1233. The Futures of Climate Modeling

In contrast to weather forecasts, climate predictions cannot be verified in real time and have relied on building physically based models. The paper "The futures of climate modeling," published in npj Climate and Atmospheric Science, examines the current state of climate modeling and future developments.

As climate science develops, what climate models have predicted as signals of climate change as the world warms — such as land warming more than the ocean and the Arctic warming the most, etc. — are now being observed. The success of climate models lies in the establishment of the dominant paradigm of climate science, which applies physical laws under the assumption that large and small spatial scales are sufficiently separated, and is called the standard approach.

Climate modeling has traditionally improved predictions through advances in theoretical understanding and representation of physical processes. Since the birth of the first atmospheric general circulation models in the mid-20th century, rapid progress has been made along two streams, enabling theoretical understanding of global warming. One is specialized in global weather prediction on short scales, and the other is specialized in long-term climate simulation and prediction. The latter played a key role in demonstrating that anthropogenic greenhouse gas emissions warm the climate in the 1960s, leading to the Nobel Prize awarded to Dr. Syukuro Manabe in 2021. Dr. Manabe and his colleagues simplified physics and gradually increased the complexity to understand the mechanisms underlying the new response to increasing CO2. All of their predictions have since been observed, providing clear examples of how incrementally adding complexity towards all-encompassing complex simulations improves our understanding and confidence in the model's behavior.

But as warming continues, discrepancies between real-world climate change signals and predictions based on standard approaches are piling up, especially at regional scales. At the same time, disruptive computational approaches are driving new paradigms. In particular, in recent years, climate modeling has entered a unique era. It is now possible to combine satellite-based observational records with in situ measurements and reanalysis results. Increasing climate change signals and longer observational records have highlighted areas where previous climate predictions worked well, but also revealed contradictions that need to be resolved in both understanding and prediction.

While the climate is changing rapidly and society is becoming more urgent for accurate climate information, the field of climate science is changing as well, leading to a variety of proposals among scientists for the future of climate modeling. None of the proposed approaches have yet proven sufficient to address the scientific challenges facing the climate science community, with some requiring significant investments in one proposal, naturally raising concerns that it could strain resources for other proposals. Scientists say they don't consider discovering bugs or things not working as expected to be a sign of failure, but rather finding problems and figuring out errors as a sign of success and progress.

(References)

Shaw, T.A., Stevens, B. The other climate crisis. Nature 639, 877–887 (2025). https://doi.org/10.1038/s41586-025-08680-1
S. Bordoni et al, The futures of climate modeling, npj Climate and Atmospheric Science (2025). DOI: 10.1038/s41612-025-00955-8 https://www.nature.com/articles/s41612-025-00955-8

Contributor: Miyuki IIYAMA, Information Program