Frazil ice events: Assessing what to expect in the future

A question that is currently raised by stakeholders and decision makers concerned with the viability of shoreline infrastructures is: What is expected from frazil ice activity in rivers, taking into account the changing climate? The objectives of this article are, firstly, to present a brief overview of the circumstances under which frazil ice forms, namely nucleation mechanisms as well as the nature of a supercooling event. A basic understanding of these phenomena, which are highly dependent on air temperature, is deemed an informative initial step. Secondly, a brief description is presented of the main studies published so far that attempted to foresee frazil ice events and their contribution to ice jams. Lastly, it is to sum up the outcome of these studies.

Freeze-up of a water body may be either static or dynamic (Devik 1942, 1949; Mason 1958; Michel 1978; Beltaos 2013). A static state is when the surface is relatively undisturbed, in areas with low flow velocity, which is dominated by thermal effects. Growth proceeds from the solid matter along the shoreline and at the water surface from pre-existing ice nuclei. Heat transmission is from conduction – as such, the very low thermal conductivity of water causes a considerable thermal gradient in the water column immediately below. Air temperature, wind and short- and long-wave radiations are the main controlling factors in growth. A dynamic state is when the water column is turbulent due to currents and wind. In that case, the heat transmission is through convection and there is a very small thermal gradient with depth. In both static and dynamic freeze-up scenarios, the water is supercooled, also called ‘subcooling’ or ‘undercooling’ in the material science literature (Cantrell & Heymsfield 2005; Shamseddine et al. 2022). Although the temperature is below freezing point (0 °C), it remains in the liquid state. The amount of supercooling is the difference between that phase change temperature and the lowest temperature reached by the liquid, i.e., just before it starts to form crystals (Cantrell & Heymsfield 2005; Shamseddine et al. 2022). However, while supercooling for static scenarios is at or near the ice–water interface, dynamic scenarios can give rise to the supercooling of the entire water column. This happens during periods of intense heat transfer from the water surface to the cold air above it.

There are thus two basic requirements for a supercooling ‘event’ (Daly & Barrette 2022): a turbulent water column and a period of intense heat transfer from the water surface that cools the water temperature to below 0 °C.

A frazil ice event has a distinct start and a distinct end, during which time the water column is supercooled. As crystals form, the latent heat released by this nucleation process causes the temperature to go up again. The resulting temperature change is known as supercooling history, during which the ice crystallizes, flocculates and rises to the water surface.

Under the right circumstances, water can exist in the liquid phase well below the phase change temperature (Cantrell & Heymsfield 2005; Hartmann et al. 2011; Zahir et al. 2019). For the crystallization of pure water, that can happen in the upper troposphere, ‘where temperatures are consistently lower than −33 °C’ (Cantrell & Heymsfield 2005, p. 798). It leads to crystal nucleation events of a stochastic nature, which is referred to as homogeneous nucleation. When it is supercooled, however, the material is in a metastable state, such that a number of factors can trigger solidification at higher temperatures. The main one is the presence of foreign particles, which can act as catalysts or ‘seeds’, a process known as heterogeneous nucleation (Cantrell & Heymsfield 2005; Fujikawa et al. 2018). It explains why the extent of supercooling is normally much lower than that occurring in pure water. Particles of foreign matter at the surface, inside a water column or the material along river shorelines can all act as nucleation sites.

A second dichotomy refers to the nature of the interface from which nucleation arises. If it is at the interface with a foreign substance, it is referred to as primary nucleation (border ice, formed from shoreline material or anchor ice, formed from the riverbed). If it arises at the interface with already existing ice particles (snow, crystals forming above the water surface), it is referred to as secondary nucleation (Evans et al. 1974; Botsaris 1976). Various mechanisms allow the multiplication of nuclei. These include the detachment of tiny dendrites growing from existing crystals with or without liquid-shearing action, as well as the collision of crystals against an external surface or between crystals.

