO air pollution index ever hits 1700 that WE know of. The highest ones worldwide have been known to top off over 800, but that’s the highest I’ve ever known. So, don’t woeey; there is no air pollution concern with any reading of 1700. That sensor needs to be fixed. Any other of those 35 showing ridiculously high readings, you can just ignore.
To clarify, this is the uncorrected data so the real reading is somewhere around 800 ug/m^3, but I still believe it’s probably wrong because it’s sustained for such a long period of time. Any idea on what could be causing the sensor to go faulty? This particular sensor returned to normal values after about a month.
Maybe somebody checked on that sensor and maybe they fixed it? I don’t know what would have caused such a high reading. If it is still reading according to the average going rate for a lot of the stations, then it’s probably good again.
This is most likely due to debris or a bug becoming caught in the sensor and blocking the laser. We typically recommend that sensor owners with similar issues refer to our Sensor Maintenance article. For your use case, you may want to consider this data as an outlier and remove it from your dataset, as the sensor is not reliable. Our confidence in a sensor’s data quality can be seen on the map by viewing the Confidence Score.
Thank you so much for this information!