Low-cost sensors aka LCS are commonly used in an effort to measure air pollutants like particulate matter all around the world, indoors and outdoors. Their low price has driven a lot of interest from many communities. Academics, experts, and consumers have embraced them because they are cheap to get and easy to embed in an IoT solution.

Countless air quality monitors use low-cost sensors (mostly from China) and although they are great as educational tools, their low accuracy leads to wrong conclusions most of the time.

Wrong conclusions are as bad as misinformation or fake news. Air pollution doesn’t kill instantly (most of the time) and it doesn’t create severe health issues in the short-term, but after an extended period or at least when we notice the consequences. One exception is carbon monoxide (CO) as it can kill people instantly and this is the reason we don’t see many low-cost CO sensors. There are some regulations that protect the consumers. Moreover, companies don’t want to take responsibility by using a low-cost CO sensor because they can get sued easily by the family of a victim when the air quality monitor won’t notice the increase of the gas indoors. Liability!

However, particulate matter sensors don’t fall under the same principles and many companies create sensors and monitors that don’t comply with standards. As a result, we have many products in the market that work as random number generators, especially when we measure larger particles above 5 μm in diameter aka PM10.

Low-cost PM sensors rely on Statistical Estimation for the various channels (PM1, PM2.5, PM10, etc.) and Aerosol Models to determine the mass concentration of a cloud of particles. They are not precise, they cannot repeat the results, and they lack accuracy. We call them black boxes because no manufacturer discloses the algorithm they use to estimate the PM values.

• Not Precise: They see a cloud of particles that’s the reason they are called nephelometers from the Greek word nephos which means cloud.
• Not Repeatable: These sensors don’t follow any standards so repeatability is a huge issue because without it you cannot be sure of what you are measuring.
• Not Accurate: In many cases the manufacturers auto-claim the accuracy base on what they believe is accurate.
• Lucky Guess: As they estimate mass from the cloud of particles and the algorithm they give a value that doesn’t mean that they know what they have measured.

Common tactic, manufacturers put inside a room hundreds of sensors and one reference monitor and they program all the sensors to have the same value as the reference monitor. They call this “calibration”. Particle distribution is not homogenous inside a room and there are many restrictions and issues with this technique. Sensors need to be calibrated one by one.

Finally, the amount of effort experts spend in manipulating the data out of the sensors after the measurements were taken, is also not practical. Especially, when we talk about big projects that require experts to provide frequent air quality analysis and results.

Keep in mind that mass concentration values are very forgiving with sub-micron particles, and this is the reason we see good correlations with PM1 from LCS and reference monitors. The uncertainty though inside these PM1 measurements is tremendous and only by measuring the particle number we can determine how accurate the sensors are.

The reason I wrote this article is because I see many academics relying on LCS for scientific studies. Sensor accuracy and precision are important because solid scientific research shapes the future of the world.

References:

• Dust, Low-Cost Sensors, and a Few Lessons Learned by Kerry Kelly
• Understanding low-cost PM sensors: A “teardown” meets aerosol science by Jessica Tryner