Inversion of a glacier hydrology model

Abstract

The subglacial hydrologic system exerts strong controls on the dynamics of the overlying ice, yet the parameters that govern the evolution of this system are not widely known or observable. To gain a better understanding of these parameters, we invert a spatially averaged model of subglacial hydrology from observations of ice surface velocity and outlet stream discharge at Kennicott Glacier, Wrangell Mountains, AK, USA. To identify independent parameters, we formally non-dimensionalize the forward model. After specifying suitable prior distributions, we use a Markov-chain Monte Carlo algorithm to sample from the distribution of parameter values conditioned on the available data. This procedure gives us not only the most probable parameter values, but also a rigorous estimate of their covariance structure. We find that the opening of cavities due to sliding over basal topography and turbulent melting are of a similar magnitude during periods of large input flux, though turbulent melting also exhibits the greatest uncertainty. We also find that both the storage of water in the englacial system and the exchange of water between englacial and subglacial systems are necessary in order to explain both surface velocity observations and the relative attenuation in the amplitude of diurnal signals between input and output flux observations.