Demonstrate improved ocean model simulations with the new high-resolution MPAS-Ocean.
OVERALL PERFORMANCE MEASURES
1ST QUARTER METRIC – COMPLETED
High-fidelity simulations of the Earth System depend critically on an accurate representation of global ocean currents. Earth System Models (ESMs) in which the simulated currents do not include ocean mesoscale eddy variability have been shown to have numerous biases, including: inaccurate sea surface temperatures (Delworth et al. 2012), weaker meridional heat transport (Kirtman et al. 2012), inaccurate strength and location of western boundary currents (Kirtman et al. 2012, McClean et al. 2011), and poorly represented coastal upwelling (Small et al. 2014). A new high-resolution configuration of the ocean and sea-ice components of DOE’s E3SM has been implemented and used in a simulation forced by the CORE v2 interannual forcing data set (Large and Yeager, 2009). The ocean component of E3SM is the Model for Prediction Across Scale - Ocean (MPAS-O). This configuration is validated by comparing the simulated sea-surface height (SSH) variability to satellite-derived SSH variability from the AVISO (Archiving, Validation, and Interpretation of Satellite Oceanographic) data set.
In order to simulate ocean mesoscale eddy variability, the ocean model must resolve the characteristic length scale of mesoscale eddies, referred to as the first Rossby radius of deformation (RRD; Chelton et al. 1998). The E3SM ocean and sea-ice components have been configured to fully resolve the RRD throughout the vast majority of the ocean. Near the Equator the horizontal grid resolution of the high-resolution E3SM ocean model is 18 km and near the poles the resolution increases to 6 km. This configuration is expected to fully resolve mesoscale eddy activity in most of the global ocean and is similar to 0.1o resolution in traditional-gridded ocean models (e.g., Maltrud and McClean, 2005; McClean et al. 2011, Kirtman et al. 2012). Unlike previous high-resolution ocean simulations, the E3SM high-resolution configuration also fully resolves the first RRD in the vertical direction as well. Stewart et al. (2017) show that an improved representation of mesoscale eddies in the vertical increases SSH variability, eddy kinetic energy (EKE), and the magnitude of the meridional overturning circulation. The vertical-grid-generation algorithm of Stewart et al. (2017) was used to generate the MPAS-O grid for the E3SM high-resolution configuration. The resulting 80-layer vertical coordinate has a characteristic resolution of 2m in the top layer expanding to 90 m at ~1 km depth.
The AVISO satellite has measured SSH variability for more than 20 years. The observations are also at a high enough spatial resolution (0.25o ⨉ 0.25o) to compare directly to high-resolution simulations. The standard deviation of SSH from E3SM is compared with AVISO observations in Figure 1. In general, E3SM SSH variability compares well to AVISO.
2ND QUARTER METRIC – COMPLETED
High-fidelity simulations of the Earth System depend critically on an accurate representation of global ocean currents and their eddies. Earth System Model (ESM) simulations that do not include ocean mesoscale eddy variability have been shown to have numerous biases, including: inaccurate sea surface temperatures (Delworth et al. 2011), weaker meridional heat transport (Kirtman et al. 2012), inaccurate strength and location of western boundary currents (Kirtman et al. 2012, McClean et al. 2011), and poorly represented coastal upwelling (Small et al. 2014). In order to simulate ocean mesoscale eddy variability, an ocean model must resolve the characteristic length scale of mesoscale eddies, referred to as the first Rossby radius of deformation (RRD; Chelton et al. 1998). The quality of eddies produced at different resolutions with respect to the RRD demonstrates the ability of the model to reproduce mesoscale eddy mixing, which is essential for transport of heat, freshwater, and biogeochemical constituents into the global ocean (Dutay et al. 2002, Gnanadesikan et al. 2004, Siegenthaler 1983).
A common way to assess the capability of an ocean model to produce an accurate eddy climate is via the classic eddying-double gyre benchmark (Berloff et al. 2002, Figueroa & Olson 1994, Holland & Lin 1975, Poje & Haller 1999, Straub & Nadiga 2014). The Simulating Ocean Mesoscale Activity (SOMA) test case (Wolfram et al. 2015) mimics a strongly eddying-double gyre similar to the North Atlantic Gulf Stream. Here, the SOMA benchmark is used to investigate the ability of local mesh refinement to resolve the eddy climate relative to a uniform, high-resolution-mesh simulation.
A high-fidelity capability for simulating eddies with a variable-resolution ocean is demonstrated in the DOE-ocean model, as validated against the SOMA benchmark.
The ocean component of the U.S. Department of Energy (DOE)’s Energy Exascale Earth System Model (E3SM) is the Model for Prediction Across Scale - Ocean (MPAS-O; Ringler et al. 2013). A new variable-resolution-mesh configuration of MPAS-O has been used for the SOMA test case to understand the capability of local mesh refinement to resolve a strongly eddying ocean climate. SOMA has been configured in MPAS-O at a uniform high resolution of 4km, as in Wolfram et al. (2015), and with variable-resolution using local mesh refinement ranging from 32 to 4km (herein). Here we assess the accuracy of the variable-resolution-mesh approach relative to the uniform, high-resolution approach.
