Abstract
Nearly 70% of the world's fresh water resources are currently locked up in the ice sheets of Greenland and Antarcticata. Satellite observations reveal that our polar ice sheets are losing ice at an unprecedented, accelerating rate. Since 1992 the polar ice sheets have contributed, on average, 0.6 mm per year to sea level, which is roughly the yearly water usage of the United States. Rising sea levels are expected to cause adverse economic and social effects. To estimate the ice sheets' future contribution to sea level, we need ice sheet models.

The skill of a predictive ice sheet model depends on the quality of data available for testing and on the implemented model physics. Testing, i.e. validation is a critical component of model development, yet notoriously challenging in ice sheet modeling. One approach to validation is hindcasting, i.e. forcing a model with known or closely-estimated inputs for past events and comparing model results to time-dependent observations. In other words I ask the question: "How successful is the state-of-the art Parallel Ice Sheet Model (PISM) in reproducing observed changes over the last two decades?" I show that PISM reproduces some observations while struggling with others. This excercise highlights the importance of observation-based validation to reduce the uncertainty in projections of sea level rise.