In northern rivers, turbulent water becomes supercooled (i.e. cooled to slightly below the freezing point) when exposed to freezing air temperatures. In supercooled turbulent water, frazil (small ice disks) crystals are generated in the water column, and anchor ice starts to form on the bed. Two anchor ice formation mechanisms have been reported in the literature: either by the accumulation of suspended frazil particles, which are adhesive (sticky) in nature, on the riverbed or by in situ growth of ice crystals on the bed material. Once anchor ice has formed on the bed, the accumulation typically continues to grow (due to either further frazil accumulation and/or crystal growth) until release occurs due to mechanical (shear force by the flow or buoyancy of the accumulation) or thermal (warming of the water column which weakens the ice-substrate bond) forcing or a combination of the two. There have been a number of detailed laboratory studies of anchor ice reported in the literature, but very few field measurements of anchor ice processes have been reported. These measurements have relied on either sampling anchor ice accumulations from the riverbed or qualitatively describing the observed formation and release. In this study, a custom-built imaging system (camera and lighting) was developed to capture high-resolution digital images of anchor ice formation and release on the riverbed. A total of six anchor ice events were successfully captured in the time-lapse images, and for the first time, the different initiation, growth, and release mechanisms were measured in the field. Four stages of the anchor ice cycle were identified: Stage 1: initiation by in situ crystal growth; Stage 2: transitional phase; Stage 3: linear growth; and Stage 4: release phase. Anchor ice initiation due to in situ growth was observed in three events, and in the remainder, the accumulation appeared to be initiated by frazil deposition. The Stage 1 growth rates ranged from 1.3 to 2.0 cm/h, and the Stage 2 and 3 growth rates varied from 0.3 to 0.9 cm/h. Anchor ice was observed releasing from the bed in three modes: lifting of the entire accumulation, shearing of layers of the accumulation, and rapid release of the entire accumulation.
Anchor ice is described as ice that is attached or “anchored” to the bed of natural water bodies (rivers, lakes, or sea floors) as defined by the World Meteorological Organization (1970). Although observations on the formation of anchor ice in rivers have been documented since the 18th century (Barnes, 1908; Altberg, 1936), the mechanisms of formation, growth, and release, as well as its overall effect on river ice processes, are still a relatively unstudied phenomenon (Tsang, 1982; Beltaos, 2013). Anchor ice formation and release can cause significant changes to the riverbed geometry, affecting water levels and discharges, and consequently leading to the loss of hydropower production during freeze-up seasons (e.g. Girling and Groeneveld, 1999; Jasek et al., 2015). Anchor ice released from the bottom often contains significant amounts of bed materials which contribute to the sediment transported in river systems (e.g. Kempema and Ettema, 2011; Kalke et al., 2017). Recently it has been shown that the duration and extent of anchor ice cycles have an effect on algae growth rates and its total biomass (Suzuki et al., 2018). Moreover, fish habitat, in particular fish spawning, can be affected by the formation of anchor ice on the bed which can block oxygen supply to substrate water and freeze the eggs under the ice (e.g. Prowse, 2001; Brown et al., 2011). Available river ice numerical models have attempted to include the effects of anchor ice in hydraulic modelling. These models have mostly relied on empirical or semi-empirical relations (e.g. Shen, 2010; Lindenschmidt, 2017; Blackburn and She, 2019; Makkonen and Tikanmäki, 2018), but the development of physically based models has been challenging due to the lack of accurate field measurements of anchor ice formation, growth, and release.
