Physics-based parameterisation framework for basal melting in ice-ocean boundary layers over dynamically stable pycnoclines

TL;DR

This paper introduces a new physics-based framework to improve predictions of ice melting in ocean models by accounting for low-diffusivity layers near ice bases.

Contribution

The novel contribution is a physics-based parameterization framework that addresses overpredicted melting in ocean models due to low-diffusivity pycnoclines.

Findings

Low diffusivity in pycnoclines restricts heat transfer, causing models to overpredict melting.

Reducing boundary layer depth or setting an upper melt rate limit can improve model accuracy.

The proposed framework better emulates observed physics in low-resolution ocean models.

Abstract

Accurate basal melt prediction is crucial for assessing ice sheet stability and sea level rise. Recent observations at eastern Thwaites Glacier reported low melt rates despite warm ocean waters. Weak vertical mixing due to low current speeds and strong density stratification suppresses melting. However, the basal melt parameterization approach in ocean models overestimates the melt rates there. Hence, we revisit the parameterization by applying an ice-ocean boundary current model to a simple horizontal ice base. This setting creates a boundary layer (BL) over a dynamically stable pycnocline. We show that the pycnocline’s low diffusivity restricts heat transfer, causing models to overpredict melting, especially for weaker far-field currents. While reducing the prescribed BL depth can minimize this overprediction in ocean models, a better fix might be prescribing an upper melt rate limit for slower currents. We also propose a physics-based parameterization framework that more accurately emulates physics in models and observations.

A physics-based framework for parameterizing ice-ocean heat exchange in low-resolution ocean models accounts for a low-diffusivity pycnocline beneath the buoyant meltwater layer, thereby avoiding overprediction of melting found within current models.

Figures