What do you need: more expensive or a lot?

This article was supposed to be about our comparative data analysis between the Nebo sensor and reference air monitoring sensors. However, my introductory text has grown to more than one page. So let’s organize it into a separate article.

The two devices you see in the photo, one of which looks like a space container, are often called reference-grade monitors. The smaller one is a BAM (Beta Attenuation Monitor). Reference-grade means they possess the highest scientific quality, essentially the gold standard. These instruments are very expensive, ranging from tens of thousands to hundreds of thousands of dollars, just for the initial purchase. Add to that tens of thousands more for maintenance. Therefore, such instruments are used in developed countries where the government “can afford” them, or in less developed countries with grants from foundations, which is still beneficial. Sometimes, data from these stations are not published to avoid skewing statistics or “alarming the public.”

When you search “Air quality in Parma, Italy,” you might land on the first weather service links. Very convenient — real-time data.

However, next to the value, there’s often an asterisk hiding a mysterious note: “Modeled Using Satellite Data.” In areas with less coverage, meteorological services use this crafty method to provide data to users and expand coverage. You often see interpolated surfaces reflecting air quality. Lower costs — calculated on a server and provided to hundreds of thousands of users. But I would approach such sources with skepticism, as I previously wrote in a short article (link). Therefore, our entire team supports increasing the number of monitoring stations: the more sensors on the ground, the more accurate the modeling and forecasting.

Due to the lack of ground data, devices known as low-cost sensors (LCS) emerged in the market. These are consumer-grade products, meaning you or I can buy such a sensor for personal use, using household funds. The main thing is to justify to your spouse why you need this gadget. And as you can see in the photo, they are significantly smaller.

In the early stages of development, low-cost sensors were not very accurate, drawing criticism from the scientific community. And rightly so. Initially, LCS equipment had a high margin of error due to the specifics of optical measurement technology, influenced by factors like humidity and negative temperatures. Laser technology is limited to a temperature range of +10 to +40 degrees Celsius.

Inside the sensor, there is a laser emitting a narrow beam of light. Air is drawn into the sensor through a fan or pump and passes through the laser beam. Dust particles scatter the light, which is detected by a photodetector. The electronics analyze the data and convert it into the concentration of PM2.5 particles in the air.

The Swiss-German company IQAir was one of the first to use this method, installing a sensor next to a BAM. They calibrated their sensor data against the reference-grade data. The company Airly went even further, using AI to calibrate data against nearby “benchmark stations.” PurpleAir inserted two optical sensors for reliability and also used calibration.

The German company Palas combined mathematical fitting and elimination of the technical limitations of optical technology. They used a dryer and air heater in their Fidas 200 sensors, improving accuracy. However, such instruments cost around $20,000.

We wanted cheaper, more compact sensors that used both methods. Thus, our window sensor Nebo was born: we integrated an air heater and, albeit with some apprehension, decided to test its accuracy. We did this in Germany and in freezing Siberia.

In one of our project presentations, we calculated that instead of installing just one “reference” station, over 100 low-cost stations with good data quality could be placed in a city. This approach would provide broader coverage and exceed the data quality of a single reference station.

I thank Prof. Dr. Achim Dittler for fruitful discussions and contributions. I hope to jointly publish an article on the findings very soon.