A Stanford study uses artificial intelligence to estimate how soon global warming could cross major temperature thresholds. Its findings put the 1.5-degree mark in the early 2030s across three climate pathways, while warning that even rapid emissions cuts may not prevent warming from reaching two degrees.
How the researchers tested the forecast
The team trained an AI system on results from current climate model simulations and historical temperature measurements from around the world. The model used annual temperature anomalies over the 1951-1980 reference period to estimate how many years remained before a selected warming threshold would be reached.
To assess its performance, researchers asked the system to estimate the current 1.1-degree level of global warming for each year from 1980 to 2021. It correctly predicted the level for 2022, with a probability range of 2017 to 2027. The analysis also accounted for how the estimated time to reach 1.1 degrees has shortened in recent decades.
Study lead Noah Diffenbaugh described the historical test as passing the “acid test.” He said the researchers had been skeptical of the approach, but its ability to predict the course of warming so far raised their confidence in forecasts of future temperatures.
Three pathways point to the 1.5-degree mark
The researchers compared three climate pathways that vary in the intensity of warming caused by greenhouse gases. In each one, the global average temperature reaches 1.5 degrees above pre-industrial levels in the early 2030s.
The paths diverge more sharply when the study considers the two-degree threshold. If emissions stay high over the next few decades, the model estimates a one-in-two chance of reaching two degrees by mid-century and a more than four-in-five chance by 2060.
That forecast links the temperature outlook to the pace of emissions reductions. The model suggests that the timing of net-zero emissions matters, while also indicating that the threshold could be reached even when emissions fall rapidly.
Why the two-degree forecast matters
Diffenbaugh said the model is very confident that warming will exceed two degrees if reaching net-zero emissions takes another half century. He contrasted this with other projections that assume the two-degree threshold can be met if net-zero emissions arrive before 2080. He also said the study’s findings may prove controversial among scientists and policymakers.
Even in a scenario where emissions rapidly fall to zero by 2076, the AI forecast puts the two-degree limit at serious risk. It estimates a one-in-two chance of reaching that level by 2054, and a two-in-three chance of exceeding it sometime between 2044 and 2065.
These estimates are model forecasts, not a certainty about the exact year temperatures will cross a threshold. Their value is in comparing possible pathways and showing how the projected timing shifts under different emissions assumptions.
Climate goals and the path ahead
Some countries have set net-zero emissions goals between 2050 and 2070. Diffenbaugh said those ambitions are intended to meet the 1.5-degree goal, but may also be needed to stay within the two-degree threshold.
The study therefore raises a practical question for climate planning: whether emissions cuts can happen soon enough to change the risk implied by the forecast. The AI approach offers one way to bring historical temperature observations into that assessment, while its conclusions remain part of a wider discussion about climate projections.
The research was led by Stanford University climate scientist Noah Diffenbaugh and Colorado State University atmospheric scientist Elizabeth Barnes. It was published Jan. 30 in the Proceedings of the National Academy of Sciences.