Procuring and deploying low-cost sensor networks

By Dr Nicole Cowell, Dr Catherine Muller and Prof Lee Chapman, WM-AIR, University of Birmingham

Low-cost sensors are becoming an increasingly popular option for monitoring air quality due to their reduced capital costs and small stature, making them suited to sampling in previously unattainable locations.

Many users are turning to low-cost sensors to supplement regulatory sampling or for targeted air pollution monitoring. Low-cost sensors are increasingly being referred to as ‘small form sensors’, as whilst their initial costs may be low, they can often be associated with significant ongoing costs past initial investment.

The initial device costs of a low-cost sensor in the current market usually have a range of between £150 and £5000 per device, which is at least an order of magnitude cheaper than the regulatory equivalent. However, in addition to these prices, units may require subscription fees; maintenance fees, staffing fees (for installation & maintenance, as well as data quality assurance and control) and the effective lifetime of a device may not be as long as a regulatory device. Therefore, whilst the term low-cost is used in this guidance, it’s important to consider that the phrase generally refers to the initial hardware costs and not the total sampling costs which may be associated with such devices.

It is important to carefully consider your project aims and resources and whether low-cost sensors are a practical solution which will support your goals. Low-cost sensors are not likely to be able to monitor pollution to the same resolution as regulatory instruments. If you are wanting to confidently detect a small level difference of concentrations (aka 21µgm^-3 to 20 µgm^-3) then low-cost sensors are unlikely to be suitable. They are more likely to be suited to providing coarse insights into concentrations, and give an idea if an area has ‘higher’ or ‘lower’ concentrations.

If after these considerations, low-cost sensors seem suitable there are a wide range of options available for purchase. The simplest devices offer raw air quality data from a low-cost sensor stored on an SD card or downloaded via internet connection. Some other providers offer sensing as a service, offering QA/QC, online data storage and visualizations and in some instances, sensor install and maintenance. When procuring such sensors, there are five key considerations that should be made:

1. Are low-cost sensors the best approach for me? Are indicative values enough for my application? Consider the length of sampling, whether you are targeting a specific source or intervention, if the reduced accuracy of low-cost samplers to regulatory instruments is enough for your monitoring goal and the practicality of installing low-cost sensors including what infrastructure, power and permissions would be needed for a successful install and who owns the data/device/infrastructure.
2. Cost: is the sensor really low-cost beyond the short term? Consider how many pollutants you want to measure and what is included in the price of your sensor. Whilst many low-cost sensors have reduced capital costs compared to regulatory sampling, they also may have ongoing costs such as maintenance, subscription fees, data storage and comms. Enquire about the lifetime and expected ongoing costs of a device and consider if hiring a unit is an option.
3. How is the device hardware different from the others on the market? Whilst there are many different commercially available sensing units, many of the devices actually operate the same sensing hardware to detect air quality concentrations. It is worth asking what the sensing hardware components within a device are, and if the sensor operates on a cartridge system which could allow them to be easily replaced if needed. Often the main difference in cost from such devices comes from the service and software provided, not the hardware!
4. Data quality: How are the devices calibrated and how long does that last? A frequent concern of low-cost sensors is data quality and a common way to address this is to use a calibration against a reference instrument, with the best calibrations also taking into account pollutant source/composition and meteorology. It is important to understand how a sensor is calibrated, how long a calibration lasts and how sensors compare between themselves to support any data analysis.
5. How transparent is all this? Whilst manufacturers can offer sensing as a service, they may not disclose full details of how they calibrate and offer QA/QC of a sensor. From a scientific and regulatory monitoring perspective, the metadata around sensor accuracy, precision and selectivity is key in supporting any data analysis and decision making. Without transparency, data is devalued and we recommend using a range of sources, not just manufacturers statements, to assess the potential of low-cost sensors.

Check List: Before procuring low-cost AQ sensors it is best practice to consider the purpose and goals of the network. Consider:

• Is this a long-term or a short-term network?
• Is the project aiming to capture a specific AQ intervention or source?
• Could the network act as an alternative method of screening for exceedances of AQ objectives? Whilst a low-cost network may not confidently report a level difference of 20 to 21, it could report at coarser resolutions and for example give an idea if an area has a concentration of 20μgm³ or 100μgm³.
• What locations are available to install the sensors? For example, what street furniture is accessible? And are special permissions necessary to install with the local Council?
• Do these locations have power (if required by the sensor)?
• Will specialist equipment such as cherry pickers be required or can the installations be done safely by ladder?
• Is the installer fully insured for public liability?
• Will the installer be able to provide full risk assessments and seek necessary permissions from Local Authority Highways if footpath closures are required?
• How vulnerable are locations to tampering from the public? Higher install heights can reduce this risk but can then make installations more challenging and reduce relevance to public exposure.
• Does this project provide opportunity for education and community engagement resources?
• Would a % or ± error or limit of detection of a sensor impede on the sensors ability to deliver the project aim? (i.e. If a sensor has a limit of detection of 10µgm-3 and you are trying to detect concentrations below or close to the limit then the sensor is unlikely to be effective at getting the results required)

For further guidance regarding procuring and deploying low-cost sensors, you can access our full briefing note which builds on lessons learned from the WM-Air and Birmingham Urban Observatory projects at the University of Birmingham online here