A document by Aleah N Sommers. Click on the document to view its contents.

Are the Sermilik Fjord glaciers terminus-controlled, runoff-controlled, or runoff-adapting? Decomposition of glacier speed maps at Helheim, Fenris, Midgard, and Pourquoi-Pas Glaciers Previous research has identified three common seasonal patterns (“types”) of ice flow at Greenland glaciers (Moon et al., 2014; Vijay et al., 2019). Some glaciers have a consistent type, while others change from year to year. Neighboring glaciers may have the same or different type. Previously, types were identified by examining flow at a single point. This limitation may affect the inferred variability. We use principal component (PC) / empirical orthogonal function (EOF) analysis to decompose maps of ice speed (Joughin et al., 2018; Joughin, 2021; Howat, 2020; Scambos et al., 206; Rosenau et al., 2015; Solgaard & Kusk, 2021) at four glaciers feeding Sermilik Fjord over 2014-2021. This improves on the previous single-point method by yielding temporal patterns (PCs), which allow types to be identified, plus their associated spatial patterns (EOFs). Helheim Glacier shows the most spatial and temporal heterogeneity of the four glaciers. PC #1 (95% of the variance in 2014-2021 speed) suggests primarily terminus control (p=0.003) but also some runoff control (p=0.05). PC #2 (1% of the variance in 2014-2021 speed) shows only runoff control (p=0.006). Previous work found that Helheim can be either terminus- or runoff-controlled. Fenris Glacier is runoff-adapting (PC #1, 66% of the variance in 2014-2021 speed). This disagrees with previous work that classified it as terminus-controlled. On Midgard Glacier, PC #1 (79% of the variance in 2014-2021 speed) is consistent with runoff-adapting behavior (p<0.0001). EOF #2 shows that the lowest 4 km is more runoff-controlled (p<0.00001) or terminus-controlled (p=0.04) than the upstream area. Our conclusion agrees with previous work that classified Midgard as runoff-adapting. Since separating from Midgard in 2009, Pourquoi Pas Glacier has slowed near the terminus while accelerating upstream. EOF #1 shows this pattern (67% of the variance in 2014-2021 speed); its PC shows runoff control. Previous work classified Pourquoi Pas as terminus-controlled. Overall, these results agree moderately with previous, simpler analyses. Thus, application of EOF/PC analysis to the popular “glacier type” problem holds some promise in the quest to discover what controls the seasonal flow patterns of Greenland glaciers.

Ice motion in land terminating regions of the Greenland Ice Sheet is controlled in part by meltwater input into moulins. Moulins, large near-vertical shafts that deliver supraglacial water to the bed, modulate local and regional basal water pressure and ice flow by influencing subglacial drainage efficiency on daily to seasonal timescales. Our previous modeling work found that the geometry of a moulin near the water line has substantial effect on subglacial water pressure variations. Here, we develop a new physically based moulin model which can help constrain moulin shape across the ice sheet and its influence on hydraulic head oscillation, and inform the englacial void parameter used in glacier hydrology modeling. The Moulin Shape (MouSh) model (in Matlab and Python) provides new insight into the evolution of subsurface moulin size and shape at hourly to multi-year timescales. The modeled moulin is initialized as a vertical cylinder. The moulin walls melt back above and below the water line due to the dissipation of turbulent energy, open or close due to viscous and elastic deformation, and freeze inward in winter when cold air temperatures and an absence of meltwater allow refreezing. We combine MouSh modeling results with geometric data from two moulins in Pâkitsoq, western Greenland, which we mapped to the water line. The moulins have heterogeneous shapes and volumes in the top 100 m. This suggests that the size and shape of the upper portion is controlled by local and regional pre-existing fractures, which provide preferential paths for water flow and melting, creating stochastic karst-like conduit shapes. Modeling results show that moulin geometry below the water line is influenced by the hydraulic head, which controls the depth-dependent elastic and viscous closure rates, and by the roughness of the walls, which enhances melt-out rates that oppose moulin closure. We show that subglacial water pressure across the ice sheet is likely influenced by moulin geometry, underscoring the need for including moulins in subglacial models.

The increasing ubiquity of high-resolution imagery has yielded many observations of water-filled crevasses across the surfaces of glaciers and ice sheets (e.g., Figure 1a). The subsurface character of these features, however, is not apparent in imagery, nor can it be fully elucidated even through field geophysics. Thus, what visible surface water in crevasses indicates about englacial hydrology, including whether there is a surface-to-bed connection, is currently subjective and interpreted differently by different scientists. Application of a physically based crevasse model to this problem shows that if a crevasse visibly holds water, it likely does not connect to the bed. The crevasse model incorporates depth-dependent visco-elastic deformation and refreezing to evolve the size and shape of a water-filled crevasse over hourly to decadal timescales (Figure 1b). Seasonally, visco-elastic closure tends to form a neck at the water line of most crevasses. Over a year or more, this neck can pinch off, isolating a pod of water that can extend hundreds of meters beneath the surface. The area above the neck persists as a 1–5 meter wide, 10–40 meter deep “ditch”: the surface expression of a dead crevasse that no longer receives surface melt. Accumulation of meltwater in these ditches is consistent with observations; the model results show that the ditches are not hydrologically connected to the crevasse or to the bed. These findings are consistent with recent observational work by Chudley et al. (2020), who concluded that visible water in crevasses sited in compressive stress settings was not connected to the bed. Observations of sudden drainage of these ditches show that reconnection to the englacial system, and potentially the bed, must be possible. The smooth bathymetry of the ditches, however, discourages formation of the starter crack needed to reactivate these hydrofractures. Thus, an external forcing, such as advection into a more-extensional stress setting, may be required to drain them. Overall, model results suggest that these water-filled ditches are shallow (<40 meters), overlie an englacial pod of liquid water that is in the process of refreezing, and are not connected to the bed.

Subglacial models represent moulins as cylinders or cones, but field observations suggest the upper part of moulins in the Greenland Ice Sheet have more complex shapes. These more complex shapes should cause englacial water storage within moulins to vary as a function of depth, a relationship not currently accounted for in models. Here, we use a coupled englacial--subglacial conduit model to explore how moulin shape affects depth-dependent moulin water storage and water pressure dynamics within a subglacial channel. We simulate seven different moulin shapes across a range of moulin sizes. We find that the englacial storage capacity at the water level is the main control over the daily water level oscillation range and that depth-varying changes in englacial water storage control the temporal shape of this oscillation. Further, the cross-sectional area of the moulin within the daily oscillation range, but not above or below this range, controls pressures within the connected subglacial channel. Specifically, large cross-sectional areas can dampen daily to weekly oscillations that occur in the surface meltwater supply. Our findings suggest that further knowledge of the shape of moulins around the equilibrium water level would improve englacial storage parameterization in subglacial hydrological models and aid predictions of hydro-dynamic coupling.