How to evaluate model-derived deglaciation chronologies: a case study using Antarctica

Abstract

We address the evaluation of model-derived deglaciation chronologies using observational data. The study has been undertaken using the Antarctic ice sheet as the focus, however, the issues addressed and the methods described are applicable to the evaluation of other ice sheet reconstructions. Within this context, we present an initial database of observational data for constraining Antarctic ice sheet deglaciation chronologies (AntICEdat). The database constrains present-day ice sheet configuration, relative sea level, past ice thickness and grounding line retreat and is made available as a spreadsheet. We consider the non-trivial translation of an observation to model-applicable constraint data. Through observational error models and data-weighting we address the main issues that arise from evaluating modelled reconstructions – generated with a glacial systems model that has, like all such models, inherent structural deficiencies – using heterogeneous observational data. The evaluation method uses observational error models to quantify model to observational misfits that also incorporate the measurement uncertainties for each data-point. The data-point misfits are adjusted by data-weighting and combined to generate a score for the model output. As such, different chronologies can be evaluated and compared. We examine the sensitivity of the score to the different data-types and associated weighting using model-derived reconstructions. In addition, suggested reporting requirements are proposed to ensure that maximum value can be extracted from observational data.