Quality Control for Community Based Sea Ice Model Development
Using the output from five coupled and uncoupled configurations of the Los Alamos Sea Ice Model, CICE, we formulate quality control methods that exploit common statistical properties of sea ice thickness, which are used to efficiently test for significant changes in model results when the code is altered. Modifications to the code are assessed using criteria that account for the high level of autocorrelation in sea ice time series, along with a skill metric that searches for hemispheric changes in model answers across an array of different CICE configurations.
Quality control of community sea ice codes has, until now, been somewhat subjective. These statistical tests grade new additions and changes to CICE into four categories, ranging from bit-for-bit amendments to significant, answer-changing upgrades. In addition to providing a classification procedure for modifications to existing code, these metrics also provide objective guidance for assessing new physical representations and code functionality.
Understanding whether or not changes in CICE code may also alter the climate of the model can be nontrivial. The CICE Consortium has developed an efficient and automated acceptance testing method for controlling the quality of new contributions to CICE, thereby guarding against inadvertent bugs or numerical inaccuracies. The method exploits statistical properties of sea ice thickness evolution common across a range of sea ice models, and is demonstrated in both stand-alone and coupled model settings. The CICE software and data are publicly available through an open-source repository and data portal to facilitate community involvement and improvement.