| 02 Sep 2026
Ice-sheet evolution controlled by meltwater connectivity across ice-shelf gaps
Abstract. In regions of intense sub-shelf melting, ice shelves can substantially weaken, potentially leading to the formation of gaps within the ice shelf. Using a coupled ice-sheet–meltwater-layer model, we investigate two end-member representations of meltwater transport across such gaps: meltwater either exits the cavity (sink) or remains dynamically connected across ice-free regions (connected). For an idealised ice-sheet–ice-shelf geometry based on the MISMIP+ framework, our results show that gaps form under both moderate (0 °C at depth) and warm (1 °C at depth) ocean forcing, driven by enhanced melting associated with a strong western boundary current. When gaps are treated as sinks, downstream melting is suppressed. In contrast, retaining meltwater connectivity allows heat and momentum to propagate downstream, enhancing further ice-shelf thinning along the western shear margin. This thinning reduces stress transmission between the ice shelf and the grounded ice, resulting in greater ice-volume loss in the connected configuration. The difference in ice-volume loss between the two meltwater representations can be comparable in magnitude to that caused by a 1 °C increase in ocean temperature forcing. These results highlight meltwater connectivity across ice-shelf gaps as an important source of uncertainty in projections of ice-sheet evolution and mass loss.
This paper compares idealized simulation outputs resulting from two different modeling choices of how a sub-ice-shelf mixed layer model can treat gaps in the ice-sheet domain. The original choice in Laddie was to remove excess heat and momentum from the mixed layer upon encountering an ice-shelf gap, while a new modeling choice is to remove no heat and momentum from the layer. The results show that the ice loss is faster when no heat and momentum is artificially removed from the mixed layer.
The newly introduced treatment is certainly more physically appropriate - if heat is removed by sea-ice formation in an ice-shelf gap, the newly formed sea-ice will quickly insulate the layer from further heat loss (how often do you actually observe sea-ice free rifts under freezing (heat loss) conditions?). So while opting not to artificially remove heat is almost certainly an improvement over the previous modeling assumption, this paper doesn't really address and solve a problem relevant to the broader (but still quite small) ice-ocean modeling community or modeling groups. Instead it corrects a rather unphysical modeling assumption made previously by some of the authors, who are also themselves the developers of Laddie. So I feel that the subject of this paper would be more appropriate as a footnote, or supplementary information in the next release of Laddie, than a standalone paper.
Qualitatively, the results, that when no heat and momentum are artificially removed then melting is higher and ice loss over time is larger, are not surprising. At the same time, the quantitative assessment of these effects lacks any realism. Laddie is coupled to an ice-sheet model unable to calve and form rifts, which is why (I assume) the authors talk throughout the paper about ice-shelf gaps rather than rifts, which are a far more plausible ice-shelf hole making mechanism. Have any melt through gaps actually ever been observed, to even justify such an assumption? The inability to calve creates further issues - unphysical looking ice-shelf fronts or gaps - see Figure 3. The calving fronts or extended ice-shelf gaps/rifts produced here would have certainly went on to calve off, which adds to my comment about the lack of realistic quantitative assessment, even in the context of an idealized model domain.
While the authors try to illustrate the relevance of their findings using sensitivity to other variables (temperature) the way it is done is not very fortunate. The statement in the abstract comparing the effect of the different modeling choices to 1 temperature degree change is taken out of context and overstates the impact of this paper. Because of the partly suppressed positive feedback in the original modeling choice, the ice volume decreases at a different rate for each modeling choice - which after some initial years is nearly linear for both cases. As such, following the author's logic, the effect of the modeling choice could be stated as any number of degrees of temperature change, as long as we wait long enough - a statement that is not particularly useful.