Researchers created a new rating system to assess the credibility of climate model simulations in the latest IPCC report, emphasizing that models predicting severe warming should be taken seriously.
What does the future hold for our climate? Scientists globally are investigating climate change by developing models of the Earth’s systems and compiling extensive observational data to comprehend and forecast the climate over the next century. But how do we determine which models are the most realistic and align closely with the planet’s future climate?
To tackle this question and assess the reliability of different models, scientists from EPFL have established a rating system. They examined climate model outputs from the global climate community as presented in the recent IPCC report. Their findings reveal that about one-third of these models fail to accurately represent current sea surface temperature data; another third display robustness without being significantly affected by carbon emissions; and the final third are also sturdy but forecast a notably hotter future due to high sensitivity to carbon emissions. These results are detailed in Nature Communications.
“Our results indicate that the carbon-sensitive models, which predict considerably more heating than the IPCC’s most likely estimates, are credible and warrant serious consideration,” explains Athanasios (Thanos) Nenes, a professor at EPFL in the Laboratory of Atmospheric Processes and their Impacts, and co-author of the study with graduate student Lucile Ricard.
“This means that the current initiatives aimed at reducing carbon emissions, which are founded on lower sensitivity estimates, may insufficiently prevent a dangerously hot future,” Ricard adds.
Assessing the credibility of climate models: a big data approach
Since the mid-1800s, scientists have systematically studied the planet, tracking meteorological variables like temperature, humidity, atmospheric pressure, wind, precipitation, and the status of oceans and ice. Particularly in recent decades, advancements in observational networks and satellites have led to an overwhelming amount of data, making predictions about the climate’s future a challenging endeavor.
To evaluate specific climate models, the EPFL team introduced a tool named “netCS” which employs machine learning to group climate model outputs, synthesizing their regional behaviors and comparing them with existing data. Thanks to netCS, researchers can identify which climate simulations most accurately reflect observations and rank them accordingly.
“Our method offers a rapid way to evaluate a climate model, as netCS can process terabytes of data within a single afternoon,” says Ricard. “Our rating system represents a new form of model evaluation that complements insights gained from historical data, paleoclimate records, and our understanding of processes discussed in the 2021 IPCC AR6 assessment report.”
Nenes, who has been invited to participate in the IPCC AR7 scoping meeting in Malaysia, recalls a piano concert he performed in Athens nearly thirty years ago during a summer with temperatures peaking at 33 to 36 degrees Celsius, which were considered extraordinarily high at that time. “The heat made it incredibly difficult to play the piano. Now, Greece frequently faces summertime temperatures exceeding 40 degrees, with forest fires invading urban areas and burning neighborhoods where I once lived. It’s only going to worsen. The planet is literally on fire, and worldwide temperatures are breaking records year after year, with serious consequences.”
“At times, I feel like climate scientists are akin to Cassandra from Greek mythology,” Nenes reflects. “She had the gift of prophecy but was doomed to be ignored. However, this inertia or inaction should fuel our motivation rather than discourage us. We must collectively awaken and take substantial action against climate change, as it may be accelerating more than we have realized.”
Other contributors to this study include Fabrizio Falasca from the Courant Institute of Mathematical Sciences at New York University and Jacob Runge from the Technical University of Berlin. The research was funded by the European Union’s Horizon 2020 research and innovation program under the Marie SkÅ‚odowska-Curie grant agreement No. 860100 (iMIRACLI), by the FORCeS project under Horizon 2020 with grant agreement No. 821205, and by the CleanCloud project under the Horizon Europe program with grant agreement No. 101137639.
