Low-Cost Sensors for Air Quality Monitoring

The WM-Air project will be developing and deploying low-cost particulate matter monitors across the West Midlands. These sensors will be fully tested, validated and calibrated prior to and throughout the deployment period. Details of this will be made available in due course. Information about what the WM-Air project will and will not be able to do with project partners (both existing and new) regarding air quality monitoring can be found here.

The WM-Air Team often receive enquiries from stakeholders, project partners, external companies and community groups regarding the use of low-cost air quality sensors for their own requirements. They are often seeking our advice on using instrumentation that has been offered to them by a growing number of companies who are commercialising such low-cost sensors. We recognise that (on paper, at least) they are an appealing, cost-effective, and user-friendly solution that can offer a step-change in air pollution monitoring, yet it can be overwhelming when hundreds of companies are offering low-cost products, ranging from hundreds to thousands of pounds, monitoring various pollutants and with differing capabilities, accuracy and power and data transmission requirements. Whilst we cannot provide advice on which companies or particular sensors to work with, we can outline some guidance to help with deciding on whether low-cost sensors are a viable option for your application(s), and if so, what you should consider when choosing what sensors to purchase or which company to work with. 

In general, the benefit of such sensors is the increased spatiotemporal coverage and automated data collection and transmission in real-time, which is useful for a number of applications, particularly within urban areas, along roadsides or in remote environments.  However, low-cost sensors – by their very nature of being up to three orders of magnitude cheaper than standard reference instruments – may be unstable, imprecise and inaccurate if relied on in isolation. Whilst there are a number of companies offering well tested (i.e. in controlled lab environments and out in the field), calibrated and more reliable low-cost products (with documentation, publications and/or independent evaluations to accompany them), it is still important to understand their limitations. 

Low-cost sensors benefit from being inexpensive and compact. They require fewer or indeed no skilled operators with technical knowledge for maintenance or manual data collection, and may be operated in locations with little existing infrastructure. Overall, they are more useful for exploring general trends (i.e. improving/deteriorating air quality) in real-time, at a high spatiotemporal resolution, rather than providing ‘true’ and absolute values at specific locations.  Realistically, the data from one sensor will never provide the confidence needed to ‘trust’ the observation. Furthermore, if the variability in air quality at the monitoring site is known to be small, it may be that it is within the error of the instrumentation, and therefore won’t produce reliable information. As such, buying one or two low-cost sensors will be of very limited use. However, if a larger network is purchased, then the ‘noise’ between sensors becomes much more evident and the greater the chance of obtaining useful data.  For example, it is advisable to cluster a number of low-cost sensors at a location, and, if possible, to co-locate one or more of the low-cost sensors at an official air quality monitoring site. This ‘clustering’ approach will potentially allow for rogue data and/or faulty or inaccurate sensors to be identified. In short, it is not recommended to solely rely upon such devices in isolation or as the main method of data collection, but they can indeed be useful for supplementing air quality monitoring networks for specific applications. However, this does mean that low cost isn’t as low cost as it may first appear since dozens of sensors are actually required to achieve this approach.

After consideration of these points, if a ‘low-cost’ solution is still a viable option, please consider the points below when deciding on which is the most appropriate sensor to deploy.

Things to consider

  1. Pollutants. What air pollutant(s) does the sensor monitor? What sort of sensor is it and does it monitor the pollutant(s) you require data on?
  2. Cost. Are they within budget for the number required? Low cost sensors should be viewed as consumables and therefore need to be priced as such in order to be procured in suitable numbers for the desired application.  As per above, one or two sensors just isn’t sufficient.  Unit rotation may also be needed for calibration and should be accounted for with regards to budget and the number of sensors purchased. Are there any hidden costs beyond the initial purchase? (i.e. comms). What is the shelf life of the instrument (again, think consumables)?
  3. Power. Does it require mains power, which can often be problematic, or will it work on battery? ‘Off the shelf’ batteries or bespoke? Does it use integrated solar / other energy harvesting? Will this be appropriate for your application, chosen site(s) and timescale needs (i.e. should it last at least a season? Minimum of 3 months or longer)?
  4. Comms. What IP/comms protocol does it work on, and/or what is available at the site(s)? Is it future-proofed or will this be obsolete in a few years? If the sensor is self-contained, comms costs need to be low power. There can be numerous issues with utilising WiFi, Zigbee etc and dealing with sim cards for 4G/5G (which are also very power intensive), and these may not be a sustainable option long term.  Sigfox, NBIoT, LoRa are useful low-power options to consider but there are presently limited options in the Air Quality market.
  5. Staff support. Is it plug and play? (i.e. no specialist skills needed for deployment) Or will it require someone with technical knowledge? If so, and the sensor is self-contained and semi-mobile, the management of the device can be left in your hands to deploy, move, change the battery, send unit back for calibration etc.  
  6. Unit size and weight. What size and weight are the sensors? Is this appropriate for your siting needs (i.e. is the weight acceptable for wind loadings / health and safety purposes?) Is it inconspicuous (i.e. to limit vandalism if in unsecured locations)?
  7. Testing and calibration.  What literature is provided to support hardware/software accuracy and precision claims? Are the errors satisfactory? It needs to be able to produce ‘acceptable’ readings (i.e. stable with minimal drift) over the timeframe.  What support will the sensor manufacturer offer? Do they provide post-processing, correction or calibration of the data? 
  8. Data. Will you own the data? Will the sensor company host the data on their servers (is there a hidden cost?) or will it use a smart server, or in-house server? How secure is it (i.e. will the network be open to attack)? Does API need to be available to access data for own platforms? Will the data be ‘open’ or made available to others? What are you needs/requirements for this? 
  9. Visualisations and tools. Are there pre-existing visualisation or data analysis tools available for you to use (if required)?
  10. Recommendations. Do you know anyone else (i.e. a similar company, research institute, respected colleague(s)) who has used these sensors previously, and do they come recommended?