Modelling air pollution with AI

Last week I got part way into my bike commute when I was hit by the weight of the smog. We had an air quality warning in the city for a number of days, but in some areas it was painful to breathe and in others it wasn’t noticeable. Urban air pollution is one of the downsides of our car dependent transportation grid. Aside from the contribution to climate change there are also major health concerns. Landrigan reports that 4.2 million people died prematurely from ambient air pollution (as cited in Desai, Tayarani, and Gao, 2022).

Having more refined models of air pollution in urban spaces means being able to do things like create specific maps of which neighbourhoods and even streets have poor air quality. Desai and colleagues (2022) propose that this has the potential to impact environmental justice. It does. Tracking where a company’s pollution goes can potentially make it easier to track and fine heavy polluters. It can also impact urban planning as we learn more about how urban infrastructure impacts air flow and air pollution. It can also make neighbourhoods with good air quality more expensive than those with poor air quality. This already happens of course but having a map to prove it will only make it more significant.

But getting such refined models has been challenging. As anyone who has been surprised by the weather even after checking the weather report knows, predicting and modelling what air is doing is hard. This is where artificial intelligence (AI) comes in. The authors tested different models using AI to fill in gaps in information. This actually resulted in more accurate models without as much data as what is currently required. It also required fewer modelling steps.

Obviously, my hope as a cyclist is that we change our transportation grids to reduce cars and increase modes like cycling, walking, and even public transit. Can having better models contribute to this change? I hope so.