The climate is changing faster than ever before. Machine learning could be a key understanding how--and to what extent. 

That's according to Karen A. McKinnon, assistant professor at UCLA's Institute of the Environment and Sustainability. During the Conference on Neural Information Processing Systems, which is being held online this week, McKinnon spoke about the ways scientists can use artificial intelligence, in this case machine learning, to make critical observations about the ways that weather patterns changing.

Here are three things machine learning can help scientists do when it comes to climate change, according to McKinnon.

1. Identify cause and effect.

It's one thing to observe that the climate is changing. It's another to be able to understand why. "In climate science, we're always thinking about this high-dimensional data, and we're always wondering about causality, which can be really hard to establish with only an observational record," says McKinnon.

With machine learning, it's possible to go beyond correlation--knowing that the climate tends to be warmer when there's more carbon dioxide in the atmosphere, for example--to causation. This is important to help us understand what's happening as well as what kinds of behaviors to avoid in order to prevent the worst possible outcomes.

2. Make the most of climate data.

Using AI, scientists can take the data they have and use machine learning to fill in the blanks. McKinnon points out that this field--combining physical measurements with AI for climate modeling purposes--is young and challenging, but promising. 

"It takes people who are experts in AI and machine learning and in climate models to really know where these two intersect," she says. "You're kind of getting the best of both worlds in order to improve our simulations and, ideally, predictions of climate change."

3. Help us understand climate change at the local level.

Climatologist Syukoro Manabe won the Nobel Prize in Physics 2021 for his wide-scale, computer-aided climate change modeling. Harder to forecast, though, is what climate change will look like on what McKinnon refers to as the human scale.

"If you think about trying to predict what climate change will look like in your city or county or state, sometimes we can't go from our theory to those types of predictions," she says. When combined with statistics, she says, machine learning can help give us a better idea of what climate change will look like in a given location, from temperature to humidity to water level. This will allow cities and other locales to more effectively prepare for the future.