We are super thrilled about launching this and the examples that I describe in the blog post are quite encouraging that we can get some really useful data out of it.
How do you account for inaccuracy in the measurements? I have two different sensors (from different companies) side by side. Sometimes they are spot on, but right now they are 100 ppm different.
At the moment we have the automatic baseline calibration set to a default of seven days. This means the sensor modules auto calibrate to around 400ppm every week. This typically gets them very close to each other when co-located.
However, as the article points out, the measurements are not reference grade and should not be taken as an absolute correct value but our tests show that the sensor modules are actually performing very well [1].
Rather it is important to look at the patterns of the measurements, e.g. do they correlate with traffic patterns or wind directions (e.g. downstream from a power plant etc).
I think the example I give in the original post about the pollution from the speed boats captures this quite well.
We do hope that with this open-source hardware project, we can motivate much more people to build a monitor and contribute their data. Then with the growing data set there can be extremely interesting applications with machine learning and adding additional data sources like wind direction, emission inventories etc.
The data is freely available through our public API [2] and if somebody wants to start modelling it, please get in touch with us. It would be great to co-operate.
As the planet gradually climbs above 400ppm as an average - or even low - everywhere, won't that mean these sensors become less and less accurate? Is there a plan to handle that?
edit: still precise, probably highly precise, but .... the mark just goes up and up?
We are super thrilled about launching this and the examples that I describe in the blog post are quite encouraging that we can get some really useful data out of it.
Happy to answer any questions.