Delving into climate dynamics: How we test the performance of climate models that are used for climate projections

30 Jun 2025

Author: ZHANG Yi Lin

Climate models are powerful tools that help scientists understand and project how the Earth’s climate might change in the future. But before we can rely on these models to make predictions, such as how much it might rain in the future, we first need to ensure they replicate past data accurately. Most of the time, the match is not perfect because our understanding of the processes driving the climate is still incomplete—it's always a work in progress and something our research group is deeply passionate about.  

Global climate models (GCMs) rely on complex rules, called parameterisations, to simulate how clouds form and rain falls. However, some of these processes are still not fully understood, especially in the tropics, where climate and weather processes tend to be highly complex due to tropical convection.  

fig1_v1
Schematic representation of the grid structure and physical processes used in General Circulation Models (GCMs) (Source: History of climate modeling, Edwards, 2010) 

Our previous study showed how stable isotopes in rainwater, like δ¹⁸O, d-excess, and ¹⁷O-excess, act as natural tracers of water's journey through the atmosphere and help us learn more about how our climate works. By studying isotopes in rainwater, we can better understand key processes like evaporation, condensation, and cloud formation. 

  
fig2
Schematic representation of the rainout effect on δ2H and δ¹⁸O values (Source: Coplen et al. 2000) 

In a new study recently published in the Journal of Advances in Modeling Earth Systems (JAMES), we used water isotope data from the Global Network of Isotope in Precipitation (GNIP) and other published data to evaluate the performance of the NASA’s GISS-E2.1 climate model, which is one of the few models that incorporate isotope tracers. GISS climate model outputs contribute to the CMIP (Coupled Model Intercomparison Project) archive, which is used by the IPCC reports. We added a rare isotope tracer, ¹⁷O, to evaluate how well the model simulates three key isotope signals in precipitation: δ¹⁸O, d-excess, and ¹⁷O-excess.  

Overall, we find the model performance to be satisfactory in simulating most precipitation isotope patterns. The model does a good job of matching the real-world patterns of 𝛿18O and d-excess, but produces slightly low d-excess values in the tropics. Most of the simulated ¹⁷O-excess values are within the range of observed values, but it is still hard to constrain the accuracy of the model because there is not enough observational data on ¹⁷O-excess. In tropical regions, we found that evaporation of raindrops and cloud processes have a strong impact on precipitation d-excess and 17O-excess.  

Fig3
Long-term annual mean of precipitation (a) δ18O, (b) d-excess and (c) 17O-excess simulated by the GISS-E2.1 model (color shading) compared to the observational data (plotted as points) (Source: Yilin Zhang/NTU Singapore)  

This study helps explain how these isotope tracers behave in the atmosphere and will support efforts to improve climate models—a crucial step toward making future climate predictions more accurate and reliable. By improving these climate models, we also hope to provide better tools for understanding droughts, floods, and long-term water resource planning in Southeast Asia and beyond.  

This work was supported by a Singapore MOE Tier 2 Grant (MOEMOET2EP10121‐0008), Singapore MOE Tier 3 Grants (MOE2019‐T3‐1‐004; MOE-MOET32022-0006) and a Stephen Riady Geoscience Scholarship. 

 

Blog Category

Research

Subscribe to the EOS Newsletter

Stay in touch with the latest news, events, research, and publications from the Earth Observatory of Singapore.

Email is required

Email is wrong format

You Can Make a Difference

Partner with us to make an impact and create safer, more sustainable societies throughout Southeast Asia.
Make A Gift