PurpleAir and other sensors repeatedly give higher concentration values for PM2.5 than the sensor community values. This can be as high as 5x
This is significant if the difference is 5 or 15 ug/m^3
I show a snap shot of the concentration readings for the same location.
Please can you explain why this is the case.
Between 2 and 5, the difference is not that big. The resolution of the SD011 is not that small. The same for the PMS7003 in the PurpleAir. And we have no exact colocations here.
We could ask: Are PurpleAir sensors readings 3-5x higher than others?
A huge comparative study (with different kind of low cost sensors and official stations) will start soon in France.
Before people were complaining because SC readings were too high because of humidity etc… 
Thank you for the reply. At these low figures I agree you could argue that the difference is low. The median of the two Sensor Community (SC) sensors is 1, so I was using that figure as I have also seen a difference of 5 and 30. Local to me in Salisbury, I have 8 sensors that give an annual average PM2.5 concentration of around 3ug/m^3 the reference sensor in the city is reporting 18ug/m^3, so this is again around 5x higher.
When I compare the SC sensors in these semi rural areas even at 3-4 km away the sensors are in reasonable agreement with each other. I attach a screenshot showing this. The least squares of the fit of the comparison of the data is high at 0.9, so I interpret this as meaning that the particles in the air mass are non local and therefore all sensors should be reading the same to within the accuracy of the devices even when 3 or more kms away. I would be grateful if you would let me know if there is a different interpretation possible as I would like to know how to interpret hot spots from this data. It is interesting that you say people complained that they were reading too high as I do recall from 3 years ago higher readings that seemed more accurate to me for my location. That said, I am glad that you will be testing multiple sensor types together. I hope to be gifted a PurpleAir sensor from my council and I will be able to test them side by side. I look forward to the report on your results too. I hope this further information is useful, my main concern is that the SC sensors suggest my local air is reasonably safe, but the reference sensor is reporting something different. As always thank you for your support.
I’ve been running an SDS011 sensor since 2018 and a PMS5003 and PMS7003 since 2021 in the same location. My PMS sensors almost always give higher readings than the SDS011. My SDS011 readings are similar to others in my area (around Gloucester), whereas my PMS7003 readings are the most similar to those of the official monitor nearby that was set up by DEFRA only 6 weeks ago.
https://plumeplotter.com/monitor/
1 Like
As for the quantitative difference between the sensors, here is a plot of the latest year’s readings from my colocated SDS011 and PMS7003 sensors. The correlation is very good but the PMS7003 readings are 1.91 times the SDS011 readings (for this example).
1 Like
Not sure about SDS011, but PMS5003 (probably newest version as well) tends to have an upward bias.
Speaking about PurpleAir in particular, there are some studies that have come with various correction algorithms.
You can apply it directly on the Purpleair map. Just click on the little gear at the top left corner and change conversion to “woodsmoke”.
1 Like
Thank you for the added information.
Thank you for this. I have quickly looked at using the woodsmoke correction and see there are a number of other correction algorithms. Woodsmoke probably does apply to some of the larger differences I have seen, but perhaps not all. I need to investigate it further, but an initial glance at the woodsmoke algorithm suggests it adds a correction for humidity, if that is the case then this correction would also need to be applied to all sensors that do not dry the air.
I realise it is difficult to compare devices as there are many factors involved e.g. inlet efficiency and even the reference sensors can differ by 20%. So it is not exactly known what the ground truth is. However my main concern is that the SC sensors in my location suggest that the air is relatively healthy when the reference system reports a different scenario.