Cook et al. describe in their manuscript a three-step approach to investigate controls on the crevasse pattern observed at Totten Ice Shelf surface near its terminus. First they use the particle-based model HiDEM (first application of a discrete element model to an Antarctic ice shelf) for the given terminus geometry and find no adequate representation of the observed crack pattern, in particular with regard to the kilometer-wide across-flow features. In a second step the authors apply HiDEM to the grounding zone of the Totten Ice Shelf and find formation of narrow basal (tensile and shear) basal crevasses at pinning points in the grounding zone.
In order to understand how those crevasses control the fracture pattern at the terminus and eventually the formation of rifts and icebergs, Cook et al. assume that those grounding zone crevasses can be advected with the ice shelf flow to the terminus while they evolve in shape and size. In a third step two additional independent models are used to simulate the evolution of the shape and size of basal crevasses along that path downstream. Crevasse widening due to ice-internal deformation is represented by the full-Stokes finite-element continuum model Elmer/Ice, while for the basal-melt driven crevasse growth they use a generic non-linear partial differential equation (of type KPZ univerity class).
The paper is logically structured with a new separate discussions part, also the steps taken to answer the research question are clearer stated now in the revised manuscript. It is a valuable contribution to the Cryosphere community presenting applications of the latest model developments to an Antarctic hot spot at the interface of ice, ocean and fracture processes. However, it is difficult for the reader to actually learn from the results about the dominating processes involved.
As the HiDEM model is computational extremely expensive the sensitivity of the induced fracture pattern to chosen parameters (such as critical yield stress) or initial/boundary conditions (initial bond failure, setup orientation with respect to artificially added crevasses, mentioned as lattice bias, uncertainty in bedrock data and basal friction coefficient, isostatic pressure instead of longitudinal stresses at the boundary) can not be tested for easily. The authors honestly name most shortcomings of this black box, point to previous studies for parameter choices, but the scientific message remains somewhat weak.
The HiDEM model validation is based on the key assumption, that transversal crevasses at the calving front control calving processes and hence the iceberg sizes. The observed iceberg (fragment) size distribution, however, does not clearly support this hypothesis, as the iceberg sizes are rather linked to locally formed crevasses, „perpendicular to the calving front“, independent of transversal crevasses. The authors state that their study focusses rather on crevasse orientation than crevasse density, but how does orientation then translates into calving rates or iceberg sizes? And how can across-flow features change the calving style?
As HiDEM produces narrow crevasses in the grounding zone, which can hardly been observed, but cannot produce broad across-flow crevasses near the terminus, the second part of the study investigates two possible mechanisms that could explain basal crevasses widening during the transition time of about 100 years from crevasse source to the ice shelf terminus. As HiDEM seems not capable to simulate crevasse evolution over this time span, the authors choose two other model approaches. The widening of initial basal crevasses is actually represented well in both Elmer/Ice and the KPZ model, although the ice-dynamical model is run only for 1 year instead of 100 years as in the KPZ model. Is it the computational cost that prohibits direct coupling of the two models? In contrast, the deepening of the basal crevasse is reproduced only by the melt model, while strain thinning in the Full-Stokes model leads rather to a lowering of crevasse depth. The authors mention this discrepancy with previous coupled model studies that find refreezing and hence a lowering in crevasse depth, while other ice dynamical model studies including effects such as stress concentration can enhance crevasse deepening. A result of such a conflict should be a critical investigation of included terms, i.e. the physical processes they stand for, rather than stating the applied model suffers from „significant oversimplifications“. A possible association of the driving term could be the melt-driven flow which dependent on the sub-surface-slope as in the Holland and Jenkins, 1999 model. How could the diffusion-like first term (called here „bending stiffness“) be motivated in the KPZ model? Such a constructive discussion would also help to motivate the plausible range of involved parameters (all set to 1m/a here). What can a model with several unconstrained parameters, that has been tuned to reproduce one observed pattern, tell us about the involved dominant processes and how these may change for different boundary conditions?
Holland, David M., and Adrian Jenkins. "Modeling thermodynamic ice–ocean interactions at the base of an ice shelf." Journal of Physical Oceanography 29.8 (1999): 1787-1800.
Fig 2 caption „perpendicular to the calving front“ is somewhat misleading here, as the front line is highly heterogeneous. Better stick to „parallel to the flow“ with a main flow direction indicator or vector. Also in p.8 l.12.
Fig. 2b and Fig.6: If every pixel in this plot is associated with one surface particle, what is then the process leading to the thinning of large blocks of the ice (shown in grey)? This is not discussed in the text but might be important in the rift process?! How is the distance of initial crevasses motivated, it seems to be of the same magnitude as the indicated width of observed basal crevasses (1-3 km)?
Fig. 3 Produced crevasses seem to form aligned to defined lattice / computational domain? The main flow direction could be indicated by an arrow, to visualize the grounded-to-floating transition.
Fig. 4 Switching colors of model and observations would be more intuitive with regard to initial conditions in blue.
Fig. 7 A power-law relationship such as the theoretical tensile fracture with exponent -3/2 should be represented as a straight line in a log-log-plot?
p.6 l.1 just for consistency should be also „km2“
p.6 l.10 You could indicate that from Eq. 4 follows that the crevasses have 115m depth and spacing between 200m and 600m. This seems rather wide and dense compared to ICECAP observations?
p.7 l.8 „glacier“ or better „floating ice shelf“?
Supplement: „educed“ should be „reduced“