The Design Document page provides a description of the algorithms, implementation and planned testing including unit, verification, validation and performance testing. Please read Step 1.3 Performance Expectations that explains feature documentation requirements from the performance group point of view.
Design Document
Expand | ||
---|---|---|
| ||
The first table in Design Document gives overview of this document, from this info the Design Documents Overview page is automatically created. In the table below, 4.Equ means Equations and Algorithms, 5.Ver means Verification, 6.Perf - Performance, 7. Val - Validation
Use the symbols below (copy and paste) to indicate if the section is in progress or done or not started. |
In the table below, 4.Equ means Equations and Algorithms, 5.Ver means Verification, 6.Perf - Performance, 7. Val - Validation, - completed, - in progress, - not done
Page Properties | ||||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ||||||||||||||||||||||
|
Title: O_24_LI MPAS Land Ice in ACME Design Doc
Requirements and Design
ACME Ocean and Ice Group
Date: 2015-9-23
Summary
Requirements
Requirement:
Support for Trilinos and Albany third-party libraries in ACMEConservation of mass, energy, and momentum
Date last modified: 2015-9-23
Contributors: Stephen Price
The MPAS Land Ice model requires linking to an "external dycore", the FELIX-Albany dycore developed under the PISCEES project (Tezuar et al., 2015a; 2015b, below). The dycore is built using the the Trilinos and Albany software libraries, which therefor must be made available to the ACME build system.
land ice model will conserve mass, energy, and momentum.
Requirement: Accurate marine ice sheet dynamics
Date last modified: 2015-9-23
Contributors: Stephen Price
The momentum and mass conservation components of the dynamical core will provide an accurate simulation of marine ice sheet dynamics. Specifically, the simulation of retreat and advance of the grounding line (position at which ice goes afloat due to buoyancy) will be verified according to standard benchmark test cases.
Requirement: Iceberg calving
Date last modified: 2015-9-23
Contributors: Stephen Price
Requirement: blahThe momentum and mass conservation components of the dynamical core will allow for the retreat and advance of floating ice margins (the fronts of ice shelves) through the implementation of iceberg "calving" physics.
Requirement: Support for optimal initial conditions
The initial condition for the ice sheet model should be both a good representation of present-day observations of the ice sheet state (e.g., the geometry and velocity fields) and should be approximately equilibrated with present-day forcing from the climate model (in order to avoid undesirable ice sheet model transients that could mask actual trends of interest).
Algorithmic Formulations
Design solution: N/A
Date last modified: 2015-9-23
Contributors: Stephen Price
The design solutions are primarily software and build-system related and as such, do not require any particular algorithmic solutions
Algorithmic formulations for all of the above requirements are discussed in their respective design documents, already published papers, or existing model documentation (referenced as appropriate in the design solutions / implementation sections below and listed completely in the References section below).
Design and Implementation
Implementation
: Support for Trilinos and Albany third-party libraries in ACME: Conservation of mass, energy, and momentum
Date last modified: 2015-9-23
Contributors: Stephen Price
- Conservation of mass: Conservation of mass is expressed through the equation for ice thickness evolution. It uses an explicit forward-Euler scheme in time and explicit upwind-based advection in space. Within MPAS-LI, this is implemented in a way consistent with it's implementation in CISM2.0. Additional details on that scheme can be found in the model documentation links in the References section below.
- Conservation of energy: Conservation of energy is expressed through the advective-diffusive heat equation. Within MPAS-LI, this will be implemented largely in the form of a column-physics package copied over from the CISM2.0 model, with additional hooks to MPAS-LI advection routines. Additional detail on the heat balance in CISM2.0 can be found in the model documentation links in the References section below /wiki/spaces/OCNICE/pages/34113116.
- Conservation of momentum: Conservation of momentum is expressed by the (3d) solution of the first-order accurate approximation to the Stokes equations for ice flow. Discretization of the governing nonlinear PDEs uses the finite element method on unstructured, Delaunay meshes (the dual mesh to an MPAS CVT mesh). Within MPAS-LI, the discretization and solution of the resulting linear and nonlinear systems of equations is handled by the FELIX-Albany solver, which is discussed in detail in Tezuar et al. (2015a, 2015b) in the References list below. Considerations for Including the relevant solver libraries in ACME are discussed in a separate code review document.
