For a long time in my air quality monitor reviews, I’ve been stating that you probably shouldn’t rely on PM10 readings from any low cost air quality sensor. But why is this? It mostly comes down to the fact that low cost sensors (LCS) just aren’t very accurate when it comes to PM10. The image below (from this study) shows exactly how inaccurate LCS can be when it comes to PM10.
As you can, and as is probably surprising for many, LCS are typically far more accurate when it comes to PM1 and PM2.5 than when it comes to PM10. This can also be seen in other third-party testing, such as that by AQMD’s AQ-SPEC:
Notice how the PM10 accuracy tends to be far lower than PM2.5 and PM1? For even more examples across other databases, you can also view this article. But why is this? Well, there are a few key reasons:
1. Light Scattering and Particle Size:
These LCS mostly use light scattering - shining a light beam through the air and seeing how particles scatter it. Smaller particles (like PM1) scatter light in a more predictable pattern. Larger particles (like those in PM10) have more complex scattering, making it harder for the sensor to accurately “read” their size.
2. Binning and Classification:
Sensors have to classify particles into different size “bins” (think of it like sorting mail by size). This is called binning. For example, some sensors will sort < PM1, PM1 - PM2.5, PM2.5 - PM4, PM4 - PM10.
PM10 includes a wider range of sizes than just PM1. The sensor has to be more precise to correctly put a particle in the “PM10 but not PM2.5” bin, for example.
3. Sensor Design and Cost:
LCS often have simpler designs than more expensive reference-grade monitors. This can limit their ability to accurately measure the more varied light scattering patterns of larger particles.
More expensive sensors often use additional techniques, like particle counting or aerodynamic sizing, to improve accuracy for larger sizes, but this technology is far more expensive and not found in LCS.
4. Calibration:
Calibrating for PM10 is more complex than for PM1, as you’re dealing with a broader range of particle sizes and behaviors. This complexity can be a challenge for LCS.
Furthermore, since PM2.5 and PM1 are more of a focus these days (due to the wider health implications of these smaller particles), sensors are often calibrated for these size ranges, improving performance for these smaller categories of particles, but decreasing PM10 accuracy.
These challenges make PM10 very challenging to measure for LCS, and this is why when I get asked the question ‘What monitor should I purchase to monitor PM10?’ I still don’t have a good answer. Unfortunately, with our current technology, this is just an inherent limitation of air quality monitors. I’m excited to hopefully see this change in the future!
Glad you bring this aspect up.
The next is based on 10 years of doing low-cost (Open Source) dust measurements in a Dutch region with about 30 dust measurement DIY stations.
Probably all dust measurement products use a dust sensor from only a few manufacturers. Dust sensors which are originally applied in airco equipments where measurements does not need to be very exact as well humidity and temperature do not very much. These sensors make use of counting and sorting partikels in so called ‘bins’ (<0.4 mu,<1.0, <2.5, <4, <10mu). By the way not e.g. 2.5 mu - 10 mu. Plantower measures the other way around: e.g. >1.00 mu.
Air Quality levels are based on elder sensors e.g. NetOne BAM1020 (Beam Affection Meter) and weight an hourly collected partikels sample from dry air stream on a disc. So low-cost dust sensors need to be recalculated from the particle count. However humidity and temperature play a role in these calculations. Clearly in outdoor situations one has to adjust the algorithm constantly and use reference sensors.
As well low-cost sensors differ a lot (e.g. laser type). Later sensor types are better and give more values as well have functions to clean up, etc. I have only experience with a few (in chronologic and sorted order: Shiney (had no fan, uses IR, only PM counts), Nova (PM10 value is better), Plantower (provide bin counts, values are ca 150% higher as with others, has clen up function), and Sensirion (provide bin counts, has cleanup function).
Conclusion: it is good to look inside the box.
Bin counts: the good thing with bin counts is that one can do profiling. The bad thing is that most station manufacturers do not show these values or only PM2.5 weight values. Probably OK for indoor household measure measurements where the requirements are not strong and product lifetime of one year is ok?
At last: humidity sensors have problems with higher (>60%) humidity. Some just die. And particle sizes are (not linear) influenced by humidity… Which makes recalculation to weight values difficult. In other words: a simple PM sensor and a simple ESP without good firmware and application in outside environments requires some more efforts.
It is time that weather measurements and forecasts get also a dust component and inside how they come to those values.
This is another very good point that I forgot to mention. If you look at monitors that use Plantower, Sensirion, or Cubic sensors, that’s probably already over half of the consumer market.
Thank you for the clarification here. That was my mistake!
Thank you for providing some notes based on your experiences! This is all very helpful, and it’s great to learn more about these sensors and how they function.
I agree entirely. humidity (and less so, temperature) play a big part in the accuracy of these sensors and it would be good to have more research around this interaction.