Foreseeing frazil ice events may be envisaged for two timescales. In the short term, i.e., within a few days to a few weeks, the relevance of these procedures would be at an operational level. That is, it benefits the management of facilities such as a hydroelectric dam, a run-of-the-river facility or municipal water plant, in preparing against these events. In the long term, i.e., within the next number of decades, it would provide an indication of the number of frazil ice events at a target location or along a given river reach. The emphasis in this article is in methods that aim at foreseeing frazil ice events at both timescales, with a focus on modeling endeavors. For the sake of conciseness, accounts addressing general river ice aspects and ice jams (e.g., Turcotte & Morse 2016; Beltaos 2021) are not considered in this article, even if they mention frazil ice. Other related matter, possibly of interest to the reader, includes means of quantifying the amount of frazil ice in a hanging dam (e.g., Vergeynst et al. 2017) and ways to mitigate frazil ice jams (e.g., Gholamreza-Kashi et al. 2011; Turcotte & Morse 2016).

The aim of short-term assessments is to increase confidence in being able to foresee frazil ice events within a timeframe of more than 1 day, i.e., for an operational perspective. These also help better understand longer-term approaches in the context of climate change. Among the many parameters that influence whether or not a frazil ice event will occur, air temperature and water discharge are those that matter most for both the short- and long-term perspectives.

Two examples of short-term analyses are given. Firstly, Beltaos et al. (2007) conducted some analyses on historical flooding events due to frazil ice over 140 years for Belleville River, in Ontario, Canada. Using an index for temperature – the lowest value of cumulative degree-days in January during 15 consecutive days – they showed that the majority of flooding events occurred below an approximate value of −7 °C (for that index). When plotted against a discharge index – the maximum discharge within the January 1–January 20 timeframe – approximate values −12 °C and 60 m³/s were recommended as a tentative delineation for the likelihood of flooding events, for that particular location.

A second example, similar to the first one, was proposed by Gholamreza-Kashi (2016) for Spencer Creek in southern Ontario, Canada. Historical information on flooding events and associated temperature/flow conditions were gathered. Then, air temperature data and flow data were collected on a daily basis from stations at the site of interest. A 5-day rolling average was obtained on the number of freezing degree-days and the flow data. Based on the above information, and using appropriate threshold values, the probability of frazil ice occurrence leading to a flooding event was assessed on a qualitative basis. In their case, an upper-bound temperature value of –14 °C and a lower-bound discharge value of 11 m³/s were used. For the purpose of the demonstration, that methodology was simplified, in the sense that it overlooks a number of other factors that play a role in frazil ice development. They provide an example of this, when a flooding event would have been expected but did not occur, possibly because of the presence of a thick ice cover. The methodology also relies on the availability of sufficient historical information, as well as current air temperature and flow data at the site of interest. It is meant to be updated on an ongoing basis with new flooding events.

To what extent can a changing climate affect frazil ice dynamics over the years? For a comprehensive overview of its impact on river ice dynamics and ice jams, the reader is referred to Burrell et al. (2022). In general terms, higher winter temperatures can promote the generation of frazil ice in rivers. Combined with an increase in the amount of rain, higher temperatures can induce a mid-winter succession of breakup and freeze-up, i.e., higher volumes of surface and underwater ice.

Some examples of these procedures are provided below – they are divided into two types: Thermal analysis and Temperature and precipitation.

