Surface-layer parameterizations in atmospheric general circulation models remain a significant source of uncertainty, potentially affecting model biases and the spread of climate change projections. Alongside this work in the archive are studies on the sensitivity of circulation to surface and boundary-layer processes: the influence of bottom friction on shallow-water vortices, the role of clouds in reorganizing ocean eddies, heat pumps in Enceladus's ocean, and the effect of mountains on exoplanet climates. The present paper complements this group by focusing on the sensitivity of the large-scale circulation to changes in sea surface roughness, which has been relatively understudied compared to cloud and convection parameterizations.
The CAM3 atmospheric general circulation model was used in an 'aquaplanet' configuration with a prescribed zonally symmetric sea surface temperature distribution (Qobs profile). Paired simulations were run with the original and reduced roughness for momentum Z0m under weak winds (the coefficient c4 was decreased from 0.0027 to 0.000027 m/s). To isolate the mechanism, experiments were performed with the mechanism 'switched off': reducing roughness only in the tropics or only in the extratropics, and also only in the heat flux formulas or only in the stress formula. Additionally, ensemble simulations with a sudden onset of reduced roughness were used to analyze the temporal evolution of the response. To test robustness, simulations were run with other sea surface temperature profiles and in the more complex AMIP configuration with prescribed SSTs and with a mixed-layer ocean.
Reducing Z0m roughness under weak winds leads to a statistically significant response in the zonally averaged circulation, which is qualitatively reproduced across all configurations considered. In the aquaplanet with the Qobs profile, midlatitude westerlies shift poleward by several degrees, and the shift extends to the surface. The Hadley cell weakens by about 25% in terms of the maximum streamfunction at 500 hPa, and the Intertropical Convergence Zone (ITCZ) shifts poleward. Tropical tropospheric temperatures decrease, especially in the middle and upper levels, while subtropical warming is observed, consistent with the thermal wind. The 'switch-off' experiments showed that the response is initiated by reduced latent and sensible heat fluxes in the tropics, not by changes in surface stress. Temporal analysis revealed a fast adjustment of the Ekman balance and a slower reorganization of eddy momentum fluxes. In the AMIP configurations, the response is weaker but retains the same structure, including the jet shift and Hadley cell weakening.
The study demonstrates that even small changes in surface-layer parameters, within observational uncertainties, can cause a substantial large-scale circulation response, comparable in amplitude to natural ENSO variability. Unlike neighboring studies that focus on individual aspects of atmosphere-surface interaction (e.g., bottom friction or cloud effects), this paper links the sea surface roughness parameter to a coherent response of tropical and extratropical circulation through a thermodynamic mechanism. This underscores the need to consider surface parameterization uncertainties on par with cloud and convection parameterizations when analyzing model biases and projections.
The topic could develop toward a systematic comparison of circulation sensitivity to surface-layer parameter changes across different general circulation models. The interaction with convection and cloud parameterizations, which could alter the sign of the ITCZ response, deserves separate study. It is also important to investigate how the results depend on the specific surface-layer scheme (Large and Pond vs. others) and on the way roughness is specified for heat and moisture.
The results may influence the interpretation of systematic biases in climate models, especially in the position of jet streams and the structure of tropical convergence. They are also relevant for developing more reliable atmosphere-ocean coupling parameterizations in weather and climate prediction models.
A next step could be to perform similar experiments in several independent general circulation models to assess the universality of the response. It would also be useful to investigate the impact of roughness changes on regional precipitation and monsoon circulation in more realistic configurations.
The work is directly related to the problem of systematic model biases in jet stream position and ITCZ structure, as well as the spread in circulation projections under global warming. It shows that uncertainties in surface parameterizations can be as important as those in clouds and convection, and should be considered when assessing the robustness of climate projections.
In the original CAM3 parameterization, the drag coefficient under weak winds (less than 5 m/s) can be reduced by a factor of 100, and this still falls within the scatter of observations from field campaigns over the ocean.