Explore AQMesh

ASIC discussions look at AI vs non-AI calibration

22-May-2024ASIC | Calibration | Hybrid networks | Network calibration | Networks

ASIC discussions look at AI vs non-AI calibration

So how do you do it? Many presentations at the recent ASIC conference revolved around calibration of small sensor air quality systems, including that given by AQMesh Technical Business Development Manager, John Downie. Offered the opportunity to poll the ASIC audience, we chose to ask delegates ‘Please indicate which of these calibration methods for small sensors you would consider for your small sensor network’. Whilst numbers responding were not huge, ‘periodic co-location of each instrument with FRM/FEM creating seasonal correction factors’ came out top, indicating a thorough – if labour-intensive – approach is generally most common and taken as the ‘norm’.

So what if you don’t have the ‘boots on the ground’ to carry out the many instrument movements that would be necessary with a network of any size? Is there a short-cut? Can AI help? The poll did show that about a quarter of respondents were considering ‘desk-based calibration, using machine learning or AI’, about twice the proportion thinking about ‘desk-based calibration using network calibration method without machine learning or AI’.

We tried using AI on electrochemical sensor output a few years ago but found that – at the time – the machine learning was very good at latching onto interfering factors, such as temperature, and less good at finding the weak (valid) signal in amongst the noise. We were greatly impressed by the later (non-AI) network calibration work done by University of Cambridge, our partners in the Breathe London pilot, and have developed our own approach. This series of repeatable calculations can be applied to a network of five instruments upwards and we feel there are some clear differences (advantages, of course) comparing the AQMesh approach to local calibration with AI-driven approaches:



AI / Machine Learning

Confidence in repeatability of measurements from individual instruments



Confidence in repeatability across seasons & locations



Traceability back to only self-contained* measurements



Calibration method repeatable with the same output



Drift correction / interpolation between calibration intervals



Method approved by DEFRA / Environment Agency



The fundamental issue with machine learning and AI being used either for compensation of sensor interferences or for calibration adjustments is that their whole premise is to “evolve”: they spot differences and change to account for those changes. This means that over time the results may be closer to reference, but they will never follow the same process for adjustment as they did previously. This raises significant traceability questions about when processing changes and makes it impossible for these methods to become certified to a standard which require a fixed processing of inputs, managed by a version number.

It’s worth noting that PAS 4023 (Annex D) distinguishes between sensor systems that are ‘self-contained’ – with readings derived by the system alone, using a series of repeatable calculations – and those that require training against a reference station. The PAS also emphasises how important it is that individual sensor systems should perform very similarly to one another, so that remote comparisons across a network can be made with confidence. Another reason why a proven, global correction algorithm used by every instrument has the edge over individual site-based AI. Strong inter-instrument comparability also means that a network can be meaningfully compared against itself, in the absence of a reference station, as offered by AQMesh’s network normalisation option.

Supporting your air quality monitoring system when you can’t get to it

24-Apr-2024Fenceline | Hybrid networks | Industrial | Networks | Product | Support

Supporting your air quality monitoring system when you can’t get to it

Each time we think we have found a spectacularly remote monitoring location, an even more inaccessible spot is reported by one of our users. Full-day trips to visit a location have now been beaten by customers who need to charter a plane to reach them. So, remote diagnostics and support are very important.

Luckily, IoT communications, cloud data management and over 10 years of experience supporting AQMesh have allowed us to continually improve our ability to supply and support AQMesh in remote locations. Pods have been used from the edges of the arctic to undeveloped deserts – as well as on ships – with the help of a few features.

Robust design, low maintenance intervals

AQMesh was designed to be rugged, for use all over the world and with an expected maintenance interval of two years. We have always understood that field maintenance requirements must be kept to a minimum, and pods operating for year after year, in the harshest environments – from deserts to extreme cold – demonstrate design effectiveness. This includes protecting electronics from the elements and mitigating electromagnetic interferences, as well as taking measures to keep insects, wildlife and birds out/off.  The unobtrusive pod design has also ensured a very low rate of vandalism and theft.

QA flags and notifications

The AQMesh data stream includes vital pieces of information which allow users and the AQMesh support team to check that pods are functioning correctly and provide an early warning system. Users can register for email notifications for their pods – it is always better to find out that power is running low or data is no longer being transmitted at the time, rather than when the project ends and it’s time to review data.

