E3SM V3.0 New Features

Owned by Robert Jacob
Last updated: Mar 04, 2024

 

First release of version 3.0 of the Energy Exascale Earth System Model

Default grid configuration for the low-resolution configuration of v3 is a “trigrid” consisting of

  • Atmosphere on ne30pg2 (physics grid). Same as v2.

  • Ocean, sea-ice on new Icosahedral 30 km mesh with Ice Shelf Cavities (IcoswISC30E3r5). New in v3

  • Land and river run-off on a common 1/2 deg lat-lon grid (r05). New default in v3.

The following compsets are supported at the low-resolution detailed above: F1850, F1950, F2010, F20TR, WCYCL1850, WCYCL1950, and WCYCL20TR

The atmospheric component remains EAM. Its vertical resolution has been increased from 72 layers in version 2 to 80 layers in version 3. Atmospheric physics have had significant additions and changes since version 2:

  • Added interactive gas chemistry in stratosphere and troposphere with 32 transported species - the UCI chemistry mechanism

  • Update from MAM4 to MAM5. Additional size mode for prognostic treatment of stratospheric sulfate aerosol from explosive volcanic eruptions

  • Added explicit representation of secondary organic aerosols (SOA) sources from multigenerational chemical oxidation of biogenic, anthropogenic, biomass burning volatile organic compounds and chemical sinks including photolysis and fragmentation

  • Improved aerosol wet removal processes to reduce the overestimation of aerosols and thus aerosol forcing

  • Improved numerical coupling of aerosol emission, vertical mixing and dry deposition

  • Re-tuned Dimethyl sulfide (DMS) emission, number of monolayers for black carbon (BC) aging, and hygroscopicity of particulate organic matter (kPOM) to improve the representation of aerosol direct forcing.

  • Added a new dust emission scheme to enable the time-varying soil erodibility for dust mobilization and to include high-latitude dust sources

  • Added Predicted Particle Properties (P3) for stratiform clouds to improve representation of ice microphysical processes and aerosol-cloud interactions

  • Added sophisticated cloud microphysics for convective clouds, which allows aerosols to impact convective processes through convective microphysics

  • Added Multiscale Coherent Structures Parameterization (MCSP) to represent the effects of organized mesoscale convective systems

  • Made cloud base mass flux adjustment in the ZM deep convection based on large-scale dynamics to incorporate the influence of large-scale circulation on deep convection

  • Refined lower stratospheric vertical grid and surrogate-accelerated parameter optimization that leads to a considerable improvement of simulation of Quasi Biennial Oscillation (QBO)

  • Revised surface gustiness formulation for coupling with land and ocean.

  • Adopted rougher topography due to improved terrain-following pressure-gradient and hyper-viscosity formulations.

  • Added online conditional diagnostic capability for sampling and budget analysis for the purpose of facilitating process-level model evaluation and debugging

  • (not on by default) MOSAIC for explicitly representing nitrate and ammonium aerosols and interactions

The EAMxx (aka SCREAM) code is included as an optional atmosphere component in v3. Configurations with EAMxx are not yet supported.

The land component is ELM. The default configuration for ELM has been switched from vegetation leaf area being specified as a static monthly time series for each gridcell based on satellite data (the satellite phenology or SP mode) to a dynamic vegetation growth model that predicts leaf area dynamics and vegetation height in each plant functional type for each gridcell, and how those change over seasons and years in response to the simulated climate. This implementation uses prognostic coupled carbon, nitrogen, and phosphorus cycles (C-N-P) for vegetation and soil biogeochemistry. It uses the relative demand (RD) approach to resolve competition between plants and soil microbes for available nutrients. It uses the converging trophic cascade (CTC) model for organization of plant litter and soil organic matter pools. The model also includes:

  • A computationally efficient radiative transfer parameterization (TOP), which is based on the Monte Carlo simulations, accounts for subgrid topographic effects on solar radiation.

  • Changed the order of sublimation and dew on snow depth calculation to ensure that snow depth is not accidentally set to a negative value.

  • Updated stomatal resistance to properly account for sunlit or shaded conditions

  • New infiltration scheme (off by default)

  • New Flow of Agricultural Nitrogen scheme (off by default)

The river model is still MOSART. An optional sediment scheme has been added

The ocean component remains MPAS-Ocean. Major change since version 2.1 include:

  • Switched from coupling the mid-layer bottom level pressure of EAM to MPAS-Ocean to using extrapolated pressure at mean sea level, thereby greatly reducing bias in coastal sea surface height and the associated dynamic gradient term for sea ice.

