Validation and Verification

From Interactive System for Ice sheet Simulation
Jump to: navigation, search

Contents

Model Verification and Validation

Assurances that CISM is correctly implemented (verification), and that it is solving the correct problem (validation) are a critical part of establishing the credibility of our claims. We are aware of, and seeking to demonstrate consistency with the following published standards for ice flow. Note that many of these standards fall short of the objective of demonstrating correctness, and instead inter-compare model outputs.

Isothermal Flow Intercomparison: EISMINT I

This has been done and CISM compares favorably. We need to present some graphics here.

Thermomechanically Coupled Flow: EISMINT II

Manufactured Exact Solutions

Buler's exact solutions for isothermal and thermomechanically coupled shallow ice flow.

This is in the hands of Tim Wylie. We hope to place something here soon.

Vertically Integrated Flow, Comparison to Data

Ross Ice shelf intercomparison

To be done immediately after the ISMIP-HOM intercomparison is completed.

Higher Order Isothermal Flow

ISMIP-HOM

Test suite is written for experiments A-D; we have been focusing primarily on experiment A (flow over a bumpy bed) at the moment. The higher-order code from Pattyn falls within one standard deviation of the mean for non-full-stokes models. We have found that the use of an irregular grid spacing in the vertical dimension tends to overestimate velocities. This result is mostly independent of the number of vertical layers used, though the overestimation appears to be less severe using 80 layers rather than 40.

Preliminary results also indicate that, as expected, refinement of the vertical grid affects the accuracy of the results. Decreasing the number of vertical layers from 40 to 20 qualitatively causes a large increase in the error of the maximum velocity observed (quantitative comparison to come). Increasing the number of vertical layers to 80 causes the maximum observed velocity to match the first order mean almost exactly. However, the change from 40 layers to 80 makes less of a difference than a change from 20 layers to 40.

Smaller domain sizes of experiment C are very difficult to run to acceptable convergence using the unstable manifold correction scheme in Pattyn's model. Though convergence can be reached, it is very slow and is usually terminated early by Pattyn's error tolerance adjustment scheme.

Output that is generated by CISM for intercomparison with ISMIP-HOM A
Output that is generated by CISM for intercomparison with ISMIP-HOM B
Output that is generated by CISM for intercomparison with ISMIP-HOM C

Isothermal Greenland

We have computed ice velocities for Greenland using a simple isothermal model with no basal mechanics. A thermomechanically coupled solution is forthcoming, followed by a solution that incorporates basal mechanics derived through inverse modeling.

A simple higher-order velocity computation for Greenland using an isothermal model and no basal mechanics

MISMIP

We are awaiting results from this experiment. They will be presented at the 2009 EGU meeting in Vienna. Shortly afterward, we hope to improve the grounding line position code in CISM, and participate in the experiments.