Remote scaling / calibration

Whilst AQMesh was a leader in co-location comparison and the ‘gold pod’ technique for in-field calibration, these approaches do require regular site visits to move pods around. We have now developed a method that can provide remote calibration of a sensor network, with or without an available reference station, that does not rely on artificial intelligence.

Diagnostic information

The AQMesh team can access additional diagnostic information remotely, such as performance indicators from the optical particle counter, solar pack battery voltage or sensor failures. Some of these indicators are available to users via their secure online or API access, and some can be used by our global technical support team. The team uses the full range of diagnostic information available, including SIM connection attempts, to provide free support for the life of the equipment. Their over-riding goal is to fix any problem without asking users to visit the site.

Over the wire intervention and updates

AQMesh firmware developments now allow power cycles to be triggered remotely, firmware to be updated over the wire or remote sampling and transmissions interval changed.


We have learned from the many challenges that power supplies can present to remote operation. Whilst the original lithium thionyl chloride battery offered unbeaten long-term autonomous operation of gas sensors, increasing shipping limitations have turned our focus to direct power supply and solar. We invested in a full technical investigation to identify a mains to 12V DC transformer that could cope with ‘dirty’ power supplies, as well as in-pod measures to manage spiky or intermittent power.

Having seen so many problems from simple solar-panel-plus-battery arrangements, we designed our own smart solar pack, which squeezes the most power out of any location, manages power delivery and provides online voltage measurements. We are mindful that sampling and reading rates are defined by the project – and potentially certification – and the power supply must deliver the same sampling throughout the year. Readings should not be compromised by the difficulty of providing autonomous power.


The global SIM supplied with a standard AQMesh pod will roam across networks to find the best connection at each transmission, and has proven to be a very reliable way of transferring sensor output from hardware to our cloud server for over 10 years in more than 70 countries. Occasionally, we find that only a single, specific network is available – or a customer would prefer to use their own SIM – in which case we can programme the pod to work with a locally-sourced SIM contract. To achieve autonomous communication, the AQMesh LTE CAT M1 modem uses the latest LTE communications standard, including support for NB-IoT where available. In the most extreme cases, satellite communication is the only viable option and AQMesh can connect via an ethernet port to a suitable modem to connect this way. Reliable communications are key to remote data access and support.

The growing need for remote, long-term monitoring, in all conditions, drives our continuous development from data QA to comms, and we welcome challenges.

Some projects run more smoothly than others – what makes the difference?

13-Mar-2024Hybrid networks | Networks

Some projects run more smoothly than others – what makes the difference?

We have some thoughts here at AQMesh about the common features of successful, well-run small sensor air quality monitoring projects. This is our list, but we’d love to hear your ideas.

  • One main contact who remains constant throughout the project
  • Clear project objective(s) and timescale, even if the client doesn’t share them with us
  • The main contact requests and distributes relevant information to all the right people – guides, videos, manuals
  • Planned installation locations are reviewed – by the client and by our team – on Google maps and using photos
  • Direct communication between installation staff is actively encouraged
  • Regular data reviews, so any issues are addressed quickly
  • All available resources are used, such as our site technicians’ ‘app’

These points seem to be critical, wherever the project is, and whether it is totally confidential, ‘typical’ or novel.

What would you add to the list for successfully running a small sensor air quality monitoring network?

UK local authority uses AQMesh for cost-saving NO2 monitoring network

28-Feb-2024Accuracy | Hybrid networks | Local authorities | Networks | PerformanceUK

UK local authority uses AQMesh for cost-saving NO2 monitoring network

A UK local authority installed nine AQMesh systems at different points across a busy town, measuring nitrogen dioxide (NO2) at 15 minute intervals, monitoring 24/7. These locations were established monitoring points, where measurements had been taken previously using diffusion tubes, limited to one average reading every few weeks.

AQMesh – in common with all lower cost air quality systems – can provide near real-time air quality information, with high frequency measurements that allow daily and weekly patterns to be seen. However such systems are not certified, as are reference stations or diffusion tubes. As a result, AQMesh readings need to be ‘calibrated’ against certified readings, at some point in the network, to provide confidence in data accuracy and traceability to an approved standard.

