Models

ClimaCoupler.jl implements a few simple surface models for use in coupled simulations. These models are provided in the Models module and provide basic ocean and sea ice representations that can be used when more sophisticated component models are not needed or available.

Slab Ocean Model

The slab ocean model (SlabOceanSimulation) is a simple energy balance model for the ocean surface layer. It solves the following energy equation:

\[(h \cdot \rho \cdot c) \frac{dT}{dt} = -F_{\text{turb\_energy}} + (1 - \alpha) \cdot SW_d + LW_d - LW_u\]

where:

• $h$ is the depth of the ocean slab [m]
• $\rho$ is the density of the ocean [kg/m³]
• $c$ is the specific heat capacity of the ocean [J/(kg·K)]
• $T$ is the ocean surface temperature [K]
• $F_{\text{turb\_energy}}$ is the turbulent energy flux (latent + sensible heat) [W/m²]
• $\alpha$ is the ocean albedo
• $SW_d$ is the downwelling shortwave radiation [W/m²]
• $LW_d$ is the downwelling longwave radiation [W/m²]
• $LW_u$ is the upwelling longwave radiation [W/m²] (calculated as $\epsilon \sigma T^4$)

The model assumes a well-mixed surface layer of fixed depth and computes the temperature evolution based on the net energy flux into the ocean. The ocean surface temperature is used to compute upwelling longwave radiation and surface fluxes.

Usage: The slab ocean model is used in:

• SlabplanetMode simulations
• SlabplanetAquaMode simulations

The slab ocean model should not be used with a sea ice model, as it does not account for ice-ocean interactions.

Prescribed Ocean Model

The prescribed ocean model (PrescribedOceanSimulation) uses observed sea surface temperature (SST) data that is read from a file at each timestep. The SST is prescribed and does not evolve based on energy fluxes, making this a "data-driven" ocean representation.

The model includes:

• Prescribed SST from observational data (e.g., HadISST)
• Ocean COARE3 roughness parameterization, which accounts for the dynamic response of ocean surface roughness to wind speed and wave state
• Wind-dependent albedo calculation using ClimaAtmos regression functions
• Surface properties (roughness lengths, albedo) that can be updated based on atmospheric conditions

Since the SST is prescribed, this model does not solve any prognostic equations. It serves as a boundary condition for the atmosphere, providing realistic ocean surface temperatures for atmospheric simulations.

Usage: The prescribed ocean model is used in:

• AMIPMode simulations
• SubseasonalMode simulations

These simulation types are designed for atmospheric model evaluation and prediction, where realistic SSTs are needed but ocean dynamics are not the focus.

Prescribed Sea Ice Model

The prescribed sea ice model (PrescribedIceSimulation) uses observed sea ice concentration (SIC) data while solving an energy balance equation for the ice surface temperature. The model solves:

\[(h \cdot \rho \cdot c) \frac{dT_{\text{bulk}}}{dt} = -F_{\text{turb\_energy}} + (1 - \alpha) \cdot SW_d + \epsilon \cdot (LW_d - LW_u) + F_{\text{conductive}}\]

with the conductive flux:

\[F_{\text{conductive}} = \frac{k_{\text{ice}}}{h} (T_{\text{base}} - T_{\text{sfc}})\]

where:

• $T_{\text{bulk}}$ is the bulk ice temperature [K] (prognostic variable)
• $T_{\text{sfc}}$ is the ice surface temperature [K] (diagnostic, extrapolated from $T_{\text{bulk}}$ and $T_{\text{base}}$)
• $T_{\text{base}}$ is the prescribed temperature at the ice base (sea water temperature) [K]
• $k_{\text{ice}}$ is the thermal conductivity of sea ice [W/(m·K)]
• $h$ is the ice thickness [m]
• Other variables are as defined for the slab ocean model

The sea ice concentration is read from observational data (e.g., HadISST) and updated at each timestep. The ice temperature is constrained to remain at or below the freezing point of seawater. The model assumes a linear temperature profile through the ice layer, allowing the surface temperature to be computed from the bulk temperature and base temperature.

This formulation follows the approach of Holloway and Manabe (1971), where sea ice concentration and thickness are prescribed, and the model solves for the ice temperature.

Usage: The prescribed sea ice model is used in:

• AMIPMode simulations
• SubseasonalMode simulations

These simulation types use prescribed sea ice concentration along with prescribed ocean SSTs to provide realistic surface boundary conditions for atmospheric models.

Models in Extensions

More scientifically-complex component models are provided as package extensions, keeping them out of ClimaCoupler.jl's base dependencies. The models available in each extension are described below.