The process of anchor ice formation starts when surface water becomes supercooled (i.e. water is cooled below its freezing point) typically due to freezing air temperatures. In the presence of sufficient flow turbulence, the supercooled surface water is transported to lower layers and quickly reaches the riverbed (Daly, 1994). In supercooled turbulent water, active (sticky) frazil ice crystals are typically generated in the water column, and subsequently anchor ice may also form on the riverbed. It has been established that anchor ice formation can be initiated by two processes: in situ growth of ice crystals (i.e. nucleation of ice crystals atop the bed material) and/or accretion (i.e. deposition) of active frazil particles on submerged objects (Tsang, 1982). After the initial formation of anchor ice crystals on the riverbed, the accumulation continues to grow either by crystal growth due to heat loss to the surrounding supercooled water and/or by the further deposition of suspended frazil particles (Osterkamp and Gosink, 1983; Qu and Doering, 2007). The final thickness of the anchor ice layer is limited by several factors including the absence of supercooled water, the stream flow depth (although in some cases the accumulation can emerge above the water surface forming anchor ice dams), the growth of an overlaying surface layer of stationary or border ice, and the release of the anchor ice accumulation from the bottom (e.g. Osterkamp and Gosink, 1983; Beltaos, 2013; Turcotte et al., 2013). Increasing anchor ice thickness coincides with increases in the drag and buoyancy forces acting on the accumulation. Anchor ice release is thought to occur due to mechanical or thermal forcing or a combination of the two (e.g. Parkinson, 1984; Shen, 2005). The mechanical release of anchor ice occurs when the buoyancy and drag forces are greater than the ice-substrate bond or when these forces are greater than the submerged weight of the anchor ice and the attached bed materials (sands, gravel, or boulders). This latter mechanism results in the rafting of riverbed materials (sediments) as released anchor ice pans are advected downstream (Kempema and Ettema, 2011; Kalke et al., 2017). Anchor ice has often been observed to release in the morning following cold and clear nights (Barnes, 1908; Kempema et al., 2001; Daly and Ettema, 2006), and this has been attributed to warming of the water by solar radiation that weakens the ice-substrate bond leading to thermal release.
Many of the physical measurements available on anchor ice formation, growth,
and release are from detailed laboratory experiments (Altberg, 1936; Kerr
et al., 2002; Doering et al., 2001; Qu and Doering, 2007). In laboratories,
the ambient conditions are controlled (air temperature, discharge, and
channel characteristics), and the environment is favourable for conducting
detailed measurements (e.g. video recordings, depth and velocity profiles)
and for collecting samples of anchor ice accumulations. Altberg (1936)
conducted the first reported controlled laboratory experiments to study
anchor ice using a race-track recirculating flume. He concluded that two
fundamental conditions must exist for anchor ice formation: supercooling of
the water (thermodynamic effect) and flow turbulence (dynamic effect). Kerr
et al. (2002) conducted a laboratory study on anchor ice formation and its
hydraulic effects on a gravel bed in a refrigerated flume. In all of their
experimental runs, they observed anchor ice initiation only due to frazil
deposition or attachment to the bed (i.e. no in situ growth). They documented
three distinct stages of anchor ice growth: initial, transitional, and final
growth stages. The initial stage (referred to as “Stage 1” herein) is the
formation of the first visible anchor ice crystal layers on the substrate.
This stage was characterized by faster growth rates and uneven appearance
along the bed. The transitional stage (referred to as “Stage 2” herein)
started once the accumulations began to emerge out of the substrate and
protruded into the flow. The hydrodynamic drag forces acting on the
accumulation caused the flattening or release of the anchor ice formation.
During this transitional stage, anchor ice accumulation either continued to
increase due to frazil deposition or decreased slightly due to its release
into the flow. The final growth stage (referred to as “Stage 3” herein)
was defined as the nearly uniform (linear) and slower growth rates of anchor
ice thickness due to continuous frazil deposition. At this stage, individual
anchor ice forms were not distinguishable. The measured growth rates ranged
between
Doering et al. (2001) and Qu and Doering (2007) conducted laboratory experiments on anchor ice evolution in a counter-rotating flume. They measured growth rates between 0.3 and 0.7 cm/h, and their data suggested that anchor ice growth rates and densities increased with increasing Froude number. Although they visually observed that anchor ice always initiated with frazil deposition, with a careful interpretation of the water temperature (supercooling) curves, they were able to attribute some of the continuous growth of anchor ice thickness to the in situ growth of the crystals. They found that gravel size did not have an effect on the formation (initiation) mechanisms, but they found that anchor ice released more easily when it was attached to smaller gravel particles. They also observed that anchor ice tended to release when the Reynolds number is less than 42 000.
Several field observations on anchor ice processes have been reported in the literature using grab sampling techniques, onshore photographs, and intermittent underwater photography. Therefore, these observations were mostly limited to the qualitative description of anchor ice properties such as shape, thickness, extent of formation, and release and how these properties are related to hydro-meteorological conditions (e.g. Hirayama et al., 1997; Terada et al., 1998; Turcotte and Morse, 2011; Nafziger et al., 2017). Reasons for this limitation include the need for supercooled water, the fact that anchor ice typically forms during night time when visibility is low, the difficulty of predicting where anchor ice might form, and of course the limitation of working in cold weather. Despite these limitations, field studies have significantly advanced our knowledge of anchor ice processes.
