FAQ

Q: We have a pollution problem in my area too, can we use Yahi to help with this?

A: You can certainly reuse the components: most of Yahi’s design, source code and data are on this website and GitHub under permissive licenses which support most likely uses.

It is worth understanding that there is a problem with getting from dust sensor counts to PM2.5 estimates: these are two different kinds of measurements:

• count of particles 1μm and above, and
• mass of particles 2.5μm and below as a fraction of the mass of the air containing them.

The former has proven to be a good proxy for the latter in Singapore because:

1. Singapore’s air is unusually clean - apart from the haze from Sumatra etc. - so when PM2.5 doubles, dust particle counts approximately double too, and
2. Singapore’s NEA publishes hourly PM2.5 data for five areas in Singapore, which provides Yahi’s reference source.

If you have a lot of non-haze particulate matter in the air then its composition is likely to differ from haze’s, so it will be difficult to separate one from the other with dust particle counters. Does a particle count of 300 consist of:

• 290 haze particles and 10 others of different composition, in which case the dust count may be a good proxy for PM2.5; or
• 150 of each, in which case the dust count will tell you very little?

Further, you can only construct a mapping curve for the proxy if you have a lab publishing reliable information. This generally means government labs because of the cost of the instruments plus the full-time professionals to tend them:

• precision “gravimetric” instruments start at US$150K,
• adequate “beta absorption” instruments start at US$15K,
• everything cheaper is dependent upon the composition of the particles and, therefore, unlikely to be accurate unless you have either of the above available as a reference.

Without an accurate reference you can still generate information about relative badness, but you’ll have to come up with a way to communicate that which doesn’t look like PM2.5 or PSI/AQI/etc.