Land models (ClimaCouplerClimaLandExt)

This extension is loaded when ClimaLand.jl is available and provides two land model options:

BucketSimulation

A simple bucket hydrology and energy balance model (ClimaLand.Bucket.BucketModel) following the Manabe (1969) bucket scheme. This model tracks soil moisture and temperature using a single-layer bucket scheme and is suitable for lower-cost simulations where a full land model is not needed.

• Prognostic variables: temperature, subsurface water content, snow water equivalent
• Spatial discretization: finite difference, independent columns on the sphere
• Time stepping: explicit additive Runge–Kutta
• Land surface albedo: bare ground prescribed from files; snow albedo constant

ClimaLandSimulation

The full integrated land model (ClimaLand.LandModel), which includes soil, canopy, and snow sub-models. This is the recommended land model for high-fidelity simulations.

• Prognostic variables:
  • Soil: internal energy, water content, ice content, carbon content
  • Canopy: temperature, water content
  • Snow: snow water equivalent, snow liquid water, energy
• Spatial discretization: finite difference, independent columns on the sphere
• Time stepping: mixed implicit/explicit (IMEX) additive Runge–Kutta

For full model equations and further documentation, see the ClimaLand.jl documentation.

Note that the integrated land model requires spatially-varying parameters that are ill-defined over ocean/sea ice regions, so it cannot be used in simulations that ignore the land/sea mask (e.g. terraplanet).

Ocean and sea ice models (ClimaCouplerCMIPExt)

This extension is loaded when Oceananigans.jl, ClimaOcean.jl, ClimaSeaIce.jl, and KernelAbstractions.jl are available. It provides:

• OceananigansSimulation: A full ocean circulation model implemented in Oceananigans.jl and initialized from EN4 or ECCO reanalysis data via ClimaOcean.jl. The ocean evolves prognostically in response to atmospheric fluxes. For details on the ocean model physics, see the Oceananigans.jl documentation.

• ClimaSeaIceSimulation: A thermodynamic sea ice model implemented in ClimaSeaIce.jl. Sea ice concentration and thickness evolve prognostically, with ice-ocean heat fluxes computed at the ocean-ice interface at each coupling step.

Atmosphere model (ClimaCouplerClimaAtmosExt)

This extension is loaded when ClimaAtmos.jl is available and provides:

ClimaAtmosSimulation

The CliMA atmospheric model, which handles radiation, turbulence, convection, microphysics, and dynamics. This is the only atmosphere model currently supported by ClimaCoupler.jl.

Dynamical core:

• Equations: non-hydrostatic and fully compressible
• Prognostic variables: density, velocity components, total energy, total specific humidity
• Spatial discretization: spectral element in the horizontal, finite difference in the vertical, cubed-sphere geometry
• Time stepping: implicit-explicit (IMEX) additive Runge–Kutta

Parameterizations:

• Radiation: RRTMGP (Pincus et al. 2019) — a scheme based on RRTM for GCM applications
• Turbulence and convection: Diagnostic or prognostic Eddy-Diffusivity Mass-Flux (EDMF) scheme with prognostic turbulent kinetic energy (Tan et al. 2018, Cohen et al. 2020, Lopez-Gomez et al. 2020)
• Microphysics: 0-moment scheme — cloud condensate removed in-situ with a constant timescale; 1-moment scheme
• Surface fluxes: Monin-Obukhov similarity theory with constant roughness lengths over land and ocean

Model Selection

The choice of component models is determined by the simulation type (mode_name). For a full description of each simulation type and its component model configuration, see Available Simulation Types. Models are selected via configuration files or command-line arguments; see the Input documentation for details.

Julia Environments

ClimaCoupler provides two Julia environments under experiments/, each targeting a different set of simulation modes:

experiments/AMIP

Use this environment for amip, slabplanet, slabplanet_aqua, slabplanet_terra, and subseasonal modes. It omits the Oceananigans-related dependencies (Oceananigans, ClimaOcean, ClimaSeaIce, KernelAbstractions), resulting in a lighter environment with faster precompilation.

julia --project=experiments/AMIP experiments/AMIP/run_simulation.jl

experiments/CMIP

Use this environment for cmip mode, which requires a prognostic Oceananigans ocean model. It is a superset of the AMIP environment with additional ocean and sea-ice dependencies.

julia --project=experiments/CMIP experiments/CMIP/run_simulation.jl --config_file config/ci_configs/cmip_oceananigans_climaseaice.yml

See the Input documentation for the full list of mode_name options.