Hirayama et al. (1997) conducted a three years study on a small gravel
stream (6 to 9 m wide and 0.3 to 0.6 m deep) in Hokkaido, Japan. They mapped
the Froude number contours within the study reach and showed that anchor ice
only forms when the Froude number is between 0.2 and 1.5. They reported that
sampled anchor ice masses consisted mainly of needle-like crystals which
were growing either in the upstream or the downstream direction of the flow.
They showed that the thickness and volume of anchor ice accumulations
increased with the cumulative degree hour of freezing of air temperature
and reported anchor ice thicknesses ranging between 3 and 17 cm. Moreover, they
measured anchor ice accumulation densities between 300 and 700 kg/m
Kempema and Ettema (2009, 2011) studied anchor ice on the Laramie River, USA, and reported that daily anchor ice cycles (formation at night and release in the morning) can generate anchor ice accumulation thicknesses between 0.2 and 0.3 m with no apparent relation between crystal sizes and Froude number. They examined the morphology of anchor ice accumulations and reported that large plate-like ice crystals exceeding 5 cm in length were most dominant. They also observed that these large crystals result from in situ growth of frazil crystals that become attached to the bed. While studying anchor ice rafting by sediments, they concluded that anchor ice release and associated ice rafting were diurnal events, which suggests that solar radiation was an important factor. Kempema and Ettema (2013, 2016) used time-lapse images from an underwater camera to measure the growth of anchor ice crystals forming on wedge-wire screens deployed on the riverbed. They reported that anchor ice formation was initiated by the accretion of frazil disks followed by relatively rapid in situ ice growth. They observed ice crystals up to 3 cm long with growth rates ranging between 1.0 and 4.0 cm/h. They also measured anchor ice accumulation porosities between 42 % and 68 %.
Stickler and Alfredsen (2009) conducted a detailed study on anchor ice
formation at three sites: two sites on unregulated rivers (Southwest Brook,
Canada, and the Sokna river, Norway) and one site on a regulated stream (Orkla River, Norway). They concluded that anchor ice formation is mainly due to
frazil deposition and is dependent on the flow turbulence (i.e. Reynolds
number) with no apparent correlation with the Froude number. They reported that
in low-turbulence areas (median Reynolds number of
Dubé et al. (2014) investigated the characteristics of anchor ice accumulations on the Montmorency River, Canada, using thin section analysis and computed axial tomography (CAT) scans of the collected anchor ice samples. Elongated columnar ice crystals were observed only in ice dam samples, and disk-shaped ice crystals were observed in both ice dam samples and submerged anchor ice samples. In both cases, the crystals grew preferentially perpendicular to the flow. Their results also suggested that in situ growth of disk-shaped ice crystals was the dominant process for the formation of anchor ice and ice dams. They reported individual ice crystals between 3 and 6 cm in length with mean accumulation porosity between 38 % and 44 %.
Jasek et al. (2015) and Jasek (2016) monitored anchor ice release in the
Peace River, Canada. They showed that anchor ice formation and release
caused significant fluctuations in discharge and water levels which caused
ice cover instability and consolidation and consequently freeze-up jams.
Their observations showed that anchor ice release appeared to be mainly due
to hydraulic effects rather than the thermal influence of the sun. Acoustic
scanning of the river bottom indicated length to thickness ratios of
Nafziger et al. (2017) studied three streams in New Brunswick, Canada (two
unregulated and one regulated), and 161 anchor ice formation/release events
were observed using time-lapse photographs from the shore. A correlation was
found between the increase in water depth (stage) during the formation of an
anchor ice event and the corresponding accumulated freezing degree hours of
air temperature. Although there were no direct measurements of growth of
anchor ice accumulation, the “trend” of the increase in water levels
showed good agreement with the laboratory growth rates reported by Kerr et al. (2002). On days with a net heat gain at the water surface and air
temperatures
As discussed above, previous field and laboratory studies have provided considerable insight into the formation, release, and properties of anchor ice, but there are still considerable gaps in our knowledge, for example, the relative importance of frazil deposition versus in situ growth, mechanical versus thermal release and single versus multi-day cycles. However, one of the most critical gaps is that the anchor ice growth rates and mechanisms observed in the laboratory have never been confirmed in the field. The primary goal of this study was to address this gap by making direct measurements of anchor ice growth in the field. For this purpose, a custom-built underwater imaging system (camera and lighting) was deployed on the North Saskatchewan River in Edmonton. The imaging system was able to capture for the first time high-resolution digital time-lapse images of anchor ice formation, growth, and release mechanisms. This paper describes the deployments of the imaging system, results from the continuous measurements of anchor ice processes, and the effect of ambient hydro-meteorological conditions on these processes.
