The sea-level methods¶
Three published ways of turning a change in ice geometry into a change in global mean sea level. All three are written for every run. The spread between them shows how much of a model’s sea-level contribution depends on the accounting rather than on the ice.
Each is a function in the ismip7_scalars.slc subpackage and can be called from your own code.
Volume above flotation (slvaf)¶
The ISMIP6 method. Only ice above the flotation thickness moves sea level when it is lost; the rest already displaces its own weight. The volume above flotation is converted straight to a fresh-water depth:
hf = max(S - B, 0) * ρsw / ρi
Vaf = Σ max(H - hf, 0) * A
slc = -(Vaf - Vaf_ref) / A_ocean * ρi / ρfw
Simple, and the most directly comparable with earlier intercomparisons. It ignores bedrock motion and the density difference between the ice’s fresh water and the sea water it displaces.
Goelzer et al. (2020) (slg20)¶
TC 14, 833-840. Adds two corrections to the volume-above-flotation term:
Potential ocean volume, for bedrock that moves. A bed that rebounds displaces water it used to hold; a bed that subsides makes room for more.
Density, for the difference between the fresh water the ice becomes and the sea water it mixes into.
slc = slc_af + slc_pov + slc_den
The implementation folds the geoid height into the bedrock elevation, which matches the published equations when sea level is fixed at zero. A second function, slc_G2020_publ, restates the paper’s equations in its own notation as a cross-check, and the test suite checks that the two agree.
Adhikari et al. (2020) (sla20)¶
TC 14, 2819-2833. Works in an absolute frame, with bed and sea level both measured from the same ellipsoid, and explicitly accounts for cells that switch between grounded and floating. The other two methods only see grounding-line migration through the flotation criterion.
This method measures the change between two consecutive states rather than the departure from a reference, so a time series has to be accumulated. The processing accumulates from the first historical timestep and subtracts the value at the reference year, so the projection continues the historical curve without a step at the join.
From the ice base (slvafbase, slg20base)¶
With --base-variants, slvaf and slg20 are also computed with the ice base
(the model’s base variable) in place of the bed B in the volume above
flotation, and written as slvafbase and slg20base. This is a trial: the
defaults are unchanged, and the base variants are there to compare with them.
The two differ only in a cell that is partly grounded and partly floating, where thickness and bed are cell means. With the bed, the water under the floating part is counted against the grounded ice, so the cell loses volume above flotation as its shelf thins, though no grounded ice is lost. With the ice base, the cell mean gives the right answer wherever the cell is below sea level. The same holds at the ice front, as long as base is at sea level over open water.
In slg20base only the volume above flotation changes. The potential ocean volume is the space between the bed and sea level, so it still comes from the bed.
There is no sla20base. The grounded and ocean masks of sla20 come from cell means of thickness and bed, so a cell the grounding line crosses counts as all grounded or all floating, and the ice base alone does not change that.
If either run has no base file, the base variants are skipped with a warning and everything else is still written.
Densities and ocean area¶
Every method takes the densities from the model’s params.nc. Using the densities a model was actually run with is what makes its sea-level contribution comparable with another model’s.
The ocean area is the same for every model, 3.625 × 10¹⁴ m² from Gregory et al. (2019). A params.nc may hold its own ocean area, but it is ignored.
Glaciers and ice caps¶
Every sea-level method is calculated twice. The -gic variant leaves glaciers
and ice caps out of the integral, using the iaf2_GIC mask; the plain variant
includes everything on the grid. Both variants are rows in the run-level SLC
CSV; by default, only the plain variant is also written to NetCDF. Which
variant you want depends on whether the glaciers around the ice sheet are
being counted elsewhere in your budget.