Introduction
Dr. Ruaraidh Dobson is an air quality and health data specialist who leads South London Scientific Ltd, an independent consultancy focused on evidence-based approaches to understanding, communicating, and improving air quality.
He holds a PhD in Health Science from the University of Stirling and serves on the OpenAQ Advisory Board. His research spans exposure science, environmental epidemiology, and the critical challenge of translating air quality data into action. Dr. Dobson previously led the data portfolio at the Clean Air Fund, the world’s largest philanthropic funder of clean air projects, and served as Clean Air Champion for Scotland.
The Local Haze team recently had the pleasure of chatting with Dr. Dobson about the challenges of working on AQ monitoring and our shared goals of making air quality information accessible and actionable. We recently caught up with Dr. Dobson to ask him about his thoughts on air quality monitoring, his current work, and the challenges of translating data into action.
Citizen Science and Air Quality
In recent years, there has been a proliferation of low-cost air quality sensors that use optical diffraction (nephelometers) connected to Internet of Things (IoT) networks such as PurpleAir (shown above at left), AirGradient (shown above at right) and Sensor.Community.
How do you see the usefulness of this citizen science data as compared to the classic government reference monitors and regulatory sensors? (Quality, Coverage, Data Type, etc.)
I think the most important element isn’t the sensor technology – it’s what you’re trying to use the data for. Ultra-low-cost instruments will never replace reference monitors because they can’t provide the kind of data a gravimetric instrument can provide. If you need to know detailed size fractions, volatility, etc, low-cost instruments won’t help.
But for many uses – maybe most uses – that kind of information isn’t important. If you want to communicate with the public, “this is your street” is usually a better message than “here’s a representative site several kilometres away”. Or if you’re primarily interested in the direction of change, a low-cost nephelometer (see example Plantower PMS5003 below) is a great instrument – you don’t need extreme precision to tell when someone’s started burning waste in an alley!
Beyond that, quantity has a quality all its own. Very large deployments can even out sensor error in aggregate (if a sensor can be miscalibrated ‘up’ or ‘down’, an average will tend to be more accurate than any individual unit). That’s not a panacea, but it does help.
The main problem I see when governments or campaign groups install these networks is that they don’t really know what they want from them. A network set up to detect changes in woodburning before and after an intervention needs different considerations from one looking at traffic, for instance, and both are different from one you might use for campaigning. Quite often, people will want to install sensors but without defining what they’re for. That should be the first question you ask, not tacked on at the end.
Measuring the Health Impact of Air Quality
The long-term health impact of air quality on an individual can be thought of in terms of the total inhaled pollution over the entire lifespan. Recently, a new generation of personal, dosage-meter-style air quality monitors has entered the market, such as the AirGradient Go (shown above at left) and Atmo Atmotube PRO 2 (shown above at right).
How do you see these devices impacting the world of air quality measurement?
These instruments are very cool! New sensor modules are making smaller and smaller wearables possible – the Bosch BMV080 in particular is extraordinarily capable and ridiculously tiny (I bought one and nearly lost it the second I took it out of the box).
There are lots of concerns around data quality with wearables, and while I understand it, I think we can overdo the worry. As long as we’re confident the monitors produce comparable data among themselves (and can be calibrated with reasonable confidence to the major sources likely to be encountered), they’re more than usable for epidemiology. That could get us much further along the line from analysing big grid squares of ambient air pollution retrospectively to a more nuanced understanding of exposure and health.
From a scientific perspective, the main issue is data collection – are we likely to be able to get meaningful data for long enough to conduct meaningful epidemiology for chronic diseases? I suspect modelling might be the answer there – the more we know about real exposure, the better our models can become.
The other problem I can see is really about business models. There’s a risk these monitors fall between the cracks of personal vs occupational vs governmental monitoring. They might not provide enough useful information for an individual to use them (if you can’t do anything with the information, would you use one?) which could limit adoption and the breadth of data collected. Governments and occupational health people are understandably conservative – when the rules say you should collect data in a specific way to see if there’s been an exceedance, you only see risk, not opportunity.
But these are solvable problems!
Is Particulate Matter the best way of quantifying Air Quality?
The measurement of particulate matter in terms of PM1.0, PM2.5, and PM10 has become the default way of assessing air quality. However, in terms of human health impact, the composition of particulate matter (carbon, heavy metals, etc.) is highly significant, making PM concentration not a great health indicator.
How do you think that air quality monitoring should evolve to better quantify health impact? Should we be measuring more gaseous pollutants (NOx, VOCs) in addition to PM?
We tend to treat PM2.5 (or other size fractions) as one thing, a bit like ozone or nitrogen dioxide. But PM covers a multitude of sins! Particle size, shape, number, and chemical composition are the real drivers of ill health. A lot of the problems we now see with low-cost nephelometer and Optical Particle Counter (OPC) deployment are exacerbated by the need to calibrate them to a largely unrelated measurement. Mass concentration is a measure of convenience – one that was great when all we had were filters and microbalances. But there’s no consistent way to determine how mass relates to particle size and number other than lab calibration, and that’s going to be a problem until we come up with a new paradigm. That’s a collective action problem – answers on a postcard, please…
Gas sensing is a separate issue. Ideally, I think we should be monitoring NO2 and ozone as widely as we’re looking at ambient PM. But it’s easy to forget the sheer scale of the challenge. CO2, for example, can be detected fairly accurately using NDIR sensors (albeit needing frequent calibration). But CO2 is measured in parts per million (around 428ppm in ambient air at the moment, terrifyingly enough). By contrast, a busy road in London might have a nitrogen dioxide concentration of 30 parts per billion – with a ‘b’. That’s 10,000 times less! And that’s without considering cross-reactivity, which is a huge and underappreciated issue.
I think there would need to be a significant breakthrough in solid-state trace gas sensing before we were able to make a big shift, and I don’t see one on the horizon. MOx sensing is, if anything, more vulnerable to cross-reactivity than electrochemical sensing; photoacoustic sensors don’t work outside of the lab; NDIR sensors are at the mercy of the absorption spectrum of the target gas (helpful for CO2, a nightmare for NO2).
Thank you Dr. Dobson, we appreciate you joining us today to share your insights on air quality monitoring.
Disclosure:
Some of the links in this post are affiliate links. This means if you click the link and purchase the item, Local Haze will receive a small commission. Thank you for supporting this blog.