The North Saskatchewan River, Canada (length
Map showing the study site on the North Saskatchewan River in Edmonton at the Quesnell Bridge (the base map was downloaded from the Atlas of Canada's Toporama website). The inset is a photo from the right bank looking north showing the deployment site under the bridge.
In order to capture high-resolution photographs of anchor ice properties in
the field, an underwater imaging system and artificial substrate were
designed and built as shown in Fig. 2. The imaging system consisted of a
36 MP Nikon D800 digital single-lens reflex (DSLR) camera (equipped
with a Micro-Nikkor 35 mm f/1.8D lens) coupled with a Nikon SB-910
Speedlight flash. Both the camera and the flash were contained in underwater
housings which were mounted side-by-side on a MiniTec aluminum rail. The
imaging system was secured in a 100 cm long, 50 cm wide, and 20 cm high PVC
frame. The aluminum rail was designed to be released from the frame using two 20 cm high handles and a pivot hinge assembly. This feature allowed the rail to
be lifted out of the frame so that images could be examined and batteries
changed without removing the entire system from the river. The frame was
equipped with two 10 pound (
Anchor ice imaging system and artificial substrate after its retrieval from the river on 12 November 2018.
An artificial substrate was constructed and bolted to the imaging system frame as shown in Fig. 2. Although imaging anchor ice formation directly on the natural riverbed would be ideal, it is very difficult to pre-adjust the camera settings and lightings to acquire clear images of the forming anchor ice. However, using a constructed substrate allowed us to conduct preliminary laboratory experiments to adjust these settings in a controlled environment. In addition, a constructed substrate offered the opportunity to observe multiple anchor ice events growing on an identical substrate, eliminating variation in bottom sediment properties as an experimental variable. Lastly, it also allowed us to closely examine and photograph any anchor ice deposits that had not released once the system was removed from the water. To make the constructed substrate as similar as possible to the natural riverbed, bed material was sampled from the river at the deployment site in early October 2018. The bed samples were oven dried and sieved in the University of Alberta geotechnical lab. Sediment particles that were greater than 3.8 cm (1.5 inch) in size were used for the substrate. The substrate materials were hand-picked so that they had one relatively flat side and ranged in size from 3.8 to 12.5 cm, which is classified as very coarse pebble gravel to fine cobble gravel (Valentine, 2019). The gravel particles were then glued (with the flat side facing down to increase contact area) to a 50 cm by 50 cm wide plywood base. Multiple 2.5 cm diameter holes were drilled into the base to reduce buoyancy forces on the substrate.
Initial imaging settings (including ISO, aperture, focus, and duration of the flash pulse) and the distance between the substrate and the lens face were determined in the laboratory by immersing the system in a tank of tap water. These camera settings and the setup configuration were modified over the course of the field deployments to improve image quality and increase the camera battery life. These modifications included decreasing the distance from the camera housing window to the in-focus bed material from 60 cm to 40 cm, adding a 25 mm extension tube to the camera lens, changing the underwater camera housing from a clear Ikelite D800 housing to a coated aluminum Aquatica AD800 housing, and increasing the image sampling interval from 30 s to 5 min. Adding the extension tube and moving the substrate closer resulted in a field of view in the images of 34 cm by 18 cm as opposed to 45 cm by 30 cm for the original configuration. Increasing the image sampling interval extended the camera battery life from about 6 to 24 h.
