Figure 1. North Klondike River under hydrostatic pressure in the Tombstone Park, Yukon, in April 2024.
People are right to assume that ice generally floats. H2O is among the few solids that are buoyant in their own liquid form. This can easily be verified by looking at an ice cube floating in a glass of water: 8% of the ice mass should emerge, whereas 92% of it should remain below the surface*. This simple physical process also applies to cold region lakes and slow-flowing rivers: a surface ice cover forms during winter, and if we drill a hole through this cover, water will fill that hole almost to the surface of the ice. Similarly, if we cut a piece of the ice cover (i.e., using a saw), it should float at the same elevation as the surrounding ice.
Interestingly, water can be seen on the ice surface of lakes and rivers during winter. This water-on-ice phenomenon is often associated with the presence of a heavy snow layer on the ice cover: the water underneath percolates through cracks that naturally form in the ice, and it spreads horizontally on the ice surface, soaking the bottom layer of the snowpack.
Other examples exist in nature where ice is not exactly floating. In steep streams, at the beginning of the cold season, water is often seen flowing on top of a porous ice mass called anchor ice. This ice is frozen to the bed material and cannot float freely. Later in winter, the ice cover on the same steep streams can be found in a free-spanning (bridge-like) state rather than in a floating state. This state occurs once some anchor ice has released (floated away) or melted, reducing the water level. More about steep channel ice processes can be found here and here.
However, can water flow under pressure beneath an ice cover? Can an ice cover withstand hydraulic pressure while preventing percolation through cracks? If this were possible, drilling a hole in the ice cover would release some of that pressure, creating a water fountain, just like in Figure 2 below.
Figure 2. Water flowing out from a 5 cm-diameter perforation in the aufeis of the North Klondike River, Yukon, in April 2024.
Our team is currently in the middle of a four-year research project that aims to document and quantify the formation of overflow ice. Often termed icing, aufeis, or glaciation, this process partially depends on the occurrence of pressurized flow events in streams. Why is this important? Icing commonly affects culverts under highways and impacts winter driving conditions; it can also cause significant damage to the transportation infrastructure, especially during the snowmelt period. Icing processes are poorly understood and are therefore often not adequately mitigated. Our attention has recently been driven to a complex ice process closely associated with overflow and icing: “ice blisters.”
What is an ice blister?
In streams, an ice or icing blister, is an obvious bump in the ice cover standing above the surrounding ice (including upstream) and presenting an oval or elongated (crest-like) shape. Our team observed blisters that were 1.5 m high, but the scientific literature suggests that they could be taller. Figure 3 shows an example of a couple of ice blisters in the Blackstone River, Yukon.
Figure 3. Colleague Stephanie Saal standing on a 1.2 m-high ice blister in the Blackstone River, Yukon, in March 2022 (another blister of similar size is visible on the left).
Ice blisters can contain water, or they can be dry (the water they use to contain has drained or frozen in place, usually leaving an air gap, a condition our team rarely observes). Their name, blister, is probably inadequate for most stream environments because they do not represent a closed environment or an isolated chamber for water to freeze (as is typical for similar features found inland in permafrost regions). The river ice blisters that we have documented were apparently directly connected to the main flow path under the ice.
Here is what we think we know about ice blisters:
The forces acting on the ice cover to form an ice blister likely originate from streamflow under pressure under the ice. Confirming the amount of pressure causing the blister over time would require a complex monitoring strategy: deploying a water level sensor near the location of a blister before an ice cover forms (hard to predict) and comparing this aquatic dataset with a corresponding continuous survey of the ice surface elevation. One objective of our team is to obtain those results through strategic monitoring.
Pressurized streamflow conditions are likely the result of two factors: 1. an ice cover that is largely frozen against (or fused to) the stream bed and banks, and 2. a streamflow that exceeds the carrying capacity of the ice-covered stream channel. The former condition means that extreme cold weather and relatively cold groundwater or permafrost, commonly found in subarctic regions, are needed for blisters to form. The latter does not occur because of a rain event or snowmelt. We believe that it either happens because of local ice cover thickening during a very cold period (constricting the flow path) or because of a rise in discharge at the end of a cold period during which a portion of the flow was being converted in ice in the upstream stream network (as air temperatures become mild, ice thickening stops in the stream and its tributaries, and the streamflow rises back to what the upstream groundwater can supply). This rise in flow could be as small as 5%, and this would be more than enough to cause a significant rise in the under-ice pressure.
Figures 4 and 5 present a drawing of a hypothetical cross-section of an ice blister on the East Blackstone River near the Dempster Highway, Yukon. In this case, the blister was 1.7 m high, and the thickness of the ice forming the blister was approximately 1.0 m. The size of the blister was about 25 m long (in the direction of flow) and only 5-10 m wide. When first observed, it was leaking at its foot, with an overflow rate of approximately 20 litres per second (l/s). Drilling the ice near the blister summit revealed a hydrostatic pressure (or pressure head) about 1 m above the surrounding ice cover. Drilling the ice near the blister’s foot created a fountain of approximately 0.4 m in height (outflow of about 5 l/s). Drilling more holes reduced overflow rates from other holes and cracks, meaning that the pressurized condition was driven by a relatively small excess flow rate.
Figure 4. Cross-section representation of an ice blister with a background photo from the East Blackstone River, Yukon, in January 2024. Colleague Médéric Girard is standing near the top of the blister.
Figure 5. Longitudinal section representation of the same blister as presented in Figure 4.
