Making climate models more accurate by improving their tuning

Earth’s climate is a very complex system, and it is not easy to understand with all its components – the main ones being ocean, land, atmosphere, and sea ice. Nevertheless, scientists have been trying for decades to predict future changes in climate with numerical models. These models keep getting better, but all components have systematic errors to some degree. Decreasing errors and thereby predicting the future climate more reliably will benefit society by allowing us to better adapt to climate change. 


The ocean biogeochemistry and its effect on Earth’s climate

One important component of the climate system is the ocean. Processes involving the ocean and going on in the ocean play a big part in how Earth’s climate changes over time. The ocean’s biogeochemistry is of special interest to climate scientists – this covers biological, geological, and chemical processes – because of the carbon cycle. Increasing amounts of carbon in the atmosphere (as CO2) lead to increasing temperatures worldwide, and we humans release more and more carbon into the atmosphere by burning more and more fossil fuels. This carbon in the atmosphere enters the ocean over time and it can be taken up by small plants there, called phytoplankton. When phytoplankton is eaten by animals in the ocean and those die and sink to the seafloor, carbon is removed from the atmosphere and ocean, and captured. Thanks to this cycle, the ocean and its inhabitants contribute to slowing down the CO2 increase in the atmosphere, thereby affecting the climate. It is crucial to understand how carbon capture and storage processes will evolve in the future.

The problem with using ocean biogeochemistry in models

But to predict future climate change reliably, numerical models need really good information on ocean biogeochemistry dynamics. Today, climate models get this information from laboratory experiments, and scientists apply the results to the ocean worldwide. The ocean biogeochemistry is defined by many parameters (values for different processes) that are included in climate models, and these parameters get slightly adjusted in accordance with observations from the ocean. This adjustment process is called tuning. Problematic with that approach is that many slightly wrong values can easily build up to large errors. It is then challenging to identify the main source of error as there are many parameters that could be the cause.


A new approach to make the ocean biogeochemistry of climate models less biased

A group of researchers under the lead of Tarkeshwar Singh (Climate Dynamics and Prediction Group at NERSC) and for the project EU-TRIATLAS took on the challenge to improve the tuning method – how to best estimate biogeochemical parameters of the ocean. The findings were recently published in Frontiers in Marine Science. Their successful approach uses observations from the ocean and a data assimilation technique to estimate regional biogeochemistry parameters in climate models very efficiently. Singh and his colleagues show that their new tuning method drastically reduces errors in the model. This method will help to provide more reliable information on climate change in the future.  They tested their approach in an idealized configuration (this means the observations used were not real, but produced by a model), and it now needs to be demonstrated in a real framework (this means the observations will be real-world data). This approach can be applied to all components of the Earth System, not just the ocean – and it may present a new way towards better tuned and more efficient climate models!

Copyright: Stine Hommedal / Havforskningsinstituttet, CC BY-SA 4.0Copyright: Stine Hommedal / Havforskningsinstituttet, CC BY-SA 4.0


Singh T, Counillon FS, Tjiputra J, Wang Y, Gharamti ME. Estimation of Ocean Biogeochemical Parameters in an Earth System Model Using the Dual One Step Ahead Smoother: A Twin Experiment. Frontiers in Marine Science. 2022;9.

Add comment

Login to post comments