Andrishak & Hicks (2005, 2008) evaluated the thermal regime of the Peace River, in Alberta, Canada. To do so, they developed a model based on River1D, a one-dimensional finite element code, which solved the conservation of water mass and longitudinal momentum. For the hindcast exercise, the model was adapted so as to take into account additional considerations, such as water temperature, frazil ice dynamics and surface ice. The model did not address the more complex mechanical processes, such as ice jam formation or ice cover consolidation. The parameters required to conduct this exercise included water temperature, ice conditions and inflow information, such as water level and discharge. The calibration of model temperature against observations yielded a value of 15 W/m² for the heat exchange coefficient. The procedures incorporated the calibration of an empirical parameter meant to capture the effect of reduced ice velocity of the frazil ice pans as they approach the leading edge of the ice cover, and the effect of ‘associated crushing, under-turning or consolidation of floes’. This was done against historical records of, namely, the ice front progression. Frazil floe porosity, frazil rise parameter, Manning's value and ice–water heat exchange coefficients were also obtained by these procedures.

Their forecast exercise relied on the Canadian CGCM2 climate model. Using parameter values produced in the hindcast exercise, a comparison was made between the migration of the ice front upstream historically and that into the future (2040–2060). It was determined that the latter was reduced compared to the former. A similar comparison of the ice cover duration was made, with the future showing a reduction ranging from 41 to 63%. These studies are based on the assumption that the ice cover evolution at these locations is the outcome of frazil ice production. The downstream to upstream ice cover progression is attributed to the juxtaposition of frazil ice pans against the leading edge of the ice cover. The model does not factor in the amount of border ice, which can be considerable in natural rivers. This leads to an uncertainty in evaluating the relative proportion of frazil ice that contributes to the ice cover progression.

Aaltonen et al. (2010) followed up on Huokuna et al. (2009) with a study on the same river but with a refined approach where temporal and local temperature variations were accounted for. Several climate scenarios were used – see Aaltonen et al. (2010) for more information. Daily discharges for the 30-year periods 1971–2000 (the reference period), 2010–39 and 2040–69 were simulated with the same hydrological model, with flow regulation also factored in the modeling exercise. The number of frazil ice risk days increases in all climate scenarios, with December and January being the most frazil ice prone. Compared to the reference period, an increase in monthly energy heat flux from water to air is observed for the frazil ice risk days, for the months of January, February and March.

A freeze-up ice jam, defined as ‘an unconsolidated cover that eventually becomes, fully or in part, a sheet of solid ice’ (Beltaos 2013, p. 181), is an important part of river ice formation. It occurs at the beginning of the winter, thereby representing a risk to shoreline infrastructures from flooding at that time of the year (in contrast with breakup jams, which typically occur at the end of the winter). Along frazil-prone river reaches, extreme ice thicknesses can be achieved. Flood modeling and forecasting capabilities have been developed and implemented for operational use – these tools were used to foresee climate impact.

For instance, in the case of the Churchill River in Labrador (Lindenschmidt et al. 2021), freeze-up and freeze-up jamming are expected to be delayed in future fall seasons due to increased air temperatures (KGS Group 2020). However, increased flows due to increased precipitation could lead to freeze-up ice jamming events becoming more severe. A freeze-up ice jam flood forecasting system was also developed for the Exploits River in Newfoundland (Hatch 2021). It was determined that, with a warming climate, ‘warmer fall temperatures will lead to a delay in the freeze-up date, and warmer spring temperatures will likely lead to earlier breakup dates and shorter winter seasons. These changes will affect the nature and character of the ice cover that forms on the river, and the processes that drive its formation’ (Hatch 2021, p. 6–63). The warmer temperatures will lead to less production of frazil ice; hence, shorter freeze-up ice covers are expected, reducing the hazard of freeze-up ice jam flooding.

A workshop was held in October 2022 in Poznan, Poland, on ice jam flood hazard and risk assessment, with representation from Norway, Sweden, Finland, Germany and Poland (Lindenschmidt et al. 2023). Most ice jam flooding in these European countries occurs during freeze-up. An important question posed at the workshop was if ice jam flooding should be specifically addressed in the European Union (EU) Floods Directive. It was concluded that ice jam floods are already receiving attention in the directive since all types of floods need to be cataloged in the flood management plans, including pluvial and open-water fluvial floods. Special mention regarding ice jam floods is not required in the directive since most ice jam flooding events in these regions of the Baltic Sea have exceedance water level probabilities that are below those of open-water fluvial floods, probably because most ice jams occur during freeze-up. This does not mean that freeze-up ice jam floods are less severe than open-water floods.