MPAS-O’s ability to simulate mesoscale eddy variability using local mesh refinement and to produce a solution with the fidelity of uniform high resolution provides confidence that E3SM can provide improved ocean simulations at lower computational cost. Below, we assess SOMA simulation fidelity both qualitatively (by visual inspection) and quantitatively, using the coefficient of determination r2 (Crow et al. 1960).
3RD QUARTER METRIC – COMPLETED
Antarctica is surrounded by ice shelves that alter ocean properties and circulation through melting, freezing, and calving processes. These shelves extend over large ocean cavities, where warm waters melt the ice, in turn altering the shelf geometry and dynamics, with impacts on Antarctic-sourced sea-level rise (Asay-Davis et al., 2017). Ice shelves and ocean cavities in Antarctica have previously not been included in earth system models (ESMs) due to added model complexity and lack of observations for validation (Fyke et al., 2018). Ice shelves vary from 100 to 2000 m in thickness, and protrude over the ocean 30 to 600 km from the grounding line (the point where ice shelves come afloat). The two largest ice shelves, the Ross and Filchner-Ronne, are each the size of France. Waters from the Antarctic continental shelf flow into the ocean cavities below ice shelves, melting the ice and increasing in buoyancy because of the influx of freshwater (Figure 1a). The meltwater plume then ascends the ice-shelf base to the open ocean where it affects regional ocean properties. These sub-ice-shelf processes alter the temperature and salinity impacting ocean currents and mixing, which through the process of “water mass transformation” (Abernathey et al., 2016) eventually impact global ocean properties.
The U.S. Department of Energy’s Energy Exascale Earth System Model (E3SM) can now be run in configurations that include ocean circulation within cavities below ice shelves so that these processes may be modeled explicitly, compared against observations, and then used in a predictive sense. This will lead to a better understanding of the potential for long-term changes in ocean circulation, ice shelf destabilization, and sea level rise. Here, we demonstrate that global simulations using high spatial resolution lead to improved representation of ocean properties proximal to and beneath ice shelves, leading to simulated sub-ice shelf melt rates that are also closer to observed values.
This research improves upon standard IPCC-class simulations of the earth system by other groups, which do not include ocean cavities below ice shelves (Fyke et al., 2018). Rather, other models assume that the ocean domain ends as a vertical wall at the ice shelf edge, and the addition of basal melt water is simplistically included in the continental run-off parameterization at the ocean surface (Large and Yeager, 2009). Thus, standard models ignore the overturning circulation, the freshwater and salinity exchanges at depth beneath ice shelves, and the resulting feedbacks with regional ocean properties that follow from explicitly simulating melting and refreezing processes below the shelves.
The ocean component of E3SM, the Model for Prediction Across Scales (MPAS-Ocean, Ringler et al., 2013, Petersen et al., 2018), now has the capability to include ice shelf cavities within a global ocean domain and fully-coupled ESM simulations. The implementation relies heavily on the arbitrary Lagrangian-Eulerian vertical grid (Petersen et al 2015, Reckinger et al., 2015) and also required improvements for treating highly-tilted ocean layers, new methods for stable initialization of the ocean model under ice shelves, the addition of boundary layer physics for computing fluxes at the ice sheet and ocean interface, and the coupling of freshwater and salinity fluxes between the ice sheet and ocean at depth.
4TH QUARTER METRIC – COMPLETED
Demonstrate ability to improve simulation of ocean sea-surface temperature using an eddy-resolving high-resolution E3SM-MPAS-Ocean configuration as compared to an eddy-parameterized low-resolution E3SM-MPAS-Ocean configuration
Sea-surface temperature (SST) plays a major role in the Earth’s hydrologic cycle by affecting the exchange of water, heat, and momentum between the ocean, cryosphere, and atmosphere. As the climate warms, changes in SST will likely modify the location, frequency, and intensity of precipitation. In the ocean, it is expected that a high-resolution simulation (e.g., 10-km grid scale) will produce a more realistic evolution of SST due to inclusion of smaller flow scales and the need for fewer parameterizations of unresolved processes (such as mesoscale eddy effects). Previous studies (e,g,, Bryan et al. 2010, Small et al. 2014) have shown that the ability of an ocean model to realistically resolve mesoscale frontal structures in the SST can have a significant effect on, for example, global precipitation, suggesting that fine-resolution simulations may be necessary to produce high-fidelity climate projections. In this report, the improved realism of SST in high-resolution (HR) simulations compared to low resolution (LR) is investigated and confirmed using the DOE E3SM in fully coupled (ocean-atmosphere-cryosphere-land) mode.
In order to investigate the effects of resolution on the SST, the low-resolution E3SM (of order 100 km for atmosphere/land and 60-30 km for ocean/ice) and high-resolution (of order 25 km for atmosphere/land and 18-6 km for ocean/ice) coupled simulations are analyzed and compared. The high-resolution E3SM Water Cycle experiment has currently completed 30 years, which will be compared with a companion low-resolution case of the same duration. Both simulations are time-slice experiments using perpetual 1950 greenhouse gas (GHG) and aerosol forcing. The ocean and sea ice states of both simulations were initialized from a short spin-up (3 years) of a CORE-II (observed-atmospheric; Large and Yeager, 2004) forced run. SST characteristics for the LR and HR ocean configurations are compared with historical climatological observations as well as satellite-derived products. Based on several metrics, the HR case is found to be more realistic than LR, although HR appears to significantly over-represent the SST variability in regions of strong eddy activity.