NOAA GFDL - Geophysical Fluid Dynamics LaboratoryGFDL - Geophysical Fluid Dynamics Laboratory

Role of Ocean Model Formulation in Climate Response Uncertainty

September 13th, 2018

Key Findings

  • It is possible to construct coupled climate models that produce the same equilibrium climate response, despite very different formulations (i.e. vertical coordinate and numerics) in their ocean model components.
  • Both GFDL-ESM2Mb and GFDL-ESM2G produce remarkably similar equilibrium climate responses among a range of climate diagnostics after 5,000 years of quadrupled CO2 forcing (4X CO2), even though starting from preindustrial climate states where global average sea surface temperatures differ by 0.4 °C.
  • Different Atlantic Meridional Overturning Circulation (AMOC) responses are seen in the two GFDL ESMs under the 4X CO2 forcing, highlighting their differences in ventilation and deep water formation. Such differences may also contribute to model differences in the millennial scale accumulation of dissolved inorganic carbon.
  • Increasing the background diapycnal diffusivity in the GOLD ocean model used in ESM2G changes the spatial patterns of the equilibrium climate response of SST, reduces the stratification of the upper ocean, and increases the oceanic heat uptake. The changes impact the climate responses of the model.
  • These results highlight the importance of constructing ocean model components where lateral and vertical mixing parameterizations are better-constrained by observations and sources of spurious numerical mixing are kept to a minimum.

J. P. Krasting, R. J. Stouffer, S. M. Griffies, R. W. Hallberg, S. L. Malyshev, B. L. Samuels, and L. T. Sentman. Journal of Climate.

Oceanic heat uptake (OHU) is a significant source of uncertainty in both the transient and equilibrium response to increasing the planetary radiative forcing. OHU differs among climate models and is related in part to their representation of vertical and lateral mixing. This study examines the role of ocean formulation — specifically the choice of vertical coordinate and strength of background diapycnal diffusivity (Kd) — on determining the millennial-scale near-equilibrium climate response to a quadrupling of atmospheric CO2.

Using two fully-coupled Earth System Models (ESMs) that have nearly identical atmosphere, land, sea ice, and biogeochemical components, it is possible to independently configure their ocean model components with different formulations and produce similar near-equilibrium climate responses. The near-equilibrium climate response was obtained from the ESM2Mb and ESM2G through a “brute force” approach of running 5,000-year preindustrial control simulations and 5,000-year 4X CO2 experiments with both models. A similar suite of 5,000-year control and 4X CO2 experiments was also performed with the GFDL-CM2G coupled model to evaluate the sensitivity of the climate response to increasing amounts of Kd.

The patterns and magnitude of the SST response are similar between the two models (r2 = 0.75, global average ~4.3 °C) despite their initial pre-industrial climate mean states differing by 0.4 °C globally. The surface and interior responses of temperature and salinity are also similar between the two models. The ESMs differ, however, in their AMOC responses. While both models exhibit an initial weakening of the circulation during the first 1,000 years of forcing, AMOC remains weak in ESM2G throughout the entire 5,000-year simulation while it strengthens relative to the pre-industrial baseline in ESM2Mb. The efficiency of AMOC in transporting heat poleward from the tropics to the Northern Hemisphere high latitudes also diminishes in both models, but to a lesser degree in the ESM2G model. Changes in ocean ventilation and deep water formation associated with these AMOC responses also impact the accumulation of dissolved inorganic carbon in the ocean interior. A parameter sensitivity analysis demonstrates that increasing the amount of Kd produces different SST response patterns at near-equilibrium and increases the total oceanic heat uptake. These results suggest that the impact of the ocean vertical coordinate on the climate response is small relative to the dependence on representing sub-gridscale mixing.

This work examines two non-eddying ocean model simulations and it remains to be demonstrated if similar responses could be found in eddy-permitting and eddy-active models. Ocean models are increasingly using “hybrid” vertical coordinates, and more work is needed to evaluate centennial and millennial scale climate simulations using this class of models. This work also highlights the importance of constructing ocean model components where mixing parameterizations are well-constrained by observations and theoretical understanding.

Figure 1: GFDL-ESM2Mb and GFDL-ESM2G produce similar global mean temperature responses after 5,000 years of forcing from quadrupling atmospheric CO2. Sea surface temperature response (left) and volume mean ocean potential temperature (right) in the 4X CO2 simulations for ESM2Mb (magenta) ESM2G (green). The linear trends from the preindustrial control simulations are subtracted from the responses and responses are normalized by adjusting the curves for ESM2G upward and ESM2Mb downward by half of the mean difference between the two models as determined by the detrended control simulations (∼ 0.4 °C top; ∼ 1.1 °C bottom).