AI’s Role in Tracking Air Quality

Air pollution is a global health crisis and has been linked not only to many diseases but climate change as well, which is another existential threat to humanity. Ambient, referring to outdoor, and household air pollution exposure causes about 7 million premature deaths every year from increased mortality from stroke, heart disease, chronic obstructive pulmonary disease, lung cancer, and acute respiratory infections.

“Poor air quality also has enormous economic impacts, increasing medical costs, reducing workers’ productivity, and damaging soil, crops, forests, lakes, and rivers” according to the latest report by the European Environment Agency (EEA). Although people often have an idea of air pollution having peaks and episodes, which is true of course, long-term exposure to lower doses is an even more serious threat to human health and to nature.

Carbon dioxide may be the main cause of global warming and climate change, but it is not the only cause. Many other gases or particulate compounds affect the amount of solar energy, including heat, that the Earth can retain. Methane, for example, is a very powerful climate forcer and an agricultural air pollutant closely linked to livestock and meat consumption. Particulate matter is another pollutant, impacting both climate change and air quality. Depending on its composition, it may have a cooling or warming effect on the local and the global climate.

Air pollution and our health

We take thousands of breaths a day, yet, in many parts of the world clean air feels like a luxury item nowadays. According to the World Health Organization, more than 90% of the world’s urban population live in areas where air quality levels exceed safe recommended limits. “Estimates of the health impact of exposure to air pollution indicate that in 2018 long-term exposure to particulate matter with a diameter of 2.5 μm or less in Europe (including 41 countries) was responsible for approximately 417 000 premature deaths, of which around 379 000 were in the EU-28” according to the report by EEA. What a fabulous world it would be to be able to decide if you are prone to air pollution and its above-mentioned health risks. Unfortunately, this is not possible. However, with the use of artificial intelligence, we can now predict air pollution better than ever and we can at least minimize our exposure to pollutants and have better control over our health risks. BreezoMeter set out the aim of improving the health and quality of life of people by providing real-time, location-based air quality and pollen information.

BreezoMeter’s Journey

In 2014, BreezoMeter’s CEO & Co-founder Ran Korber, who is an environmental engineer and husband of an asthma sufferer recognized the severity and complexity of the problem and the lack of available solutions. When the couple was trying to look for the city with the least amount of pollution in Israel they had to realize that there is no available information on such features of places. The idea of BreezoMeter stemmed from Ran’s frustration. He believed that people should be able to have some sort of control over their health amidst their polluted environments. Therefore, he set out to create a tool that forecasts 40 pollutants within the categories of pollen, air pollution, wildfires, and weather. He aimed at implementing a technology that enables every single person to have access to information related to air pollution and makes them able to forecast pollution exactly the same way as we forecast weather changes.

The major reason why individuals generally do not have access to such detailed information about air pollution is that most monitoring stations are operated by governments and they are located in specific locations measuring only certain pollutants. Moreover, there is no real-time reporting and forecasting specifically stemming from the previous reason: air quality changes rapidly and at the street level. Last, but not least, many physical sensors can be damaged during certain events, such as wildfires. BreezoMeter’ research scientists wanted to develop a more reliable solution that could help people reduce their exposure to polluted air: by knowing the air quality in specific places, in case of a health risk, they can plan better alternative routes to and from work or even schedule their personal meetings to other locations, change their event venues, and walking and jogging routes to safer, less polluted places.

Detailed information also helps the development of businesses and can be a source of innovative ideas. Additional information on the air we breathe may incentivize industries to create healthier products for their customers: air conditioning manufacturers could develop systems that are responsive to environmental conditions, and healthcare providers could provide early warnings and advice to their customers.

When you consider creating a solution based on artificial intelligence aiming at perfect accuracy, several roadblocks are ahead of you. First, you need to make sure that the data you are incorporating into your analysis is accurate and reliable. Some other mobile applications that aim at delivering air quality reports to users actually rely on unverified sources from individual reports or satellites. The latter one does not provide timely information and can often be influenced by atmospheric factors, such as clouds.

The research and development team behind BreezoMeter has experience in areas of atmospheric science, turbulent flow, and convection and essentially led to the application of scientific principles and mathematical modeling to define how air pollution algorithms should behave. Since BreezoMeter’s aim was to provide the most accurate information possible, they had to have huge computational power due to atmospheric circulations.

The technology at the heart of BreezoMeter

BreezoMeter calculates the dynamic nature of air pollution dispersion between the monitoring stations. In addition to relying on the information provided by government monitoring stations, BreezoMeter factors in information from multiple data sources. In total, the technology relies on more than 14.000 sources around the world relying on more than 47.000 sensors. “Information from monitoring stations, low-cost sensors, satellites, meteorological data, live traffic, land cover information and more are combined to increase the accuracy of prediction, together with air quality dispersion models” – stands on the website. All the data are validated and tested by different methods in order to see whether there are any data points that do not fall within the normal trend. The combination of different sources of information allows BreezoMeter to validate data by comparing them. Then, having the dataset, BreezoMeter uses physical equations, statistical analysis to track the evolution of air pollution. The calculations result in information on air pollution on an hourly basis and at the street level.

The team at BreezoMeter continuously eliminates prediction errors and reports on them. For wildfire events, there is a smoke model incorporated into the reporting. Data also comes from traffic data providers every 12 minutes and machine learning algorithms which take into account the parameters and geographical location to calculate the pollution emission of each traffic jam to its local surroundings.

The user experience

There is a grid system where each location belongs to a tiny square. When a user would like to request information about the air quality at their location, they receive the information based on the closest grid point. Since traffic updates have been incorporated into the solution, instead of giving information at the resolution of 500 meters, BreezoMeter can now provide information at the resolution of 5 meters. There is a standardized universal index based on academic research for reporting on air quality which individuals can use to understand air quality wherever they are.

Thanks to BreezoMeter, air quality information with a resolution of 5 meters on pollen, pollutants, and fire data is now available in more than 100 countries.

Propeller incorporating BreezoMeter’s predictions

Propeller by Resmed uses the forecasts of BreezoMeter and tells chronic obstructive pulmonary disease and asthma patients what to do to improve their health. Since Propeller started using the information provided by BreezoMeter, Propeller patients have had around 50% fewer asthma attacks and fewer emergency room visits.

Governments and leaders should take air pollution seriously because it is a threat to humanity on many levels. No matter where we are in the world – regardless of age, race, religion, nationality or origin – we all need to breathe the same air to thrive. Until then, at least there are tools we can use to be informed about the level of pollution.