GEOPHYSICAL RESEARCH LETTERS, VOL. ???, XXXX, DOI:10.1029/, Modeled and observed fast flow in the Greenland ice sheet Ed Bueler Dept. Mathematics and Statistics, University of Alaska, Fairbanks, Alaska, USA Constantine Khroulev Geophysical Institute, Fairbanks, Alaska, USA Andreas Aschwanden Arctic Region Supercomputing Center, Fairbanks, Alaska, USA Ian Joughin Polar Science Center, Applied Physics Lab, University of Washington, Seattle, Washington, USA Ben E. Smith Polar Science Center, Applied Physics Lab, University of Washington, Seattle, Washington, USA Satellite surface velocity measurements covering 86% of the Greenland Ice Sheet were used to evaluate a prognostic ice dynamics model on a 3 km grid. A small, but system- atic, exploration of the parameter space considered changes in just three critical model parameters, describing ice soft- ness, nonlinear basal rheology, and basal water pressure, re- spectively. Parameter combinations were evaluated by com- paring the modeled and observed surface speeds. Best fit to the observed distribution of fast flow occurred with no en- hancement of ice softness, nearly-plastic basal material, and high basal water pressure under fast-flowing ice. The use of a standard amount of ice flow enhancement was seen to generate a distribution of fast flow which is fundamentally different from that in the observed flow, while a specific pa- rameterization of basal sliding generated a close-to-observed distribution. 1. Introduction Recent studies have led to a better understanding of the present condition of the Greenland ice sheet (GrIS). Im- proved surface temperature [Fausto et al., 2009] and pre- cipitation [Burgess et al., submitted] maps are available, for example, as are the horizontal surface velocities for a ma- jority of the GrIS area [Joughin et al., submitted]. In terms of the response of the ice flow to possible cli- mate changes, however, critical quantities like ice softness and basal material strength remain poorly-constrained on a whole-sheet scale. Ice flow models are therefore needed to understand even the present flow state of the GrIS. If reli- able predictions for future behavior are to be made, model- ers must connect relatively-rich present-day surface observa- tions to a carefully-chosen set of parameters controlling the modeled three-dimensional ice fluid and its basal sliding. A primary connection is the “inversion” of surface veloci- ties to compute basal stress. Typically hundreds to millions of adjustable spatially-distributed basal parameters are set in such procedures [e.g. Joughin et al., 2004]. Though such inverse modeling is vital to understanding ice flow physics, Copyright 2009 by the American Geophysical Union. 0094-8276/09/$5.00 and potentially so to forecasting ice sheet behavior, it raises the concern of model error, which is to say the error from fitting an inappropriate model to the data. Inversion pro- cedures might use ice temperatures from a time-dependent model to determine ice softness, for example, but model er- ror occurs if the inversion yields a description of ice flow which is greatly-different from that which determined the temperature field. Inversion may be used for the initializa- tion of the future runs of prognostic models, but the avoid- ance of model error requires evidence that the prognostic model can do a reasonable job without inversion. For these reasons, we asked in this study how a model can match the observed surface velocities using just three scalar parameters. Our goal was to explain, by example, what kind of prognostic model might best supply ice temperature and basal melt rate to an inversion of surface velocities, or supply velocity boundary conditions to a regional ice flow model. We used a physically-based formulation of ice flow and basal sliding, and we imposed the present-day climate. We created a small set of century-length model runs on a 3 km grid, and we evaluated the (transient) final state of each model run by comparing its surface velocity to newly- assembled surface velocity measurements. 2. Model The open source Parallel Ice Sheet Model (PISM) has a unified treatment of stresses, sliding resistance, and ther- modynamics, with the same physics applied at all points of the ice sheet. It uses a new hybrid membrane- and shear- stress balance scheme for ice flow [Bueler and Brown , 2009]. Additionally, for GrIS the runs here, a new conservation of energy scheme determined the ice temperature in cold ice, the liquid water fraction in temperate ice, and the basal melt rate, all from a single enthalpy field [Aschwanden and Blatter , 2009]. The basal material of the GrIS is actually a spatially- heterogeneous combination of liquid water, deforming wet- and/or-dirty ice, deforming till, and cold-or-temperate ice sliding over, or frozen to, bedrock. The spatial distribution of these cases is not known in detail for the whole GrIS. All of these conditions were modeled by one basal-sliding power law. This power law permits many interpretations, includ- ing classical sliding [Weertman , 1964] and till deformation [Clarke , 2005]. We choose to describe our basal deforma- tion model as a partially-saturated, nearly-plastic till with 1