The autonomous meteorologist is developing advanced air pollution forecasts for Airly.
The forecast of air pollution is a very difficult conceptual and computational task, requiring considerable research and simulation expenditure.
Many research centers and commercial companies are working on effective procedures to try to indicate the state of air pollution in the time to come. Unfortunately, most of them determine only the average daily level of pollution the next day.
On map.airly.eu you can see the pollution forecast specified in the CAQI index. What is more, the air quality prediction includes estimations of the state of pollution for the next day, hour after hour. Airly’s forecast accuracy remains at + 80%.
* the chart presents actual and forecast data in January 2019 for the city of Częstochowa, Poland
In order to implement this task, innovative algorithm solutions in the field of Machine Learning were used. The currently used predictive algorithm is based on methodologies of artificial neural networks, whose task is to predict the state of air pollution based on data from the past and estimation of forecasts of selected meteorological parameters (such as wind strength, temperature or air humidity).
As explained by Piotr A. Kowalski PhD – an expert in machine learning at Airly: “artificial intelligence translates directly into the possibility of applying ever new structures of neural networks in areas such as ecology, which by its nature is rather far from computer science. The synergy of such different fields is able to meet the challenges that are very close and necessary to people both on a daily basis and in a much longer time horizon”.
To check the current air quality in your area visit our online map HERE.