In addition to photographing anchor ice, the water temperature was measured
to investigate the effect of temperature variations on anchor ice formation
and release. Measurements were made using two RBRsolo T (accuracy
Meteorological data were downloaded from the Alberta Climate Information Service (ACIS) website. The closest weather station (approximately 2.0 km southeast of the study site; see Fig. 1) was the “Edmonton South Campus UA” station (Climate ID 3012220) which provides hourly weather data for the air temperature, solar radiation, wind speed and direction, rainfall/snowfall depth, and relative humidity. Real-time hydrometric data for the North Saskatchewan River at Edmonton were obtained from Water Survey of Canada gauge #05DF001 at a 5 min interval. The gauge is located approximately 10.7 km downstream of the study site (Fig. 1).
A careful examination of the images showed that the imaging system was able to capture individual anchor ice crystals growing on the artificial substrate and also the thickness of anchor ice accumulation. Therefore, the images were processed to primarily estimate these two quantities. All of the captured anchor ice images were processed in two steps. First, the images were enhanced using image processing software (BatchPhoto Pro®). The enhancements included stamping the date and time when the image was taken, auto adjusting the contrast, reducing the hue, increasing the saturation, increasing the lightness, and reducing the noise in the images. These enhancements corrected for the continuous change in ambient lighting and flow turbidity over the course of a single deployment.
Second, the enhanced images were imported into MATLAB®, and the edge of anchor ice crystals and the anchor ice accumulation thickness were manually tracked as a function of time using the image processing toolbox. For the crystal growth measurements, the images were visually examined to identify a number of crystals (typically between one and four crystals) that were clearly visible in consecutive images. Then the pixel coordinates of the edge of each identified crystal were manually tracked and extracted from the series of images. The pixel distance between the edge of the same crystal on successive images was scaled by using the in-focus size of the substrate material. The total length of the crystal was then calculated to estimate the growth of the accumulation with time. A processed image showing the individual crystals forming on the substrate is presented in Fig. 3a. The anchor ice accumulation thickness was measured by manually tracking multiple points across the top of the accumulation in each image. The average accumulation thickness was calculated for each image by averaging these manually tracked points across the width of the image. A processed image showing anchor ice accumulation atop the substrate is shown in Fig. 3b.
Sample processed images from an anchor ice event on 3 and 4 December
2018 (Event C) showing
There are three possible sources of uncertainty in the anchor ice
measurements that have been identified: (1) the camera resolution, (2)
errors in identifying the same crystal in consecutive series of images, and
(3) errors associated with the assumption that the scaling factor used to
convert pixel dimensions to real dimensions (i.e. centimeters) was a
constant. The camera resolution is 36 MP which resulted in a pixel
size of approximately 40
Figure 4 presents time series of the air temperature and the river stage measured from 1 November 2018 to 31 December 2018. The first ice pan was observed in the river on 7 November 2018, and the river was completely ice covered on 23 December 2018. The freeze-up season lasted almost 46 d and was one of the longest in recent years. During freeze-up, the weather forecast was monitored, and the dates of the deployments were determined based on when supercooling of the river was expected to occur and the availability of the research team. The anchor ice imaging system was deployed a total of four times (referred to as DEP-1 to DEP-4) during the 2018 freeze-up season as highlighted in Fig. 4. Table 1 summarizes the camera settings and the duration and timing of each deployment.
Summary of field deployments and camera settings.
Time series of
Figures 5 to 8 present time series of the measured air temperature, water
temperature, solar radiation, river stage, and anchor ice thickness during
the field deployments DEP-1 to DEP-4, respectively. Prior to deployments
DEP-1 and DEP-2, the air temperature was above zero and dropped to
Time series of results during deployment DEP-1 showing
DEP-2 started at 18:00 on 15 November 2018 and ended at 12:00 on 17 November 2018,
lasting for 40 h (see Fig. 6). Although DEP-2 lasted 2 nights, anchor
ice did not form during the first night because the water was above
0
Time series of results of DEP-2 showing
After DEP-2, the air temperature stayed relatively warm until 1 December 2018
when the temperature dropped below zero. DEP-3 started at 16:00 on
3 December 2018 when the air temperature decreased from
Time series of results of DEP-3 showing
After DEP-3, the air temperature gradually warmed again, but rafts and ice
pans were still observed in the river. On 15 December 2018, the temperature
dropped from above 0 to
Time series of results of DEP-4 showing
At the end of each deployment, after the retrieval of the instruments from the river, images of anchor ice that had not released from the substrate were taken (e.g. Fig. 9). From these images, three distinct anchor ice crystal shapes were observed on the substrate as shown in Fig. 10. These shapes are (a) curved needle crystals that grew on the surface of the bed material from the contact edges between adjacent gravels towards the centre of the gravel from all sides, (b) platelet crystals that grew starting in the interstitial spaces between gravel particles and then grew vertically away from the gravel, typically angled upstream, and (c) disk-shaped crystals that look like typical suspended frazil ice crystals that attached to the substrate. It is clear in Fig. 10 that within each anchor ice accumulation, different crystal shapes are present, which demonstrates the complexity that is commonly observed in anchor ice accumulations.