  • Switched to a global equal area 30km ocean-sea ice mesh including Antarctic ice shelf cavities.

  • Improvements to the Redi isopycnal mixing scheme.

  • Improved computational efficiency of the ocean component by changing the baroclinic time stepping method

 

The sea-ice component, MPAS-Seaice, has had many improvements since 2.1:

  • Implemented the CICE Consortium’s Icepack as the default sea ice column physics, including a floe size distribution (off by default). The old column physics (colpkg) remains in the code temporarily while Icepack’s biogeochemistry is being updated, but is not being maintained.

    • Additional BFB changes in the Icepack implementation: a) refactored code to enable bit-for-bit global sums in log output for different decompositions and processor counts; b) added a shortwave redistribution option including sw_redist, sw_frac, and sw_dtemp namelist; c) updated sensible and latent heat bulk formula options set by atmbndy_in. Valid values are similarity (default), mixed, and constant (BFB).

    • Other Icepack changes are BFB except in rare instances that they are intended to fix: a) Moved shortwave absorbed in negligible snow layers to surface; b) added abort in ITD for rare case where (hicen_init(n+1) - hicen_init(n)) <= 0 ; c) modified error tolerance for rare mushy thermo failure.

  • Switched to high-frequency sea ice coupling as the default surface stress flux option for all meshes.

  • Reduced the maximum number of sea ice atmospheric turbulent flux iterations from 10 to 5.

  • Decreased the dynamics minimum concentration sea ice cutoff from 10-3 to 10-11 and the ice mass per unit area cutoff from 10-2 to 10-10.

  • Multiple non-BFB bug fixes were made to the code, including a) an erroneous velocity component used for high-frequency atmospheric flux coupling; b) a floating-point exception in MPAS-SI incremental remap code; c) errors in the 5-band delta-Eddington radiation scheme; d) order of operations in albedo calculations for restart consistency; e) thin ice/snow treatment of enthalpy; f) snow interactions with ponds; g) missing snow melt in the melts array; h) undefined meltsliq values; i) initializations for snow mass, biogeochemistry, aerosols, and zsalinity; j) erroneous divisions by aice in the old aerosol scheme (tr_aero); k) call locations of frzmlt_bottom_lateral and neutral_drag_coeffs.

  • Reduced standard monthly sea ice output by removing sea ice age, first year ice tracers, and level ice volume and area.

  • Added or updated checks on configuration settings. Checks on values necessary for coupled runs are executed in mpas_seaice_initialize.F, and consistency checks for icepack are in mpas_seaice_icepack.F. Consistency checks for colpkg were not updated. Among others, the changes a) removed config_couple_biogeochemistry_fields (not used); and b) renamed seaice_check_constants to seaice_check_constants_coupled.

  • Removed ingestion of atmospheric pressure in sea ice coupling fields (not used).

 

The land-ice component remains MPAS-Albany-landIce (MALI). The version of MALI has been updated to include higher-order advection and time integration. The land-ice component is not yet supported in production configurations with v3.

 

Version 3 of E3SM includes as an optional wave model, WAVEWATCH III. This option is not yet supported in any production simulations.

  • A global unstructured waves mesh with 225km resolution in the open ocean and refinement to match the ocean resolution at coastlines.

  • Spectral resolution options use 50, 36 (default), or 25 frequencies (all with 36 directions).

  • An unresolved obstacle source term improves significant wave height accuracy by accounting for dissipation due to subgrid scale islands.

  • The wave mesh uses a rotated pole coordinate system to avoid the north pole singularity, allowing for complete coverage of the Arctic. A basic damping scheme for waves in sea ice is used for the time being.

  • The wave model couples with the ocean to provide a Langmuir mixing enhancement factor via the Stokes drift induced by waves.

  • WAVEWATCH III is currently only available when using the V2 low resolution ocean mesh, but not yet for the current V3 low resolution Icos30 ocean configuration.

 

The coupler remains cpl7/MCT.

  • A new property-preserving map procedure enables high-order flux maps between components for selected fields. In v3 low-resolution cases, applied to all atmosphere-to-surface fluxes except precip.

  • The calculation of gustiness between atmosphere, ocean and land was rationalized. Gustiness is off for sea ice.

The test suites were updated to support one ultra-low resolution: ne4pg2 with oQU480. ne11 and oQU240 are no longer tested. All tests with productions resolutions now use the v3 low resolution. The build system now follows CMake conventions more closely.

Default processor layouts for the low resolution configuration are available on Perlmutter and other DOE platofrms.