We think that ice blisters “take advantage” of existing cracks in the ice cover in areas where grounding (ice frozen to the channel bed) is either weak or distant (a lateral distance corresponding to several times the ice cover thickness). This is comparable to a chain breaking at its weakest link. However, that weak link must be strong enough to sustain some deformation.
We believe that ice blisters (and, more generally, ice layers made by the freezing of overflow) occur in streams that are generally shallow and wide. In a large and deep river (e.g., the Yukon River at Dawson), even a slight increase in pressure could soon generate cracks parallel to the banks, whereas in very narrow channels, the ice would not have the lateral freedom to bend, and the overflow would occur without the formation of any blister.
Here is what we need to confirm about ice blisters:
We are still determining how fast ice blisters form. However, since they rely on ice deformation (e.g. bending or creeping), we assume it takes several hours to a few days to see a measurable rise in the ice cover elevation as a blister develops. It is also unclear why most observed blisters are less than 2 m tall. Does that depend on the duration of the pressure event compared with the maximum deformation rate of the ice? Is it because large blisters usually leak, and the pressure is released that way? How about the key influence of channel gradient on the pressure head? Is there a link between the size of a blister and the hydraulic conductivity of the streambank material?
It is also uncertain how ice blisters, despite being made of a generally fragile (brittle) material that contains large cracks, can seal themselves while forming (most blisters we have observed were under pressure and not leaking at all). Generally, a small water leak exposed to a cold ice layer would immediately freeze (this also happens on lakes), and an ice layer that is initially cold and suddenly exposed to water at 0°C will slightly expand (thermal expansion). However, it is puzzling that the crack around the base of the blister is so consistently impermeable, possibly because of the significant pressure that builds where ice slabs come in contact (see areas or high pressure circled in red in Figure 4). The thermomechanical properties of freshwater ice also represent our main hypothesis to explain why ice blisters form relatively slowly (otherwise, the ice would deform too fast and would leak and break).
We are also not sure why ice blisters seem to form in a specific range of ice cover thicknesses (say 0.8 m to 1.5 m) and whether their formation is restricted to a certain weather or wintertime window. If a pressure event happens too soon during the winter period, the ice cover could be too thin and weak, and it may just break and release the excess flow. In contrast, if the ice cover is too thick and strong, which would happen later, it would not bend and crack under a relatively small ice pressure. Of course, if such a range existed, there would also be an ice mechanics perspective to it, so the channel width would also define the winter window for ice blisters to form along specific stream segments.
If an ice blister is leaking or has been drilled, we assume that the outflow will continue as long as there is a pressure head or until the crack or the hole entirely freezes. Since it is hard to freeze fast-flowing water (even when its temperature is 0.0°C), it is likely that, in some instances (depending on overflow rates and channel geometry), the overflow will spread in all possible directions until it submerges the water source (at a later stage to what is presented in Figure 5). The layer of overflow water would protect the leaking crack/hole against freezing for a fairly long time. Therefore, a significant, punctual outflow location may generate a significant ice cover thickening event, and water could eventually flow downstream over long distances, sandwiched between two ice layers (the early season ice cover below and the newly frozen layer above) for the entire winter period. Our team has observed that small blisters can be drowned by overflow from other sources (such as taller blisters) and eventually become encapsulated in ice. If we push the brainstorming further, we could speculate that most blisters eventually leak, but the evidence would be hidden by newly formed icing layers.
About under-ice flow velocities: Does water accelerate during a pressurized episode. If this was the case, at least at some locations, could such an episode trigger bedload sediment transport? If so, there could be a channel stability and morphology implication to the formation of ice blisters.
Is it possible that a blister could suddenly break and even toss ice pieces in the air? The Russian literature suggests so. Our team observed an ice blister that had obviously broken dynamically (i.e., mechanically) in Engineer Creek, Yukon (Figure 6). Several ice blocks and ice slabs had been mobilized by a few to several tens of meters downstream. However, we believe that the significant overflow from that blister caused the spatial dispersion of ice blocks, not a proper explosion (otherwise, we would have found ice pieces upstream and beyond the banks, into the bush).
Figure 6. Aufeis in Engineer Creek, Yukon, with ice slabs that have detached from the original blister on the right, as photographed on April 2024. Note the flow (and the ice) in Engineer Creek carries (contains) a significant amount of dissolved and suspended matter from nearby tributaries.
Figure 7 below shows a 1 m-thick ice slab from the same collapsed/broken blister in Engineer Creek. It shows two common types of ice covers (beside the author) and icing layers at distinct angles (circled in red). This picture alone suggests that 1. The blister started to develop after a few icing (frozen overflow) layers had formed at the surface of the early winter ice cover (unlike the blister in Figures 4-5), 2. Overflow layers had started to submerge the perimeter of the blister when it suddenly broke, 3. This blister was probably already leaking when the pressure increased to a point where its resistance threshold was exceeded, and 4. The blister was actually more resistant than ice slabs around it. Several questions about this unique encounter remained as our team had to leave the site near the end of the day.
Figure 7. Icing layers at different angles (circled in red) as observed at the same site as in Figure 6.
Ice blisters and, more generally, stream aufeis represent fascinating cold region natural processes. Readers should expect more science to be presented by our research team about pressurized stream landscapes in the next couple of years.
***
* About the buoyancy of ice, I invite readers to think about what is physically wrong with this classic (yet inaccurate) representation of an iceberg. Hint: The entire ice mass wants to float, not only the top part.
Iceberg representation as found on this website.
where? [AD1]