We use different types of computer models to make prognoses about the future climate. Earth system models are an advanced type of model that simulates the interaction between the different components of the Earth’s climate system, including the atmosphere, hydrosphere, biosphere, lithosphere, and cryosphere. There are many important processes in the climate system that are small in terms of physical extent and difficult to replicate in the models. This leads to uncertainties in the forecasts. A method has now been developed that makes the models and thus the climate prognoses more accurate.
The reason why the models are unable to capture small-scale processes is that the models do not have a high enough resolution. One example is clouds, which typically have an extent of only a few kilometres, while the models often have 50-100 km between each calculation point.
In order for models to be able to include cloud formations and precipitation, something called parameterisation is used. These are simple mathematical descriptions of how smaller processes are affected by – and themselves affect – the larger weather or climate systems that the models capture. The parameters must be adjusted so that the result is as close as possible to real-life measurements. This is called ‘tuning’. Manual tuning takes a lot of time, both for researchers and computers, and can easily lead to errors because the models consist of many such parameters that must be tuned individually. These may also affect each other, so that tuning one part of the model may improve or impair the result in other areas.
To enhance Earth system models and streamline the tuning process, Tarkeshwar Singh, along with colleagues from NORCE and the Nansen Center, has developed a novel methodology: Data assimilation is used for a more precise adjustment of small-scale parameters. The process has also been partially automated, resulting in fewer manual choices, faster processing, and more accurate models.
The method has been tested on the ocean biogeochemistry model part of an earth system model. The result is a reduction in errors of 10-20%. Using well-tuned ocean models is particularly important because they can show how the ocean absorbs CO₂ and how CO2 levels affect marine life. With better models, we can predict future changes in the ocean with even greater accuracy.
The new method can be applied to many different parameters in earth system models, not just ocean models, and is therefore an important step towards more accurate climate models. It gives researchers a better tool and politicians a better basis for decision-making in the face of climate challenges.
Key researchers: Tarkeshwar Singh, François Counillon, Yiguo Wang
