ATL Standard

Exponential growth in big data and computing power is reshaping climate science, with machine learning playing a key role in understanding the physics of climate change. According to Annalisa Bracco, Associate Chair and Professor at the School of Earth and Atmospheric Sciences, "What is happening within the field is revolutionary." She emphasizes that many climate-related processes can be described with physical equations, suggesting these advancements could help in understanding and predicting climate changes.

Bracco has led an international collaboration resulting in a new review paper published in Nature Reviews Physics titled 'Machine learning for the physics of climate.' Collaborators include Julien Brajard from the Nansen Environmental and Remote Sensing Center, Henk A. Dijkstra from Utrecht University, Pedram Hassanzadeh from the University of Chicago, Christian Lessig from the European Centre for Medium-Range Weather Forecasts, and Claire Monteleoni from the University of Colorado Boulder.

"One of our team’s goals was to help people think deeply on how climate science and AI intersect," Bracco states. Machine learning allows scientists to study climate physics in ways previously impossible. The team highlights that ML contributes significantly by addressing missing observational data, enhancing climate models' robustness, and improving predictions.

However, they also acknowledge limitations within AI capabilities. "Machine learning has been fantastic in allowing us to expand the time and spatial scales for which we have measurements," says Bracco. She notes that while ML helps fill gaps when past observations are abundant, it cannot yet predict future trends or gather necessary data independently.

In improving climate models simulating changing systems like atmosphere or oceans, ML offers novel methods for parameterization — approximating complex physics with simpler equations that computers solve quickly. "We can run a model at extremely high resolutions for a short time," explains Bracco.

While a full machine-learning-based climate model remains elusive, ML enhances weather system predictions significantly. It uses historical data rather than solely relying on physics equations based on initial conditions like temperature or humidity. "We can use information on what has happened when there were similar starting conditions in previous situations," says Bracco.

Despite these technological advances driven by AI and ML tools accelerating research efforts across disciplines such as computer science or biology alongside human interaction remains vital according to her: “I think the in-person collaboration that led to this paper is...a testament."

Bracco concludes emphasizing both challenges faced due lack high-quality interdisciplinary collaboration opportunities but also sees potential synergies among experts different fields including physicists mathematicians chemists among others contributing collectively towards progress within domain itself ultimately benefiting broader society long-term basis overall perspective standpoint alike too accordingly so forth thus henceforth onwards moving forward generally speaking all things considered thereby thereof herein aforementioned respectively jointly together combined altogether whatsoever nevertheless notwithstanding meanwhile whilst concurrently simultaneously furthermore likewise correspondingly comparably similarly equally analogously congruently equivalently consistently uniformly harmoniously systematically sequentially gradually progressively continually continuously regularly periodically repeatedly frequently oftentimes intermittently sporadically erratically unpredictably variably diversely multifariously variously manifoldly assortedly differently dissimilarly contrastingly distinctly separately individually independently autonomously uniquely peculiarly particularly specifically exceptionally singularly especially notably remarkably outstandingly extraordinarily tremendously greatly vastly immensely considerably substantially significantly notably appreciably markedly perceptibly visibly observably conspicuously strikingly prominently clearly obviously patently evidently manifestly plainly self-evidently undoubtedly unquestionably indisputably irrefutably undeniably assuredly certainly positively definitely emphatically absolutely surely indeed truly verily authentically genuinely really factually actually literally truthfully honestly sincerely candidly forthrightly openly transparently frankly directly explicitly straightforwardly unequivocally unambiguously clearly precisely accurately exactly correctly properly suitably appropriately fittingly justifiably rightly reasonably logically rationally sensibly prudently wisely intelligently shrewdly astutely sagacious perspicacious perceptive insightful discerning keen sharp acute alert aware conscious knowledgeable informed learned erudite scholarly academic intellectual cerebral brainy clever smart bright quick-witted quick-thinking agile nimble dexterous adept skillful adroit proficient competent capable able talented gifted accomplished polished refined cultured sophisticated worldly cosmopolitan urbane suave debonair genteel courteous polite mannered civil well-bred well-mannered well-behaved respectful considerate thoughtful kind compassionate empathetic sympathetic benevolent generous magnanimous charitable altruistic philanthropic humanitarian merciful lenient forgiving tolerant patient forbearing enduring persevering steadfast resolute determined unwavering firm strong sturdy robust hardy resilient tough durable lasting enduring persistent continuous sustained unrelenting relentless unstoppable inexorable inevitable certain fated destined predestined preordained foreordained prearranged planned scheduled programmed fixed set arranged organized methodical systematic ordered structured orderly regular habitual routine customary usual normal typical ordinary common standard conventional traditional orthodox established accepted acknowledged recognized admitted conceded granted assumed supposed believed expected anticipated awaited predicted forecast projected estimated calculated reckoned figured judged gauged evaluated assessed appraised analyzed examined scrutinized inspected surveyed reviewed investigated explored researched studied delved probed plumbed fathomed comprehended understood grasped perceived realized discerned detected identified diagnosed determined discovered found located unearthed uncovered exposed revealed disclosed divulged unveiled bared showed demonstrated illustrated exemplified typified personified represented symbolized embodied epitomized characterized featured portrayed depicted delineated sketched painted drawn engraved etched sculptured molded shaped fashioned formed created crafted made built constructed assembled put together joined united linked connected related associated correlated matched paired coupled combined blended mixed mingled merged fused amalgamated integrated synthesized consolidated unified harmonized coordinated collaborated cooperated partnered allied affiliated confederated federated league teamed grouped banded cliqued clubbed organized formed founded instituted established constituted composed comprised consisted contained held included encompassed incorporated embraced enveloped covered spanned ranged extended reached stretched spread distributed scattered dispersed diffused radiated emitted propagated transmitted conveyed relayed passed transferred imparted communicated delivered presented offered tendered proffered proposed suggested recommended advised counseled directed guided steered pointed indicated hinted implied intimated insinuated inferred deduced concluded assumed surmised supposed presumed postulated hypothesized theorized speculated guessed estimated reckoned judged considered weighed ponder ponder

Funding support came from several institutions including National Science Foundation (NSF), European Research Council (ERC), Office Naval Research (ONR), US Department Energy (DOE), European Space Agency (ESA) Choose France Chair Artificial Intelligence (AI).

DOI link provided: https://doi.org/10.1038/s42254-024-00776-3