Typically such ‘calibration’ is carried out by mounting at least one AQMesh ‘pod’ very close to a reference station, so pod and reference are sampling the same air and readings can be compared. However this approach does require staff to move pods from position to position, which can be time-consuming and therefore costly. An alternative approach was used for this network, similar to the one developed by the University of Cambridge and used in a major project in London (Breathe London pilot). One of the authority’s reference stations (location in red on map) was used to ‘calibrate’ the network of pods and the other (location in green on map) was used to cross-check network accuracy.

AQMesh network deployment (BELOW): AQMesh locations marked in blue, reference station used for calibration in red, reference station used for control co-location in green

The four-month project demonstrated that the AQMesh network showed that stakeholders could have the same high confidence in readings when the network was calibrated remotely as when pods were co-located for calibration (the gold standard for this technology), but with significant savings in field support and reduced data loss.

Do air quality people love holidays more than everyone else?

12-Dec-2023Environmental | Hybrid networks | Industrial | Networks | Traffic

Do air quality people love holidays more than everyone else?

Everyone loves holidays, whether Christmas or anything else, right? So what’s special about ‘air quality’ people? What we get so excited about are ‘free’ experiments, where distinct changes in activity help to peel away the layers of air pollution measured. Over the years, various studies have been published, showing residual air pollution levels when other sources drop – or increase – significantly.

Around this time of year there are changes in emissions activity around schools, businesses and industry, roads (both increases and decreases), burning of solid fuel in households, domestic heating, and so on. As well as looking at changes in measurement over time (hourly or shorter intervals) and space (hyperlocal monitoring means you can literally measure at any point you wish, from a specific point on a specific road junction to a school playground), measurement of multiple parameters is an eye-opener.

Studies by the University of Cambridge have shown how small sensor air quality measurements can be used in conjunction with their scale separation technique to distinguish between local and regional or background sources. Comparing the proportion of different pollutants in this way can give a ‘fingerprint’. CO2 measurements provide a baseline combustion level against which generally traffic-related NO / NO2 / NOx can be compared. Looking at PM fractions against CO2 and other gases can also provide more insights than individual measurements alone can provide. And, of course, dramatic shifts over time – like holidays – sharpen that focus.

A network of sensor systems has the additional benefit of showing whether pollution is being displaced from one location to another, with this forming part of the analysis around other behavioural change triggers, such as the introduction of a traffic Low Emission Zone (LEZ). It can also help identify hyperlocal sources of pollution, where high levels of pollutants are only seen by one of the monitoring points.

One memorable headline from several years ago was that a higher amount of PM2.5 in one London borough over the Christmas period could be attributed to domestic solid fuel combustion (cosy wood-burners) than road traffic. So, whether it is reduced traffic around schools, increased traffic at shopping centres or chestnuts roasting on all those open fires, the holidays can provide a curious insight to local air quality data and pollution patterns.

Happy holidays from the team at AQMesh.

A few thoughts about the WHO’s report on AQ methods

21-Nov-2023Data validity | Hybrid networks | Modelling | Networks | WHO

A few thoughts about the WHO’s report on AQ methods

This a great report, full of common sense and helpful advice, applicable globally, particularly where air quality monitoring is currently limited. Limited by budgets, logistics, and many factors discussed so eloquently at the ASIC Ghana conference last month.

Central to the report is the position that “air quality monitoring is the first step for understanding a population’s exposure and taking action” and the report lays out clearly the monitoring methodologies available, from passive sampling to modelling.

Low cost sensors are put in their place – but very much given a place – offering the cheapest and simplest way to understand spatial and temporal air pollution variability. To the report’s plea that “ideally, every nation should have access to at least one reference-grade monitor” we would add “well-maintained”. It can be a huge challenge maintaining a reference station in many parts of the world – accessing calibration gases, skills, parts and budgets – but, as the report says, they do indeed open the door to many other air quality methods. We have been involved in many projects where our pods came out of the box and performed well, but that the project ‘failed’ because the objective was comparison with reference standard and that could not be achieved because the reference station could not be brought up to spec within the project timescale. Valuable real-time information, showing minute-by-minute, hourly, daily and weekly trends in air quality across a target area, can be discarded for this reason, which is a shame.

