An international research team has completed the first fully standardized comparison of isotope-enabled climate models, demonstrating that a multi-model ensemble provides the most accurate representation of the present-day global water cycle.

Water isotopes--molecules of water containing heavier forms of hydrogen and oxygen--are powerful tracers of atmospheric moisture transport and phase changes. Over the past two decades, isotope-enabled climate models have been developed independently by multiple research groups worldwide. However, differences in experimental design and boundary conditions have made it difficult to directly compare their performance and to assess the robustness of simulated isotope distributions.

To address this challenge, the research team conducted the Water Isotope Model Intercomparison Project (WisoMIP). In this project, eight state-of-the-art isotope-enabled climate models were forced with identical atmospheric circulation fields derived from the ERA5 reanalysis, together with unified sea surface temperature and sea ice conditions. The models simulated the three-dimensional distribution of atmospheric water isotopes on a daily basis from 1979 to 2023, enabling a direct comparison of isotope processes while controlling for large-scale atmospheric circulation.

The results show that, although individual models exhibit regionally varying biases, the multi-model ensemble mean consistently outperforms any single model in reproducing observed isotope distributions in precipitation, water vapor, and snow (Figure 1). In addition, the ensemble reproduces the expected large-scale spatial structure of precipitation oxygen isotopes, while revealing where inter-model uncertainty remains large (Figure 2). Together, these results demonstrate both the robustness and the limitations of current isotope-enabled climate models.

Because water isotope signals are preserved in natural archives such as ice cores, corals, and tree rings--and can now be directly observed in precipitation and atmospheric water vapor--these findings provide a critical link between modern observations, paleoclimate reconstructions, and future climate projections. The WisoMIP dataset establishes a new international benchmark for evaluating isotope-enabled climate models and is expected to contribute to reducing uncertainty in climate change assessments.

Researcher comment
"By comparing isotope-enabled climate models under fully unified conditions for the first time, we demonstrate the robustness and added value of the multi-model approach for tracing Earth's water cycle," said Dr. Hayoung Bong, alumnus of the Institute of Industrial Science, The University of Tokyo, now at NASA Goddard Institute for Space Studies. "At the same time, the spatial patterns of model spread highlight where further model development and observations are most needed."

Publication information
Journal: Journal of Geophysical Research: Atmospheres
Title: Water Isotope Model Intercomparison Project (WisoMIP): Present-day Climate
Authors: Hayoung Bong et al.
DOI: 10.1029/2025JD044985

Institutions involved
・Institute of Industrial Science, The University of Tokyo
・Chiba University
・Chuo University
・Meteorological Research Institute, Japan Meteorological Agency
・Geophysical Institute and Bjerknes Centre for Climate Research, Bergen, Norway
・Rice University, Houston, Texas, USA

Funding
This research was supported by the Japan Society for the Promotion of Science (JSPS), the Ministry of Education, Culture, Sports, Science and Technology (MEXT), the Japan Science and Technology Agency (JST), the Environmental Restoration and Conservation Agency (ERCA), and the Arctic Challenge for Sustainability Phase III (ArCS III), the Research Council of Norway through the iTRANSFER project, and the U.S. National Science Foundation.

Research contact
Kei Yoshimura, Professor
Institute of Industrial Science, the University of Tokyo
Tel：+81-4-7136-6965
E-mail：kei (Please add "@iis.u-tokyo.ac.jp" to the end)

Media contact
Institute of Industrial Science, The University of Tokyo
Public Relations Office
Email: pro (Please add "@iis.u-tokyo.ac.jp" to the end)

Figures
Figure 1 | Multi-model ensemble outperforms individual isotope-enabled climate models in reproducing observed precipitation isotopes.
Global skill comparison of eight isotope-enabled climate models (colored symbols) and their multi-model ensemble mean (black star). The ensemble shows the closest agreement with observations, highlighting the robustness of the multi-model approach.

Figure 2 | Global structure and inter-model spread of precipitation oxygen isotopes.
The ensemble mean reproduces the expected large-scale patterns of annual mean δ¹⁸O in precipitation (left), while the spatial distribution of model spread (right) identifies regions of higher uncertainty, underscoring the value of a multi-model framework.