Scientists have long known the world is running out of time to hit its international climate targets. Now, artificial intelligence has arrived at a similar conclusion.
An innovative new AI study finds that it will take about a decade for humanity to blow past its optimistic goal of limiting global warming to 1.5 degrees Celsius.
That’s the same conclusion scientists have come to when using more conventional climate modeling techniques, but the AI research adds more evidence to the growing conviction among climate scientists and policy experts that the world is all but certain to overshoot the 1.5 C target (Climatewire, Nov. 11, 2022).
Policymakers are still striving to keep global warning well below 2 C, even if they overshoot the 1.5 C target. But even that goal is in danger, according to the AI study. It found that the 2 C threshold could approach even faster than previous research has predicted.
The AI study suggests the 2 C threshold could arrive around the middle of this century, even with relatively stringent reductions in greenhouse gas emissions over the next few decades. That’s decades earlier than conventional climate models generally suggest under the same hypothetical low-emissions scenario. And while the U.N. Intergovernmental Panel on Climate Change acknowledges the world could cross the 2 C threshold before the end of the century in that scenario, it also describes it as an “unlikely” possibility.
That doesn’t mean there’s no hope for reaching the Paris climate targets.
The aggressive emissions-cutting scenario used in the study isn’t necessarily the best the world can do — it still assumes that the world spirals down to net-zero emissions some time after the middle of this century. Meanwhile, dozens of nations around the world have set net-zero timelines for themselves, many of them aiming for the year 2050. That’s a bit earlier than the scenario in the new study assumes.
Reports from the IPCC suggest that achieving the 1.5 C target requires the world to hit net-zero emissions by 2050 and that the 2 C target calls for net zero by 2070 or so. But the AI study suggests that net zero by 2050 may be necessary even for the less ambitious 2 C threshold.
“The AI predictions suggests that those [pledges] may be necessary to avoid 2 degrees,” said Noah Diffenbaugh, a climate scientist at Stanford University, who co-authored the new study with climate scientist Elizabeth Barnes of Colorado State University.
Conventional climate studies typically make climate predictions using computer models, which simulate the physical processes that cause the planet to warm. The new study uses a unique approach to address the prevailing climate question of the times: How quickly will the world warm in the coming decades?
The researchers used artificial neural networks, a type of machine learning, to investigate. Neural networks provide a way for computers to process large amounts of data and recognize patterns within the information they’re provided. They then can be trained to make predictions based on the patterns that they’ve learned.
The researchers first trained their neural networks using input from conventional climate model simulations. They then input global maps of actual present-day temperature anomalies — places around the world where temperatures were warmer or cooler than average. Then, they asked the neural networks for predictions about how quickly the 1.5 C and 2 C targets will arrive under various hypothetical future emissions scenarios.
The neural networks predicted that the 1.5 C target would arrive somewhere between 2033 and 2035. And they found that the 2 C target likely would arrive between 2050 and 2054, depending on how quickly emissions fall in the coming years.
The AI doesn’t entirely rule out the possibility the world could avoid the 2 C threshold under the low-emissions scenario it investigates. But it doesn’t find that outcome likely.
“Given how much warming there’s already been in terms of the map of global temperature anomalies in recent years, the AI is pretty convinced that 2 C is a real possibility in the low forcing scenario,” Diffenbaugh said. “If it takes another half-century to reach net zero, the AI predicts a good possibility of reaching 2 C.”
The study is “definitely new and innovative,” according to Amy McGovern, a scientist at the University of Oklahoma and head of the National Science Foundation’s AI Institute for Research on Trustworthy AI in Weather, Climate and Coastal Oceanography.
McGovern wasn’t involved with the new study but is familiar with the work. Barnes, Diffenbaugh’s co-author on the new study, works for her at the NSF AI institute.
AI is swiftly gaining traction as a new tool for weather and climate science, McGovern said. It can be used to complement conventional modeling techniques in a variety of ways, including everything from making short-term weather predictions to modeling the formation of clouds and other complex climate-related phenomena.
Climate models are highly accurate on the whole. But they require immense computational power and can’t always adequately represent all the granular processes that make up the world’s climate system, especially at a global scale.
AI can replace certain fine-scale physical processes in climate models, allowing them to run faster. And it can help process huge amounts of data more easily.
“There’s really a revolution in the amount of data that’s available right now,” McGovern said. “But there’s so much data out there right now that humans can’t really process it. AI can help bring it down to what humans can focus on.”
AI isn’t necessarily a replacement for more traditional climate and weather modeling techniques. But it can help enhance the models and improve on their limitations, opening up new possibilities for climate research.
“I really think we’re on the cusp of a revolution of how AI is going to get used for weather and climate prediction,” McGovern said. “It’s really going to change the way we can improve our predictions.”
Reprinted from E&E News with permission from POLITICO, LLC. Copyright 2023. E&E News provides essential news for energy and environment professionals.
