“Machine learning is one of the most important tools we have for advancing climate research,” says Michael Felsberg, professor of computer vision.
“We talk a lot about AI today, and machine learning is the driving technology behind AI. It’s about teaching computers to solve complex tasks and drawing conclusions from large amounts of data,” says Fredrik Lindsten, senior associate professor of machine learning.
Together with two climate researchers from Lund, the two LiU researchers are responsible for the so-called focus period this autumn, where researchers with different competencies in climate and machine learning can meet. The focus period is arranged within the framework of ELLIIT, the strategic research environment in IT and mobile communication, which is run by four universities in close cooperation with industry.
Surveillance from space
For global climate research, remote sensing – measurements from satellite and aerial images – plays a key role. Hyperspectral cameras on satellites (a powerful technology that allows the collection of highly detailed information) are used to draw conclusions about environmental impact. There are also drones that take aerial pictures.
“It’s difficult to obtain global coverage using only measuring stations on the ground. But in this way, we can monitor, for example, forest fires across the globe 24/7,” says Michael Felsberg.
Fredrik Lindsten’s research focuses on weather and climate forecasts. One question he is interested in is how we can build machine learning models that are much better than today at dealing with uncertainty in forecasts.
“For example, what is the probability that there will be a downpour? When it comes to extreme weather, it’s important to be able to deal with uncertainty in order to make the right decisions and prepare the community.”
Such an open research question can be presented to the visiting researchers participating in the focus period, who will use an interdisciplinary approach. One of the aims of focus periods is to gather different competencies under the same roof.
“Machine learning in climate research is very broad. Our focus period brings together an unusually large spectrum of competencies and problem areas compared to similar initiatives elsewhere in the world,” says Fredrik Lindsten.
Interdisciplinary meetings
Tove Kvarnström is the coordinator for the focus periods:
“They revolve around an interdisciplinary field and last for five weeks. Some visiting researchers come here at the beginning or end of the period, others stay for the whole period. In the middle of the focus period, we have a symposium where they all participate and it’s also open to all researchers at LiU.
The symposium is not like any other scientific conference,” Tove Kvarnström explains.
“The limited number of participants makes it possible to talk to everyone and we make sure there’s time for that. You can talk to someone super senior in your field at lunch. The informal meetings are important and the feedback we get is that the participants are really happy with the setup. We also follow up on the focus periods in the long term and know that they have led to new collaborations.”
AI has an important role
Michael Felsberg is a highly ranked AI researcher and believes that artificial intelligence is incredibly important in climate research.
“AI gets a lot of negative publicity, but in climate science, AI is perhaps the only chance we have to really get a grip on the problem. I hope that the focus period will lead to new, long-lasting collaborations,” he says.
“My research can be used in many areas where there’s strong interest from the industry, such as autonomous vehicles. In climate research, there are not many commercial actors and therefore I feel that it’s particularly important to engage in this type of application.
Translation: Anneli Mosell