Smart cities will have the opportunity to tackle life-threatening traffic pollution by predicting and acting upon vehicle flows following just a few days’ worth of data, thanks to cutting-edge pollution prediction tools developed by GeoSpock.
According to the World Health Organization’s World Global Ambient Air Quality Database, more than 80% of city dwellers across the globe had to endure outdoor pollution levels that exceeded health standards in 2016. The latest figures also suggest that air pollution causes more than three million deaths worldwide every year.
However, by crunching hourly vehicle counts, as well as congestion, pressure, temperature and humidity measurements, GeoSpock has been able to predict accurately air quality levels on Gonville Place, a main road in Cambridge, at specific times on a Friday and Saturday after processing a large amount of data from only the preceding five days.
How do we track air pollution?
Utilizing a range of different Machine Learning techniques, good predictions were obtained for the actual trajectory of emissions over the two days, based on only short-term data gathering and categorization.
The implications for cities that struggle to predict pollution and therefore control vehicle emissions and pollution levels could be huge.
Visualizing predictive emissions data at such short notice will give cities the opportunity to adapt traffic controls swiftly to mitigate harmful emissions as they pack in more people and vehicles than ever before.
In Cambridge, there has been a 4% year-on-year increase in traffic volume, with 50,000 workers travelling into the city alone and a total of about 206,000 vehicles coming in and out of Cambridge every day.
The average journey time in the peak hour slipped to a sluggish 4.87 minutes per mile in 2015-16 – significantly worse than the average of 4.45 minutes per mile recorded the previous year.
However, the data processed by GeoSpock, has shown that the traffic count only tells part of the story.
It’s not just traffic
“Although the number of cars passing a location is a factor, the data reflect the fact that the traffic count alone is not sufficient to estimate air pollution. As well as weather/climatic factors, it is the density of traffic – how tightly the cars are packed (a function of both the traffic count and congestion) – which is much more fundamental when predicting air quality,” GeoSpock Machine Learning Engineer Alan Roberts said.
Given GeoSpock’s capacity to handle huge volumes of data, cities will be able to establish an even more precise vision for the future by introducing data that stretches further into the past, helping to illustrate trends over a longer period of time.
“Of course, the longer the data collection periods, the more accurate the predictions are likely to be, but our work has shown that it only takes a few days of data gathering to enable us to gain insight for a smart city to act upon,” Roberts added.
“It’s up to the policy-makers to decide what should be done with this information, but we can provide them with all of the tools to see what is happening.”
If you’d like to learn more about our how our technology impacts smart cities, check out this page here.