Conditional simulation of Thwaites Glacier (Antarctica) bed topography for flow models: Incorporating inhomogeneous statistics and channelized morphology

Abstract

AbstractThwaites Glacier, Antarctica, is experiencing rapid change and its mass could, if disgorged into the ocean, lead to ∼1 m of global sea-level rise. Efforts to model flow for Thwaites Glacier are strongly dependent on an accurate model of bed topography. Airborne radar data collected in 2004/05 provide 35 000 line km of bed topography measurements sampled every 20 m along track. At ∼15 km track spacing, this extensive dataset nevertheless misses considerable important detail, particularly: (1) resolution of mesoscale channelized morphology that can guide glacier flow; and (2) resolution of small-scale roughness between the track lines that is critical for determining topographic resistance to flow. Both issues are addressed using a conditional simulation that merges a stochastic realization (an unconditional simulation) with a deterministic surface. A conditional simulation is a non-unique interpolation that reproduces observed statistical behavior without affecting data values. Channels are resolved in the deterministic surface using an interpolation algorithm designed for sinuous channels. Small-scale roughness is resolved using a statistical analysis that accounts for heterogeneity, including an abrupt transition between ‘lowland’ and ‘highland’ morphology. Multiple realizations of the unconditional simulation can be generated to sample the probability space and allow error characterization in flow modeling.