Past patterns of glaciation continue to have a crucial impact upon the modern climate system, and dynamical models are increasingly being used to simulate the long-term evolution of ice sheets.  The construction of glacial models is, however, challenging because paleoclimate boundary conditions are constrained mostly by relatively sparse and indirect modern evidence.  The very successful ICE-6G_C (VM5a) model and its predecessors have avoided detailed assumptions about ice physics altogether by selecting hypothetical ice thickness histories based on solutions
of integral equations that determine the effects of surface mass loading upon measurable modern patterns of crustal motion and upon proxy evidence for past sea-level rise.  Glacial modelers are now trying to achieve comparable consistency with observations by tuning ice dynamical parameters, but such efforts tend to achieve local fits by tailoring complex approximations to local glaciological conditions. Representations of the calving fronts and grounding line positions of ice shelves are proving to be particularly difficult to parameterize in a general way.

I will describe a theoretical and computational approach to ice modeling that builds upon the leading-order consistency of the ICE-6G_C (VM5a) reconstruction rather than trying to replace it with the results of purely parameter-driven simulation.  The global solution was used to "nudge" consistently tuned local simulations of the Antarctic and Greenland ice sheets.  Results from these simulations validate the overall approach and add to our understanding of the ice dynamical origins of glacial meltwater pulse 1a (MWP1a).