For a long time, scientists have recognized that the world is running out of time to meet its global climate targets. In the present, artificial intelligence has reached a similar conclusion.
The optimistic aim of keeping global warming to 1.5 degrees Celsius will not be reached by humanity for about ten years, according to a ground-breaking new AI study.
Although scientists have reached the same conclusion when using more traditional climate modelling tools, the AI research provides more support to the growing belief among climate scientists and policy professionals that the globe is almost certain to exceed the 1.5 C limit (Climatewire, Nov. 11, 2022).
Even if they exceed the 1.5 C goal, policymakers are still working to keep global warming far below 2 C. However, the AI study warns that even this objective may be in jeopardy. It was discovered that the 2 C threshold might approach even more quickly than anticipated by earlier studies.
According to the AI analysis, even with reasonably strict reductions in greenhouse gas emissions over the ensuing few decades, the 2 C threshold may be reached by the middle of this century. Under the same fictitious low-emissions scenario, that is decades earlier than conventional climate models typically predict. And even though the U.N. Intergovernmental Panel on Climate Change recognizes that in such a scenario, the planet could surpass the 2 C threshold by the end of the century, it also labels it as an “unlikely” possibility.
This does not imply that there is no chance of achieving the Paris climate goals.
Although it anticipates that the globe spirals down to net-zero emissions somewhere after the middle of this century, the ambitious emissions-cutting scenario utilized in the study isn’t necessarily the best the world can accomplish. While this is going on, dozens of countries throughout the world have established net-zero targets for themselves, with many of them aiming for the year 2050. That is a little early than the scenario in the new study implies.
According to reports from the IPCC, the global emissions must reach net zero by 2050 in order to meet the 1.5 C objective, and by 2070 or thereabouts in order to meet the 2 C target. Even for the less ambitious 2 C goal, though, the AI study contends that net zero by 2050 may be necessary.
The new study’s co-authors, Stanford University’s Noah Diffenbaugh and Colorado State University’s Elizabeth Barnes, are both climate scientists. The AI forecasts show that those promises may be required to avoid 2 degrees, claims Diffenbaugh.
By simulating the physical processes that cause the earth to warm, computer models are often used in conventional climate studies to produce climate predictions. How quickly will the earth warm in the upcoming decades? is the current climate topic that the new study attempts to answer.
A form of machine learning used by the researchers was artificial neural networks. Computers can handle a lot of data using neural networks, and they can also find patterns in the data that is given to them. They can then be taught to use the patterns they’ve discovered to predict outcomes.
The input from simulations of traditional climate models was used by the researchers to first train their neural networks. After that, they entered maps of actual current-day temperature anomalies—regions around the world where temperatures were either warmer or cooler than average—on a global scale. Then, they used the neural networks to forecast when the 1.5 C and 2 C limits will be reached in several speculative future emissions scenarios.
The 1.5 C target was expected to be reached between 2033 and 2035, according to the neural network’s predictions. And they discovered that, depending on how quickly emissions decline in the upcoming years, the 2 C target would likely be reached between 2050 and 2054.
Under the low-emission scenario it examines, the AI doesn’t completely rule out the chance that the globe could avoid the 2 C threshold. However, it doesn’t anticipate that result is realistic.
The AI is very persuaded that 2 C is a real possibility in the low forcing scenario given how much warming there has already been in terms of the map of global temperature anomalies in recent years, Diffenbaugh added. The AI estimates a good chance of achieving 2 C if it takes another 50 years to reach net zero.
Amy McGovern, a researcher at the University of Oklahoma and the director of the National Science Foundation’s AI Institute for Research on Trustworthy AI in Weather, Climate, and Coastal Oceanography, called the study “certainly novel and interesting.”
McGovern is familiar with the work but was not engaged in the new study. Diffenbaugh’s colleague at the NSF AI institute, Barnes, is the other author of the current paper.
According to McGovern, AI is quickly gaining popularity as a new tool for weather and climate study. It can be used in a variety of ways to supplement conventional modeling methods, from creating short-term weather predictions to simulating the production of clouds and other intricate climate-related processes.
In general, climate models have excellent accuracy. However, they are extremely computationally intensive and, particularly at the global scale, are not always able to accurately reflect all the minute processes that make up the world’s climate system.
Climate models can operate more quickly when AI takes the place of specific fine-scale physical processes. And it can make processing enormous volumes of data easier.
The amount of information that is currently available has undergone a real revolution, according to McGovern. However, the amount of data that is now available makes it challenging for humans to interpret. AI can help reduce it to a level where humans can concentrate.
AI isn’t always a replacement for older methods of climatic and weather simulation. However, it can assist improve the models and address their shortcomings, presenting fresh opportunities for climate research.
In terms of how AI will be applied to forecasting weather and climate, McGovern asserted that we’re on the verge of a revolution. It will significantly alter how we can make forecasts that are more accurate.