Anchor ice formation on the artificial substrate after the instrument's retrieval on 12 November 2018.
Observed anchor ice crystal types (highlighted with red circles)
from the 2018 freeze-up season showing
The processed images from each anchor ice event were combined in time-lapse
videos to help visualize the results. An example of such videos for Event
C is available for download at
Summary of results for measured anchor ice events.
Time series plot of the measured anchor ice thickness during Event C on 3–4 December 2018 labelled with the different stages of anchor ice formation, growth, and release. Note that Stage 4 shows the “lifting” release mechanism before the total removal of the anchor ice accumulation.
Time series of individual crystal growth (Stage 1)
measured from events B, C, and D. For illustrative purposes, linear growth
rates of
Stage 2 is a transitional period when individual crystals came in contact
with each other and were not easily distinguished in the photographs. During
this stage, the surface of the anchor ice started to become flattened by the
flow due to the increased drag force and then continued to grow through the
deposition of suspended frazil crystals and flocs and/or further
interstitial crystal growth. This stage was only distinguishable during
events B, C, and D when in situ crystal growth was observed. This stage
lasted for
Time series of anchor ice thickness growth due to frazil
deposition (stages 2 and 3) for all the measured events. For illustrative
purposes, linear growth rates of
Stage 4, the release of anchor ice, was recorded for events C, D, E, and F
but not for events A and B due to equipment malfunction or due to retrieval
prior to the release. Three modes of anchor ice release were observed in the
data: lifting, shearing, and rapid release. During the release of events C
and E, the entire anchor ice accumulation was observed lifting up and away from
the substrate until it suddenly completely released. This mode lasted
Three stages of anchor ice growth very similar to those reported by Kerr et al. (2002) were observed for the first time in the field in this study using time-lapse photographs. Three of the six anchor ice events (events B, C, and D) were observed to be initiated by in situ crystal growth (Stage 1) followed by frazil deposition. For the remaining three events (events A, E, and F), no in situ crystal growth was observed, and it appeared that the accumulations grew only by frazil deposition (stages 2 and 3). It should be noted that Kerr et al. (2002) did not report observing in situ crystal growth and attributed the faster growth in Stage 1 only to frazil deposition. Qu and Doering (2007) did not directly observe in situ thermal growth in their anchor ice images but did conclude that it occurred in their laboratory experiments based on careful analysis of water temperature time series. Kempema and Ettema (2009) studied anchor ice crystal morphology on a small riffle and pool stream and collected anchor ice samples that were comprised of large blade-shaped crystals up to 5 cm in length (see their Fig. 1). They concluded that these larger crystals were formed by the in situ growth of suspended frazil crystals that had become attached to the bed. Furthermore, they wrote that anchor ice formed initially by the adhesion of frazil ice crystals to the bed and that subsequent growth occurred through a combination of frazil accretion and in situ growth. It is unclear if they are referring to the adhesion of a relatively small number of suspended frazil ice crystals to the bed that subsequently acted as nucleation sites for the growth of large crystals or if they are referring to the adhesion of a sufficient number of crystals that a layer of measurable thickness could initially form and that in situ growth could occur within this layer. Their description of anchor ice formation is certainly consistent with events A, E, and F in which only frazil deposition was observed. However, for events B, C, and D in which the in situ growth of large crystals was initially observed, there is some uncertainty. The initial adhesion of suspended frazil ice crystals to the bed prior to the start of large crystal growth was not observed in the time-lapse images, but this could be because the suspended crystals were much too small to be visible. Therefore, it is possible that the first step in anchor ice formation is the adhesion of suspended frazil crystals to the bed, and in this case, it would follow that some of these crystals could then act as nucleation sites for the large crystals that were observed growing on the bed in this study. This process would also be consistent with the description of Kempema and Ettema (2009) of their field observations.
