Dr Juan Bernabé-Moreno. Image: IBM
As part of its annual presence at COP28, IBM has announced that it will expand its collaboration with NASA to work on a new, separate AI foundation model for weather and climate.
As climate leaders gather at COP28 from today (30 November), IBM and NASA announced a new AI foundation model for weather and climate, which goes beyond their initial commitment to build and deploy a geospatial foundation model.
By applying its AI technology, IBM says the new model aims to improve the accuracy, speed and affordability of weather forecasting and other climate applications. As well as forecasting, sample applications of the model include identifying conditions conducive to wildfires and predicting extreme meteorological phenomena. IBM says its researchers will work alongside NASA domain experts to train and validate the new model.
Earlier this year, IBM and NASA set out to build an open-source geospatial foundation model, to gain new insights on Earth’s climate through the power of AI technology, combined with the abundance of Earth observation and geospatial data that NASA has gathered.
The goal of this partnership was to provide an easier way for researchers to analyse and draw insights from these large datasets for climate research. And sure enough, back in August the two partners announced they would release the large geospatial foundation model on open-source AI platform Hugging Face to make it widely available to scientists and researchers, saying at the time that it would be the largest geospatial foundation model on Hugging Face and the first-ever open-source AI foundation model built in collaboration with NASA.
As well as using the model to estimate the extent of past floods and wildfires, IBM is also using the model to help map urban heat islands in the UAE, track reforestation in Kenya and track climate resiliency in the UK.
Now, encouraged by these results, IBM and NASA have decided to branch out and build a new foundation model aimed at making weather and climate applications faster, more accurate and more accessible. It says potential applications include helping climate experts infer high-resolution information from low-res data, identify conditions conducive to wildfires, and predict hurricanes, droughts and other extreme events. When complete, this model too will be made open source.
Open access is key
The open access to these models is key if it is to help tackle extreme weather events and other climate challenges, says Dr Juan Bernabé-Moreno, the director of IBM Research Europe for the UK and Ireland, who also has responsibility for the accelerated discovery strategy for climate and sustainability at IBM.
Bernabé-Moreno was previously chief data officer and global head of analytics and AI at Eon, leveraging data and algorithms to support the energy transition. He is greatly encouraged by the success of the initial NASA collaboration.
“We found that it was not just economies of scale we were achieving by performing many tasks more quickly, but the tasks we were doing were performing better,” he says. “That gave us a lot of encouragement.
“So, we proceeded from small tasks where we looked at whether an area had been affected or not by fire or by flood, and started on more complex tasks like understanding land use and its impacts over time, and it worked extremely well.”
He cites the work done in Kenya where, in December 2022, its president William Ruto unveiled the National Tree Growing and Restoration Campaign, which aims to plant 15bn trees across Kenya by 2032, including in areas of critically affected water towers (forested landscapes that retain water and source many rivers in Kenya). Deforestation has been contributing to increasing water scarcity in these regions.
IBM has an agreement with the government there to support the initiative and leverage IBM’s geospatial foundation model to enable users to track and visualise tree planting and tree growing activities in specific water tower areas, to monitor forest restoration and measure above-ground biomass like sequestered carbon, with the ultimate aim of mobilising local efforts to plant more trees across Kenya.
“We measured the above-ground biomass at two different points of time to compute the difference, and then applied a multiplicator to understand how many gigatonnes of CO2 they are in a position to capture. So, we made climate mitigation measures quantifiable,” says Bernabé-Moreno.
“We need to be able to quantify how successful these measures are because it will encourage people to do more, because they will now have the proof. We did something very concrete, very local, it made a difference, and we can quantify the difference. So that’s what encouraged us to do more, and to do something as difficult as weather and climate.”
Bernabé-Moreno says the open-source nature of the models is crucial. “It’s vital to support open source. Yes, these kinds of models around weather and climate, these are innovations, and we will use these innovations to differentiate our product, but instead of keeping it for ourselves, we believe that the community is going to help us make the simulations even better.
“If you think about the foundational model, it is not a solution, it is a capability. Now imagine putting that in the hands of the community and having scientists create their own applications. Imagine that these scientists start tracking sustainability measures done locally, suddenly you have the whole community quantifying sustainability progress by a particular initiative or government, it changes the game.”
When it comes to climate change mitigation, the more we can do, and the sooner we can do it the better, and this is where AI can really make a difference, says Bernabé-Moreno. “The technology we’re developing and the open science spirit, they will help us measure things, and inform climate change strategies. Instead of trying everything, measuring nothing and hoping we are doing a good job, we can be way more targeted. I believe that’s the big role that technology has to play when it comes to climate change.”
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.