A cold spell in February 2021 caused severe ice jamming and subsequent flooding on the Oder and Vistula rivers (Lindenschmidt et al. 2023). Such floods are of great concern and steps have been taken to develop and improve flood forecasting capabilities along these rivers (Lindenschmidt et al. 2019). The Torne River along the Swedish/Finnish border is also plagued by ice jam flooding (Lindenschmidt et al. 2018) and flood hazard maps have been developed by both countries to inform the public and water managers of the potential of flooding from freeze-up ice jam (examples of maps are provided in Lindenschmidt et al. (2023)). It is difficult to say how the frequency and severity of freeze-up ice jams will change in the future due to the changing climate. An interesting observation, though, is that ‘projections for the future indicate a greater increase in land areas where river floods become more frequent, compared to the fraction of areas for which fluvial floods will decrease’ (Lindenschmidt et al. 2023, p. 5).

Mid-winter breakups (MWBs) are breakups that typically occur during January and February, i.e., they precede those in the spring (Beltaos et al. 2003; Huntington et al. 2003; Newton et al. 2016; Burrell et al. 2022; Das et al. 2022; De Coste et al. 2022). The analyses by Das et al. (2023) are a recent example specifically meant to foresee MWBs in the short term, including frazil ice generation. Factors favoring MWBs include higher-than-normal air temperatures, increased precipitations, increased water discharges and thinner ice. MWBs can promote the generation of frazil ice in the expanses of open water (i.e., the outcome of the breakup) with the return of the cold air temperatures prevailing at that time of the year. The frequency of MWBs has been increasing over time as an outcome of climate change. Consequently, risks of trash rack clogging are no longer limited to the fall but extend well into the winter season. From a climate change perspective, this will add to the challenge of foreseeing these events. The reason is that climate models cannot adequately assess variability within any given winter period. Instead, the usual practice is to conduct ensemble-type analyses based on these models but over a full winter. These scenarios are sometimes labeled according to the conditions, i.e., as ‘warm’, ‘wet’ or ‘cold’ scenarios (using Huokuna et al. (2009) as an example).

Frazil ice and anchor ice, collectively referred to as underwater ice, constitute a challenge for various engineered structures whose operations rely on water intakes; that ice is also known to promote flooding, particularly during the freeze-up period. Will a changing climate promote circumstances favoring the supercooling of the water column? From the brief review presented herein, we conclude there is no generic yes-or-no answer to this question. The climate seems to have an impact in most cases, but not in others, i.e., each case should be treated separately. The present study highlights several ways in which that has been done. One challenge is that, since frazil ice forms below the water surface, it is difficult to monitor. Predictive approaches mostly rely on what is observed near the surface or above it (including hydrological, hydraulic and climate data). Approaches are also empirical, i.e., parameters such as threshold values are only valid for specific locations. The methodologies summarized herein, be it to assess frazil ice risks or anticipate water temperature and frazil ice volume, could be envisaged for problematic areas (on a case-by-case basis). An expected increase in the frequency of MWBs due to warm spells implies that some shoreline communities will have to prepare for frazil ice events and their consequences not only in the fall but also well into the winter months. From a climate impact perspective, this means that, in addition to the already existing challenge of anticipating year-to-year variability, climate models would have to capture month-to-month variability with sufficient accuracy. Overall, it is the future of fixed ice coverage in rivers, frazil ice notwithstanding, that needs to be better understood.

Funding for this work was provided by Infrastructure Canada through NRC's Climate Resilient Built Environment initiative. The authors gratefully acknowledge comments on the manuscript from two anonymous reviewers.

All relevant data are included in the paper or its Supplementary information.

The authors declare there is no conflict.