Implementation: Accurate marine ice sheet dynamics
Date last modified: 2015-9-23
Contributors: Stephen Price
This item is covered by the code review item O_21_LI Third Party Support for MPAS-LI within ACME Design Document.
Accurate simulation of marine ice sheet dynamics requires the following:
- a higher-order momentum balance solver (one that includes the effects of "membrane" stresses, in addition to resolving vertical shear stresses)
- adequate mesh resolution, in order to resolve stress transitions in the very narrow boundary layer where ice transitions from grounded to floating
Implementation: Iceberg Calving
Date last modified: 2015-9-23
Contributors: Stephen Price
Implementation of iceberg calving within MPAS-LI has it's own design document, /wiki/spaces/OCNICE/pages/20809782.
Implementation: Support for optimal initial conditions
Date last modified: 2015-9-23
Contributors: Stephen Price
Because of the long timescales associated with equilibrium of processes internal to ice sheets (primarily thermal equilibration), standard climate model "spin up" methods are problematic for ice sheet models in terms of providing optimal initial conditions and / or initial conditions that are in quasi-equlibrium with a given climate forcing. For a number of reasons (discussed in detail in Perego et al., 2014), these goals are best addressed through formal, adjoint-based optimization methods. A brief discussion of the approach is given /wiki/spaces/OCNICE/pages/34113126. In general, we are implementing methods pioneered under the PISCEES project and discussed in Perego et al. (2014).
Planned Verification and Unit Testing
Verification and Unit Testing:
short-desciption-of-testing-hereDate last modified: 2015-9-23
Contributors: (add your name to this list if it does not appear)
How will XXX be tested? i.e. how will be we know when we have met requirement XXX. Will these unit tests be included in the ongoing going forward?
Verification and testing of specific model components / features is discussed in greater detail in the publications, design documents, and model documentation discussed in the individual sections above and listed / linked to below in the References section.
Planned Validation Testing
Validation Testing: short-desciption-of-testing-here
Date last modified: 2015-9-23
Contributors: (add your name to this list if it does not appear)
How will XXX be tested? What observational or other dataset will be used? i.e. how will be we know when we have met requirement XXX. Will these unit tests be included in the ongoing going forward?
Because this is an entirely new component that has never been included in a climate model before, there are no set tools or procedures for use in model validation. Under the PISCEES project, we are developing new methods and frameworks for validating ice sheet models as coupled components of climate models and we will adopt these for validation within ACME as appropriate. In the short-term, the metrics and diagnostics discussed on the /wiki/spaces/OCNICE/pages/1867925 will be used to validate ice sheet model output.
Planned Performance Testing
Date last modified: 2015-9-23
Contributors: (add your name to this list if it does not appear)
How will XXX be tested? i.e. how will be we know when we have met requirement XXX. Will these unit tests be included in the ongoing going forward?
References
Performance of the land ice model is almost entirely controlled by the performance of the momentum balance solver. The performance of the FELIX-Albany momentum balance solver in MPAS-LI is extensively documented in Tezaur et al. (2015a; 2015b), referenced below. We do not currently have any baselines for it's performance when run as a component of a fully coupled climate model and initial baseline performance metrics will need to be captured first before improvements can be made. Presently, we anticipate being able to simulate (order) one model year per wall clock hour on between 4000 and 6000 processors (based on standalone model performance on Edison for a var. res. (14-4km) MPAS-LI mesh run on ~4 k procs).
References
- Perego, M., S. Price, and G. Stadler, 2014: Optimal initial conditions for coupling ice sheet models to Earth system models. … of Geophysical Research: Earth …, doi:10.1002/(ISSN)2169-9011. link
- Tezaur, I., M. Perego, A. Salinger, R. Tuminaro, and S. Price. 2015a. Albany/FELIX: a parallel, scalable and robust, finite element, first-order Stokes approximation ice sheet solver built for advanced analysis, Geophys. Model Devel., 8, doi:10.5194/gmd-8-1197-2015. link
- Tezaur, I., R. Tuminaro, M. Perego, A. Salinger, S. Price, 2015b: On the scalability of the Albany/FELIX first-order Stokes approximation ice sheet solver for large-scale simulations of the Greenland and Antarctic ice sheets", Numerical and Computational Developments to Advance Multiscale Earth System Models (MSESM)/International Conference on Computational Science (ICCS15), Reykjavik, Iceland link
- MPAS Land Ice Model User's Guide Version 3.0. link
- CISM2.0.5 Documentation link