Revealing the “data” and what came out of it
One day I set out to create a device that would measure air quality – not just somehow, but with high accuracy. The device development project led to the creation SkyAir – an inexpensive sensor that claims to be highly accurate. In this article I will talk about tests in real conditions and what happened in the end.
Why is this necessary?
Today there is a shortage of reliable and open data on air quality, especially at the low price segment level. Most public monitoring initiatives rely on low-cost sensors that are often not accepted by the scientific community due to their questionable accuracy. Traditional solutions like Palas Fidas 200 And BAM-1020 dominate the industry due to their accuracy, but they are expensive.
However, against the background of developing technologies and new approaches to data processing, even scientists are beginning to pay attention to such devices. SkyAir was designed to provide affordable and accurate air monitoring solutions.
Testing NeboAir
In a previous article, I talked about how the types of air monitoring devices differ. Now it's time to look at the results.
We tested the sensor SkyAir along with Palas Fidas 200 And BAM-1020which are the reference instruments for particle monitoring. Tests were carried out in two different climatic zones: Krasnoyarsk (Siberia) And Karlsruhe (Germany). These regions were chosen due to their different levels of air pollution.
Reference:
Palas Fidas 200 uses optical light scattering to measure particle concentration, which is similar to SkyAir.
BAM-1020 is based on the beta radiation attenuation method, which is considered one of the most accurate measurement methods.
Correlations or “how close are we to the truth”
We recorded data from our Nebo along with readings from the Palas Fidas 200 and BAM-1020 for three months. Palas received data every minute, our sensor received data every two minutes, and the BAM-1020 provided information once every 20 minutes.
Measurements SkyAir showed:
Correlation with Palas Fidas 200 at the level 0.91which indicates high data accuracy due to similar technology.
WITH BAM-1020 the correlation was 0.83which is also an excellent result given the difference in measurement methods.
The NeboAir sensor operated at extremely low temperatures (-30°C) and high pollution levels (150 µg/m³). Thanks to the built-in heating system, the device coped with such conditions, which became an important advantage.
Calibration and Models
Like any device, sensors require periodic calibration to ensure accuracy. We have developed statistical model based on linear regression, which further improved the accuracy of NeboAir. This model corrects data in real time and is one of the first versions of our auto-calibration system.
The next firmware update will add a model called “N1”which is based on more complex statistical algorithms, and was developed based on recent experiments to test the accuracy of our sensors.
To improve the results, the data was grouped and averaged to obtain more accurate values. After this, smoothing was applied using the moving average method. A coefficient was calculated between the readings of the “standard” sensor and NeboAir. To identify a trend, an exponential line is constructed, which helps to see the general trends in data changes, and we also take into account the influence of external factors such as humidity and temperature.
Model N1 takes into account the unique features of NeboAir, such as the volume of air inside the case and the heating system, which unfortunately makes this model not applicable to other devices.
Impact on science and prospects
Tests have shown that low-cost sensors can become a valuable part of environmental research and air monitoring along with traditional solutions. NeboAir's accuracy confirms its potential as a tool for mass environmental data collection.
SkyAir can be used to create air monitoring networks in cities, which will allow monitoring the dynamics of pollution in real time, even where this was previously impossible due to the high cost of equipment.
The next step is further calibration and testing in different climate conditions. The data obtained can become the basis for a new model “N2”.
I hope this study will spark interest in the scientific community. If you have questions or would like to test NeboAir sensors, just email me at igor@nebo.live.
Conclusion
If you're interested in exploring the data, you can access our results follow this link.
Without the help of a gas particle expert, Dr. Achim Dittler from the Karlsruhe Institute of Technology (KIT – the oldest technical higher education institution in Germany), this work would have been impossible to do. Thank you for sharing his experience in data analysis and I hope our cooperation will continue.
