Skip to main content
U.S. flag

An official website of the United States government

Publication Date
30 July 2021

E3SM Probes Uncertainties in Ocean Surface Fluxes

Quantifying model uncertainty caused by the way air-sea fluxes are calculated.
Print / PDF
Powerpoint Slide
The UA algorithm predicts less evaporation at high wind speeds (a) and a general shift to lower wind speeds (b), compared to the control experiment. The result is significantly less wintertime evaporation from the Gulf Stream region (c).

Atmospheric turbulence – the chaotic whirls in the wind – causes the transfer of heat and moisture between the air and the sea. Turbulence is also the link between winds and ocean currents. The method used to represent turbulence in models is based on a mixture of theory and experiment. Different researchers have come up with different versions of the method (algorithms), which have up until now been compared in “idealized” tests, where the output does not affect the input. However, this doesn’t tell us everything about the effects of different algorithms in models, because the outputs can affect the inputs – forming a feedback loop. In this study, we compare – for the first time - model runs with three different surface turbulence algorithms. The effects are seen at the surface – the direct effects – but also higher in the atmosphere and deeper in the ocean.


Our results show changes in how E3SM portrays Earth’s energy and water cycles. Changes in the energy cycle, especially in cloud distributions (which affect how energy is radiated back to space) and in how the ocean moves heat around, suggest that the choice of surface turbulence algorithm may affect global warming projections. Changes seen in the water cycle suggest that the choice of an algorithm may affect how the model represents heavy rainfall and drought. However, more research is required to understand these issues better – longer model runs are needed, with the atmosphere and ocean models running together (rather than separately as in this study).


With three algorithms and two model setups (ocean-only, atmosphere-only), six different simulations were run. The ocean and atmosphere tests involved different feedbacks: most importantly, the atmosphere tests allowed feedbacks involving the wind speed, while the ocean tests used pre-determined wind speeds. Therefore, the atmosphere tests tended to show larger differences between algorithms than the ocean tests. For example (see figure), changes in latent heat flux (related to evaporation) changed in the atmosphere tests due to two effects: changes at each wind speed, and changes in the distribution of wind speeds. Meanwhile, changes in the ocean tests were only caused by the first of these. Nonetheless, a few regions consistently showed up as where the algorithm choice makes a large difference: e.g., the Gulf Stream, tropical Pacific, and Indian Oceans.  

Determining which algorithm is “best” remains quite inconclusive, as no single algorithm performs best across all metrics. For example, one improves global average precipitation compared to the other two but is not the best when it comes to top-of-atmosphere radiation balance. There are even differences when it comes to the same metric in different regions and seasons. Overall, this reflects the inherent uncertainties in the method used and also the far-reaching consequences of these uncertainties in Earth system models.

Point of Contact
Jack Reeves Eyre
University of Washington
Funding Program Area(s)