The initial or Stage 1 average crystal growth rates measured in this study ranged from 1.3 to 2.0 cm/h. McFarlane et al. (2016) reported growth rates between 0.4 and 2.3 cm/h for dendritic frazil ice crystals observed in a sequence of images of a frazil floc trapped between two cross-polarizing filters at the same field site. Initial growth rates of anchor ice accumulations in a laboratory channel were estimated from the slopes of the curves plotted in Fig. 19 from Kerr et al. (1997), and these varied from approximately 1.7 to 2.8 cm/h. Kempema and Ettema (2013, 2016) observed anchor ice growing on wedge-wire screens in the Laramie River and plotted crystal growth as a function of time (e.g. see Fig. 5 in Kempema and Ettema, 2016). The initial growth rates estimated from these plots ranged from approximately 1.0 to 4.0 cm/h. In summary, the reported field and laboratory measurements of anchor ice crystal growth range between approximately 1.0 and 4.0 cm/h. The differences in observed crystal growth rates were likely due to local variations in turbulent flow properties, heat loss rates, and/or the characteristics of the substrate on which crystals grew. It is important to note that even on the same substrate and during the same time interval, individual anchor ice crystals grew at different rates as shown in Fig. 12 for Event C. In this case, the different growth rates might be because crystals growing at different positions on the substrate and in different orientations relative to the flow would be exposed to different flow conditions.
The time-lapse images of anchor ice during stages 2 and 3 indicate that the
growth of the accumulations was mainly due to frazil deposition. It is
possible that further in situ crystal growth in the interstitial spaces
between the deposited frazil crystals occurred and that this would increase the
accumulation density and strengthen the bond between crystals. The rate of
growth of anchor ice accumulation during stages 2 and 3 ranged between 0.3
and 0.9 cm/h. This rate is in agreement with the laboratory measured rates
of 0.4 to 0.8 cm/h reported by Kerr et al. (2002) and 0.3 to 0.7 cm/h
reported by Doering et al. (2001). It is interesting to note that in this
study, the average rate of crystal growth (Stage 1) of 1.7 cm/h was
Anchor ice thickness growth due to frazil deposition
(stages 2 and 3) against the cumulative degree hour of freezing (CDHF). The
linear rates of 0.05 and 0.12 cm/
Total anchor ice thicknesses measured in this study (at the end of Stage 3)
ranged from 6.1 to 15.4 cm. This range is consistent with the ranges
reported in some previous studies, e.g. 3 to 17 cm by Hirayama et al. (1997), 20 to 30 cm by Kempema et al. (2001), and 7 to 10 cm by Stickler and
Alfredsen (2009). The crystal sizes observed in this study ranged from
2.8 to 7.7 cm and are in agreement with previous field studies, e.g. 3
to 6 cm by Dubé et al. (2014) and up to 10 cm by Kempema and Ettema
(2011). However, several studies do report substantially thicker
accumulations. Tremblay et al. (2014) reported thicknesses ranging from 0.18
to 0.46 m in a small river (width 6–12 m), and Evans et al. (2017) reported
accumulations up to
Four of the six anchor ice events observed in this study started within 0.5 h of sunset, and the remaining two events started 1.6 and 4.3 h after
sunset. This is consistent with what would be expected for diurnal anchor
ice events that begin in the late afternoon or evening. At this time of day,
the combined effect of decreasing shortwave solar radiation and lower air
temperatures typically leads to an increase in the net heat flux from the
water to the atmosphere that initiates anchor ice formation. Events C and D
occurred in shallow water (
The possibility that mechanical forces triggered the release was also
considered. Buoyancy and hydrodynamic forces always play some role in anchor
ice release since they are always present. This can be illustrated by
considering two limiting cases. In the first case, the ice-substrate bond is
weakened by thermal effects, and one or both forces lift or shear the
accumulation off the bed. This would be characterized as thermal release
since it was thermal radiation and/or heating that triggered the release. In
the second case, the strength of the ice-substrate bond remains constant, and
the magnitude of one or both forces increases, triggering release. This would
be characterized as mechanical release. In some cases, both the strength of
the ice-substrate bond and the magnitude of the forces may be varying, and
then release could be triggered by both a weakening of the bond and an
increase in one or both of the forces. In this study, the four accumulations
(events C, D, E, and F) grew to thicknesses that ranged from 6.1 to 15.4 cm and
then released. Nafziger et al. (2017) estimated the strength of the anchor
ice-substrate bond using an equation proposed by Malenchak (2011) and by
