Atmospheric chemist and RADICAL team member, Dr Stig Hellebust, discusses some common misconceptions and describes the ideal air quality map.
Air quality maps rarely to help us find our way, but they can make the invisible visible and help guide our behaviour.
There are many air quality maps online that promise real-time air pollution data across the world. Well-known examples include global maps from groups such as PurpleAir, Plume Labs, OpenAQ, and sensor.community. There are also air quality maps from regional and national government networks such as the European Environment Agency, Ireland’s EPA, and the Dutch National Institute for Public Health and the Environment.
When looking at an air quality map, usually the first and only question people ask is:
“What can the map tell me about the air quality where I live?”
But there are two questions one should always ask first, in order to truly understand the quality and accuracy of an air quality map:
“Who produced it?”
“What’s its purpose?”
In order to answer these questions, we have to talk about two different kinds of air quality maps: ‘commercial maps’ and ‘national maps’, as we’ll call them here.
Commercial vs National Air Quality Maps
Many air quality maps are produced by private companies that are in the business of selling low-cost sensor devices. The maps display the data from all the devices that they have sold, predominantly to private individuals. Low-cost air quality sensors is a competitive market, and the map is part of the product, so these maps are generally very attractive and professionally designed.
But many of these maps are created purely as vessels for the data generated by the sensors and the purpose of creating them is often more to do with mining the data than anything else. The emphasis is as much on the dissemination of the data as generating it.
Fundamentally, the ‘network’ that these maps display is not a network designed for any purpose. In fact, it is not designed at all.
It is the result of customers purchasing devices to inform themselves about the air quality in their own location. Nobody has any influence over where these locations are. It depends who buys a device and where they install it. Perhaps in the majority of locations, they are deployed on the wall of someone’s house or apartment, but if the customer purchases a device in order to monitor the level of pollution next to a chimney stack or inside a busy multi-storey car park, the readings from these devices may vary greatly from the actual local area.
This is fundamentally different in scope and in purpose than an air quality map produced from a national air quality monitoring network. For example, in Ireland the Environmental Protection Agency (EPA) produces an air quality map available at airquality.ie. This map shows a network that is deliberately designed to sample a range of environments and give a true representation of the air quality experienced by the majority of the population, whether they live in a city, a town, near a busy road, or out in the countryside.
The network is designed to incorporate all relevant environments, and the selection criteria are based on legislation and implemented by experts. The purpose of the network is to ensure compliance with air quality standards. However, the air quality map of such a network may be less attractive or less user-friendly than the commercial maps. Often, compliance with national standards is emphasised more than public accessibility, so overstretched agencies are not always given the resources to develop and promote these maps for a wider audience.
There is another crucial difference between official measurements and measurements produced by private low-cost sensor devices. Accuracy.
Official air quality measurements are collected to monitor compliance with legal air quality standards. That means they have to be accurate enough to stand up in court and comply with rigorous analytical standards. The monitoring therefore employs so-called reference methods. In contrast, the term for the low-cost sensor data is indicative measurements.
This difference between reference methods and indicative measurements can be significant. It is not unknown for reference and indicative measurements to differ by a factor of two in the one location, and typically, the indicative measurements are higher.
For example, in the case of the local network of PurpleAir sensors in Cork City, our group has been able to determine that these sensor values are consistently higher than the reference measurements in our Air Quality Monitoring Station in Distillery Fields by about a factor of two. So people looking at lost-cost sensor maps could often get the impression that the air quality is worse than it actually is.
Amount of data does not equal accuracy
The key thing here to bear in mind is that the accuracy of a map is not always proportional to its number of data points. This is because many of the locations are effectively the same, from an air quality point of view. Having 20 monitoring points in one city sounds better than having 4, but if those 20 represent only three types of environment, badly, then 4 high quality measurements can contain more information if they are appropriately located.
Different environments will be affected by a different set of local pollution sources, such as population density, traffic conditions, and local industry. A residential housing estate in county Dublin could experience similar air quality to a rural town in county Offaly, and very different from the city of Dublin, although it might be much closer to Dublin than to Offaly, geographically speaking. So it is important that each sensor on a map is a good representation of its area.
Simplistically, air quality maps are created by colouring in between the dots. So each point hugely influences the colour of the area around it. A large city park could look like the most polluted part of the city if the nearest sensor is located next to a chimney or busy road junction. It doesn’t necessarily represent the typical park environment.
So we finally turn to the key question:
What can the map tell me about the air quality where I live?
Assuming the sensors are situated appropriately, air quality maps allow us to compare different locations on the microscale. So when one is going for a run in the evening or a cycle to work in the morning, one can adjust the route according to where the air is cleanest.
This is the real strength of the low-cost sensor networks. While they may not be accurate, they do enable a quick relative comparison between different streets (provided the sensors are located with some thought). But the values have to be updated in real-time, because the comparison is only useful at each point in time.
In contrast, national reference networks often focus on daily or annual averages since this is what air quality standards are based on. For example, Ireland’s EPA air quality map uses daily averages of particulate matter. But pollution levels fluctuate greatly during the day and there are certain times of day when it is not a good idea to be out jogging. Real-time air quality maps can help people make an informed decision.
The ideal air quality map
The most accurate air quality maps are always the ones that are based on the official air quality monitoring networks, such as Ireland’s EPA air quality map. Unfortunately these are not always the most accessible or attractive maps, and they don’t always show real-time data. In contrast, the maps that are based on networks of commercial low-cost sensors are usually part of the product and therefore well-designed and attractive.
So the ideal air quality map combines reference data and indicative data, merged with a degree of quality control, and presented in an easy-to-use map.
As an example of this, the Dutch National Institute for Public Health and the Environment has produced a very good air quality web portal that publishes sensor data for the whole country, and at the same time gives the reference stations for comparison. In this map, the low cost sensors are subject to a simple but robust correction algorithm that takes into account the relevant reference measurements and applies individual correction factors to all the sensors in the network. In my opinion, they have created a good model for other organisations to follow, both on a local and national scale, e.g. city councils, county councils and national agencies.
Cork Air Quality Dashboard
Our research group, the Centre for Research into Atmospheric Chemistry (CRAC) at UCC has recently been working with Cork City Council to develop a local air quality dashboard for the city of Cork, available at corkairquality.ie.
This is based on low-cost devices from PurpleAir, but they are deployed by the City Council as part of a planned strategy, and the measurements are calibrated against the reference methods in the monitoring station operated by our group.
We hope it blends the best of both types of air quality maps and provides a useful and intuitive resource for the people of Cork city.
Where can I find out more?
Centre for Research into Atmospheric Chemistry (CRAC), UCC
Air Pollution, European Environment Agency
Air Pollution, World Health Organization
About the author: Dr. Stig Hellebust is an atmospheric chemist and environmental scientist working the Centre for Research into Atmospheric Chemistry at University College Cork. His research interests have focused on source apportionment studies of atmospheric aerosols based on chemical fingerprints and automated detection of bioaerosols. He is currently involved in active research projects on linking air quality and health (INHALE), emissions from upland wildfires (FLARES), biological aerosols in the atmosphere (FONTANA) and airborne spread of SARS-CoV-2 (UPCOM) and this RADICAL project.