OEP-0003: Surface mass-balance enhancements


Fabien Maussion


Draft - not implemented




We present a list of possible enhancements to the OGGM mass-balance model(s). Each of them can be tackled separately, but it could make sense to address some of them together, since it is quite an involved endeavor.


OGGM’s mass-balance (MB) model is a temperature index model first developed by Marzeion et al. (2012) and adapted for OGGM (e.g. to be distributed according to elevation along the flowlines). The important aspect of our MB model is the calibration of the temperature sensitivity, which is… peculiar to say the least. See Mass-balance models for an illustration of the method.

This method is powerful, but also has caveats (some are listed below). Furthermore, it has not changed since 2012, and could make much better use of newly available data: mostly, geodetic mass-balance for a much larger number of glaciers.

Proposed improvements

Varying temperature sensitivities for snow and ice


Currently, the temperature sensitivity \(\mu^{*}\) (or “melt factor”, units mm w.e. yr-1 K-1) is the same all over the glacier. There are good reasons to assume that this melt factor should be different for different surface conditions.

One relatively simple way to deal with it would be to define a new model parameter, snow_melt_factor, which defines a temperature sensitivity for snow as \(\mu^{*}_{Snow} = f \, \mu^{*}_{Ice}\) with \(f\) constant and somewhere between 0 and 1 (1 would be the current default).


The implementation is not as straightforward as it sounds, but should be feasible. The main culprits are:

  • one will need to track snow cover and snow age with time, and transform snow to ice after some years.

  • the calibration procedure will become a chicken and egg problem, since snow cover evolution will depend on \(\mu^{*}\), which will itself depend on snow cover evolution. Possibly, this will need to a relatively costly iterative procedure.

Calibration / Validation

This will introduce a new parameter, which should be constrained. Ideally, it would be fit to observations of MB profiles from the WGMS.

Find a sensible algorithm to avoid the interpolation of t*


Make use of available geodetic MB data


Use Bayes