Multiple monitoring methods, or hybrid networks as we refer to them, used effectively, can most effectively achieve the baselining and measurement of progress discussed in the report. It was difficult to fully recognise the generic Low Cost Sensor described in the report, but that is not too surprising, given that the sector is so dynamic. Sources referred to, dated 2017, would be based on fieldwork from preceding years, and an awful lot has changed in Low Cost Sensor world in the last six years or so. Are they really low capital and medium operating costs? That may be true of some, but there are also slightly more expensive options with lower operating costs. We genuinely do not see drifting baselines, but that’s maybe because we have never rated metal oxide sensors. Also, since a step change in the electrochemical sensors making their way into the market in 2016, there has been continuous improvement in temperature correction of electrochemical output, and temperature effects did look like drift over certain time periods.

We would certainly agree with the comment that low cost sensors are “not yet suitable for replacing reference-grade monitors” but the remote calibration techniques and diagnostics offered by the latest generation of systems goes a long way to address the highlighted need for data quality assurance and quality control.

AQMesh keeps essential air quality data accessible during lockdown

22-Apr-2020Hybrid networks | Networks | Smart cities

AQMesh keeps essential air quality data accessible during lockdown

Technology is critical to so many essential services during the current global COVID-19 crisis, but it is also allowing local air quality to continue to be monitored, in real-time, across the world.

Small sensor air quality monitors such as AQMesh pods, which can be easily mounted on lamp posts at road junctions, near schools or around industrial sites, can use cloud data storage to ensure that air quality information is stored and accessible even when staff are not able to visit equipment. With field staff budgets cut, air quality monitoring that does not require site visits is often preferable but in the current situation it is essential. Some small sensor systems require staff to visit the equipment to download data, which is not only a hidden cost but is currently a huge disadvantage given the restrictions on movement.

The small, low-power AQMesh pods can house sensors which measure a wide range of pollutants and environmental conditions in a single, compact unit. For example, AQMesh measures common pollutants, including those identified by the World Health Organisation as significantly damaging to human health: gases NO, NO2, O3, CO and SO2 and particulate matter PM1, PM2.5, PM4, PM10. Additionally, AQMesh records environmental measurements temperature, atmospheric pressure, relative humidity and has options for wind speed and direction, plus more specific measurements for different applications: gases H2S and CO2.

AQMesh uses the cellular phone network via a globally-enabled SIM card to send sensor output to a the secure AQMeshData cloud server.  Here, sophisticated proprietary processing calculates air quality readings. The processing algorithms used are the result of nearly 10 years of analysing comparisons of AQMesh readings against readings from co-located equipment that complies with internationally recognised standards. These co-located comparison datasets are the reason for the high level of accuracy that AQMesh offers when used anywhere in the world, with no other manufacturer having access to such a broad range of comparable data and provable results. From the AQMesh server, users can securely access their air quality readings either by a website login  on their laptop or phone, or by programming an API connection so data can be called automatically and integrated into their own system. Air quality data from AQMesh pods is not published publicly unless the data owner – or equipment purchaser / renter – chooses to do so, as data security is a crucial factor for AQMesh customers.

As well carrying out air quality analysis, air quality readings can be linked to automated pollution exceedance alerts, such as around road tunnels in the south of France, to display air quality to the public, such as the Breathe London project, or to manage air quality around industrial sites, such as mining facilities in South Africa.

Whilst the Internet of Things (IoT) and Smart City integration of sensors has helped to drive the growth of small sensor air quality networks it is important that traceability and data quality are retained. Air quality measurements are recognised as more challenging to achieve from small sensors than other measurements, particularly pollution gases, such as NO2 and O3, which are unstable in ambient air. Other systems have chosen to use artificial intelligence or machine learning to overcome these challenges, but it does not provide data traceability and integrity in the same way as the fixed version processing calculations used by AQMesh. 

Luckily the chain of smart technology does not stop at the AQMesh server and many AQMesh customers are continuing to monitor air quality via their pods during these unprecedented times, which are undoubtedly having a dramatic impact on air quality around the world.

Amanda Billingsley, CEO of AQMesh manufacturer, Environmental Instruments Ltd, comments “We are as busy as ever, providing online support to our users. We continue to carry out remote support and proactive remote diagnostic work and data quality assurance for all our customers around the world, who are themselves able to review air quality data and take appropriate action whilst working from home. Additionally, our unique AQMesh pods were always designed to be as practical as possible and so they require very little ongoing maintenance – we simply advise replacing the electrochemical sensors every 2 years. This is another reason we are able to keep AQMesh networks up and running across the world during this crisis.”