Concordia-led research project receives $5M to transform weather forecasting in Canada

As severe weather events intensify and rack up societal costs, knowing what weather is coming — and having enough time to prepare — is increasingly important.

That’s where a new $5 million NSERC Alliance Society grant over five years comes in. The funding will support WxAlliance, a Concordia-led research collaboration that is developing a new generation of weather forecasting models by combining traditional atmospheric science with machine learning.

Led by Brian Vermeire, associate professor in Concordia’s Department of Mechanical, Industrial and Aerospace Engineering, WxAlliance brings together nine academic institutions and eight societal partners from across Canada. Together, they bring expertise spanning meteorology, emergency management, wildfire response, power grids, computing, insurance and climate adaptation.

Two approaches, one forecast

Traditional weather forecasting relies on numerical weather prediction (NWP), in which powerful computers use equations describing the atmosphere to predict how weather systems will evolve. These models are grounded in physics, but they are computationally expensive and limited in how finely they can capture small-scale events such as thunderstorms and sudden downpours.

Machine learning weather prediction (MLWP) takes a different approach. Instead of relying primarily on atmospheric equations, these models learn from decades of historical weather data. Once trained, they can produce forecasts in minutes using relatively modest computing power.

Neither approach is perfect. Machine learning models can be extremely fast and identify patterns that traditional models may miss, but they can also behave  less predictably and may not always respect the physical laws governing the atmosphere.

WxAlliance will bring the two approaches together.

“If we have parts of the system we understand really well, let’s use the classical technique there,” says Vermeire. “If there are parts of the system we don’t understand, let’s use the machine learning techniques to sort of fill those gaps.”

The goal is to produce forecasts that are faster, more accurate and capable of capturing finer-scale weather events, while reducing the computing resources required by current systems.

That could ultimately help governments, emergency responders and other organizations prepare more effectively for extreme weather.

“Brian is bringing together an exceptional team to tackle a challenge that matters to Canadians,” says Tim Evans, Concordia’s vice-president of research, innovation and impact. “WxAlliance has the potential to advance how we predict severe weather and strengthen Canada’s ability to prepare for its impacts.."

Vermeire adds: “To my knowledge, this is the first research grant of this scale in Canada to use machine learning for weather prediction. I see it as really a launch platform for Canada to establish itself as a leader in this sector.”

Learn more about Concordia’s Department of Mechanical, Industrial and Aerospace Engineering.