assuming that the anchor ice accumulation released solely due to buoyancy
forces. In order to make similar calculations, we assumed the anchor ice
density varied from 300 to 700 kg/m
It is difficult to quantitatively assess the role of hydrodynamic forces in the release of the four events since the only information available is the approximate local depth and the data from the Water Survey of Canada (WSC) gauge (#05DF001). The gauge data for events C and D do not appear to be ice affected, with water levels varying from approximately 3.15 to 3.45 m due to diurnal hydropeaking (see Fig. 7b). During events E and F, the water level was steadily rising from 3.16 to 3.71 m and did not follow the typical diurnal pattern, indicating that the presence of ice was affecting the gauge (see Fig. 8b). The release of Event C coincided with a peak in the daily water levels of 3.38 m. The release of Event D occurred during rising water levels 5–6 h prior to the daily peak, and the stage was 3.45 m. During events E and F, the water levels were also rising (see Fig. 8b) likely due to a combination of hydropeaking and backwater effects related to ice congestion downstream, and release occurred at a stage of 3.36 and 3.70 m, respectively. Therefore, all four anchor ice events that were observed releasing did so when water levels were rising or were approaching the daily maximum. This may indicate that hydrodynamic forces played a role in the release of these anchor ice accumulations, but it is difficult to conclude this with any certainty.
The rate of anchor ice growth is currently calculated in most river ice
process models using the following equation:
During Stage 1 when in situ growth of anchor ice was observed, the average
crystal growth rates ranged from 1.3 to 2.0 cm/h. Using these values and
porosities of 0.4 and 0.8, the resulting range in
The first continuous field measurements of the complete anchor ice cycle
including initiation, growth, and release mechanisms were captured in this
study. Three stages of growth similar to those reported by Kerr et al. (2002) were observed in the time-lapse images. A total of six anchor ice
events were captured: growth due to frazil deposition and in situ growth
was observed in three events, and in the remainder only frazil deposition
occurred. Anchor ice was observed releasing from the bed in three modes
referred to as “lifting” of the entire accumulation, “shearing” of
layers of the accumulation, and “rapid” release of the entire accumulation.
The Stage 1 growth rates measured by tracking the growth of individual
crystals on the substrate ranged from 1.3 to 2.0 cm/h, and these were
comparable to rates observed in previous laboratory and field studies. The
measured growth rates in stages 2 and 3 due to frazil deposition varied from
0.3 to 0.9 cm/h, which are comparable to measurements made in two previous
laboratory studies. It is worth noting that in this study, significantly
higher growth rates ranging from 0.05 to 0.12 cm/
All of the observed anchor ice accumulations began forming in the afternoon
or evening between 16:30 and 21:00. The release of four of the accumulations
was captured in the time-lapse images and occurred between 07:15 and 14:40.
The two events in shallow water released just prior to sunrise and the two
events in deeper water in the early afternoon. There is evidence that solar
radiation, buoyancy, and hydrodynamic forces may have all played some role in
the timing of the releases. It does not seem likely that release was
triggered directly by hydrodynamic forces because the water level and flow
rate variations were not significant at the time of release. The fact that
during all four events the water temperature remained supercooled at
residual temperatures of approximately
River ice process models currently use a semi-empirical equation to model
anchor ice growth due to frazil deposition and in situ growth. This simple
equation accounts for the two observed growth mechanisms and is based on
sound physics combined with reasonable engineering approximations (Shen et
al., 1995). Analysis of the term in the equation modelling frazil deposition
leads to the conclusion that the frazil accretion rate
Data are available from the authors upon request.
TRG and MRL designed the apparatus and performed the field work together. TRG carried out the analysis and processing of the data, prepared the figures, and wrote the paper with review and contributions from MRL.
The authors declare that they have no conflict of interest.
We would like to thank the Natural Sciences and Engineering Research Council of Canada (NSERC) for supporting this project and Perry Fedun for his valuable technical assistance. We are grateful for that support.
This research has been supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) (grant nos. RGPIN-2015-04670 and RGPAS 477890-2015).
This paper was edited by Claude Duguay and reviewed by Edward Kempema and three anonymous referees.