
The following text was sent in response to a query about low-cost particulate matter (PM) or dust sensors that are widely available these days.
We have purchased ~30 PurpleAir monitors (~$200-250 per unit) and ~50 Met-One Neighborhood PM monitors (NPM, $1500-2000 per unit). Before deployment in the field, we run them for at least a week at our campus Supersite, in parallel with other PurpleAir and Met-One NPM monitors. We compare the results with our resident TEOM (when it works, it is aging) and the nearby Allegheny County Health Department (ACHD) reference BAM at Lawrenceville. Based on these tests, we have returned a few Met-Ones and PurpleAir units to the manufacturers for replacement because the results were not consistent with the other measurements. I have not investigated the inner workings of these sensors (e.g. light leakage) - replacement is easier, and the manufacturers are responsive.
Very cheap "dust" sensors like Shinyei might work for heavily polluted or dusty environments, but I would not recommend them for outdoor monitoring. Dust sensors might be good for supermicron (larger than 1000 nm) aerosols, but urban fine particulate mass (PM2.5) in the US is usually from secondary aerosol (sulfates, nitrates, organic aerosol) and combustion sources (gasoline and diesel vehicular exhaust, fires) which are submicron (under 1000 nm), mostly below 500 nm. This website has a good introduction to aerosols found in ambient air, and I stole the top figure from them. Here are two open-access publications evaluating dust sensors including the Shinyei: Austin et al. (2015) and Wang et al. (2015). Note that the Wang et al. paper tests the sensors at concentrations in the 0.2-1 milligram/m3 range, or almost 100 times typical PM2.5 levels in the US.
Now, (a) humidity can interfere with light scattering measurements which is how these low-cost sensors work; and (b) most aerosols grow in size with increasing humidity. So the EPA standard is based on PM measurements after conditioning filter samples at 24 hours at a relative humidity between 30-40% and temperature between 20-23 degrees Celsius (Carlton and Teitz, 2002). Continuous federal equivalent methods (FEMs) like the BAM have an inlet heater and are extensively tested for equivalence to such filter methods before being certified as FEMs.
The not-an-FEM Met-One NPM is nicely designed with an inlet heater (kicks in when RH exceeds 40%) and a PM2.5 cyclone, reflecting Met-One's traditional strength in aerosol measurements (they also make federal equivalent method BAMs and other scientific instruments.) The NPM with a pre-heater only reduces the humidity artifact, and it does not completely dry out the aerosol. So we apply corrections based on an assumed aerosol growth factor (mainly a function of relative humidity, RH), aerosol chemical composition typical for Pittsburgh, PA, and collocation with the ACHD BAM before we use the data.
Note: The RH we use for the aerosol growth factor correction is measured at ambient conditions (by the RAMP), not after the inlet heater of the NPM (Met-One does not provide this data.) The NPM data is then corrected to match with the ACHD BAM data (reported at 30-40% RH per FEM requirements) using a simple linear model. The net effect is a decrease from the as-reported NPM mass at high RH, and an increase in the as-reported NPM mass at low RH.
The PurpleAir (two Plantower PMS5003 units in one enclosure) seems to work well in the sense that they are consistent with each other and measure aerosols down to 300 nm optical diameter (Plantower claims 50% detection efficiency between 300-500 nm.) If the aerosol size distribution doesn't change significantly, the reported PM2.5 mass can be scaled or corrected by collocation with a reference monitor (e.g. TEOM or BAM.) We use a linear correction that matches the PurpleAir PM mass to the BAM-equivalent NPM data. Temperature and RH appear to be equally important for the PurpleAir correction.
Our methods are so extensive and rigorous because we use the data not only for community awareness, but also for science/publication, which is a much higher bar. For community awareness, you may only need to know "is the air quality now worse than usual?" For this, a PurpleAir that shows PM levels along with relative humidity may be sufficient. Our correction equation is simple enough (though based on Pittsburgh aerosol composition, which may be different from where you live) that it can be implemented in near-real time for display.
I have restricted this blog to sensors with which we have direct and extensive experience, and I feel comfortable recommending to others. We have tested other sensors on a limited basis; feel free to shoot me an email or tweet if you are interested in knowing more. Let me know if I have missed anything, either in the comments or by email (subu at cmu dot edu).