AI model gives air quality forecasts that are similar to weather forecasts

By simulating air movements and pollutant dispersion, AI is set to enhance the accuracy and resolution of air quality forecasts.

Researchers from APL and the NOAA are harnessing the power of artificial intelligence (AI) to revolutionize atmospheric modeling.

Wildfires and air quality forecasts

In recent years, the world has witnessed an alarming increase in severe wildfires, driven by rising temperatures and prolonged droughts.

These fires are devastating to the environment and pose a significant risk to human health due to the smoke-borne contaminants they release.

The year 2023 marked a particularly grim milestone for Canada, experiencing its worst wildfire season on record, with a staggering release of over 290 million tons of carbon into the atmosphere. Similarly, California faced unprecedented fire seasons in 2020 and 2021.

The consequences of these wildfires extend far beyond their immediate vicinity.

For instance, smoke from the Canadian wildfires drifted across the Atlantic, affecting air quality as far as Portugal and Spain and triggering alerts across North American cities.

Millions were subjected to the adverse effects of this pollution, suffering from symptoms like stinging eyes, stuffy noses, and labored breathing.

The National Institutes of Health underscores the severity of this issue, estimating that air pollution contributes to 6.5 million deaths globally each year.

AI to the rescue: New Era in air quality forecasting

Recognizing the gravity of this situation, Marisa Hughes, a leading figure in climate intelligence at the Johns Hopkins Applied Physics Laboratory (APL), emphasizes the challenge in quantifying the slow, cumulative impact of deteriorating air quality.

“We know that dangerous air quality levels are a significant threat, but because exposure happens slowly, over time it is more difficult to quantify,” said Hughes, who is also the assistant manager of the Human and Machine Intelligence program.

“A more accurate, higher-resolution model can help protect populations by providing them with information about air quality over time so that they can better plan ahead.”

In response to this need, researchers from APL and the National Oceanic and Atmospheric Administration (NOAA) are harnessing the power of artificial intelligence (AI) to revolutionize atmospheric modeling.

Pioneering forecasting techniques

By simulating air movements and the dispersion of nearly 200 different pollutants, AI is set to enhance the accuracy and resolution of air quality forecasts.

This is particularly crucial considering the complexity and computational intensity involved in traditional weather forecasting methods, which have to account for a myriad of variables and interactions in the atmosphere.

Jennifer Sleeman, a senior AI researcher at APL, highlights the computational challenges in current forecasting methods, where a significant portion of the computation is dedicated to tracking pollutant movement and their chemical interactions.

“In our case, the models are looking at the movement of nearly 200 different pollutants in the atmosphere for every timestep, sequentially. That’s approximately 40% of their computation,” said Sleeman.

“And then they also have to consider how these chemicals are interacting with one another and how they’re decaying — the chemistry is approximately 30% of the computation. It takes a significant amount of computing power to perform air quality forecasting with all of the variables used.”

The AI-assisted approach developed by APL is a game-changer, streamlining the process by employing deep-learning models to simulate ensembles using fewer, shorter timesteps.

More accurate and less time-consuming

This innovative method is not only more efficient but also more practical, given the limitations in computing resources and costs.

The traditional ensemble modeling technique, which requires running multiple variations of models, becomes significantly more feasible with AI assistance.

Sleeman notes the tremendous computational savings achieved with this approach, enabling quicker and more accurate forecasts.

“Running one model is computationally challenging — so imagine running 50-plus models. In some cases, this is just not feasible due to cost and computing availability,” said Sleeman.

Collaborative efforts and real-world applications

In collaboration with Morgan State University, NASA, and NOAA, APL researchers have successfully applied this model to NASA’s GEOS Composition Forecasting (GEOS-CF) system.

This system now generates five-day global forecasts with remarkable precision, a significant improvement over traditional models.

“NASA and NOAA have been searching for ways to increase the resolution of these forecasts,” said Hughes. “If you live next to a power plant or a highway, the air quality impacts are going to affect you differently.”

The AI-driven model, trained on a year’s worth of simulation data, has proven capable of producing 10-day forecasts that closely align with actual ground data, requiring only a fraction of the input data traditionally needed.

“The amount of computation we could save with our networks is tremendous,” said Sleeman. “We’re speeding things up because we’re asking the models to compute shorter timesteps, which is easier and faster to do, and we’re using the deep-learning emulator to simulate those ensembles and account for variations in weather data.”

The future of air quality forecasting

Both Hughes and Sleeman acknowledge the crucial role of the broader AI community in these advancements.

“If we tried the same thing five years ago, it might not have been as successful as it is today, because we’re building on the momentum of this accelerating research,” said Hughes.

“We’re sharing our results and starting to see what methods and architectures are effective when you apply them to different problems around the world.”

This project is part of a series of initiatives exploring AI applications in climate intelligence, aiming to tackle challenges like forecasting climate tipping points.

These efforts are pivotal in APL’s mission to ensure climate security, showcasing the potential of AI in addressing critical environmental issues.

In summary, the integration of AI in atmospheric forecasting marks a significant stride in our ability to predict and mitigate the impacts of air pollution from wildfires.

As we face the increasing challenges of climate change, such technological advancements offer a beacon of hope, equipping us with the tools necessary to safeguard our environment and health.

This study was presented at the Association for the Advancement of Artificial Intelligence’s Fall Symposium, at the American Geophysical Union Conference and to the American Meteorological Society.

—–

Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates.

—–

Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.

—–