Researchers have developed a new artificial intelligence-based system for forecasting ozone that they say would allow ozone level predictions 24 hours in advance.
The system was developed by staff at the University of Houston’s Department of Earth and Atmospheric Sciences who say it could be used to give advance warning of high ozone rates.
“If we know the conditions of today, we can predict the conditions of tomorrow,” study co-author Yunsoo Choi told Science Daily.
While providing a protective layer against solar radiation, ozone is also associated with respiratory problems, especially among people with asthma, the elderly and young children. Car emissions and pollution from industry are contributing factors in higher ozone levels.
The model extracts data using convolutional neural networks – networks capable of combing through data intelligently.
To test the artificially intelligent model, the researchers used weather and air pollution data from 21 weather stations in Houston and elsewhere in Texas representing conditions between 2014 and 2017. They programmed the convolutional neural networks using the weather data to predict ozone levels for the following day, and then cross-referenced the forecasts with the historical ozone data to verify the accuracy of the system.
They found the model’s accuracy rates were up to 90%, though the researchers said this rate is likely to increase as the model continues to learn.