Subglacial water flow routing v2.0: taking uncertainties and supercooling effects into account

Subglacial flow routing, which simply routs water down the Shreve potential (i.e. the hydraulic potential assuming that water pressure equals a fraction of ice overburden pressure), is probably the most widely used subglacial drainage model.  It is easy to use via many ready-to-use implementations, does not suffer from numerical issues, such as non-convergence, and is fast.

Here I present two improvements to this venerable model: (1) the spatial uncertainties in the input fields (surface and bed topography, water input) and the model parameter (fraction between water pressure and floation pressure) are taken into account by representing them as Gaussian random fields. This a allows to assess the impact of these uncertainties by running Monte Carlo simulations.  (2) the dependence of the ice melting point on the pressure (leading to the so-called supercooling effects) can lead to deflections in the flow directions where descending flow-paths are favoured over ascending ones.  I present a framework on how this effect can be taken into account in subglacial flow routing.

To showcase the improved model, I present results from applying it to catchments of Antarctica which show that the sizes of sub-catchments can have uncertainties of several order of magnitudes.  This demonstrates the need of using such an improved model to make predictions of fluxes, including their uncertainties, into subglacial lakes or at the grounding line.