A comprehensive field investigation of the dynamic break-up processes on the Chaudière River, Quebec, Canada

Break-up in northern rivers is typically defined as the transition from the winter ice cover season to the open water condition. Although break-ups mostly occur in the spring, since it is associated with warmer weather and the end of season snowmelt, current climate change scenarios indicate that mid-winter break-ups will be more frequent in the future (Prowse et al. 2007; Beltaos & Prowse 2009; Burrell et al. 2022). River break-up can be thermal or dynamic. Thermal break-up occurs when warmer air temperatures and stronger solar radiations weaken and melt the ice in place with relatively little ice movement. This tends to occur when the runoff from rainfall or snowmelt events has a relatively weak effect on the river discharge. On the other hand, dynamic break-up occurs when river discharge increases (due to rain or snowmelt events) with a rate and a magnitude that is high enough to lift, break, and dislodge the ice cover, before significant thermal ice cover deterioration occurs (Beltaos 2008). As the broken pieces of ice start to move downstream, an ice jam would occur whenever the incoming ice discharge exceeds the local ice discharge capacity in a specific reach causing the sudden stoppage of an ice run and rapid backwater effects upstream of the jam (Dow Ambtman & Hicks 2012). Ice jams usually result in ice jam-induced flooding for riverine communities. As the incoming pieces of ice impound on the stationary ice jam, the upstream water levels increase creating additional pressure and drag, under which the jam will shove and consolidate (Nafziger et al. 2019). Once the driving force exerted on the jam surpasses the resistance strength of the accumulation, the ice jam would collapse (or release) causing an ice jam release wave or ‘jave’ to carry the broken pieces of ice downstream (Jasek 2003; Beltaos 2013). In a highly dynamic break-up, a cascade of ice jam formation and release events may occur as the break-up progresses downstream until the river is completely free of ice (Hicks 2016).

Understanding dynamic break-up processes and consequently predicting flooding potential has been critical for developing reliable forecasting models for flood early warning systems (White 2003; Mahabir et al. 2007). Forecasting models often combine meteorological, hydrological, and hydrometric information typically using statistically based (or data-driven) models (e.g., Madaeni et al. 2020; Becket et al. 2021) or physically based models (e.g., Rokaya et al. 2020) or a combination of both. Regardless of the type of break-up forecasting model, its reliability largely depends on the accuracy and completeness of historical field data and a good understanding of the dominant processes in the study reach.

Several publications have reported valuable information on relevant dynamic break-up processes (e.g., Gerard & Stanley 1988). The formation of freeze-up and mid-winter ice jams has been documented in the literature (Beltaos 2008), but the effect of these events on the spring break-up of an entire river system has not been assessed (Turcotte et al. 2020). Empirical and physics-based criteria have been developed for the triggering conditions for a dynamic break-up. Empirical methods assume a threshold value for (or a rate of change of) a parameter or a combination of parameters (typically river stage or discharge), beyond which ice cover break-up will initiate (Beltaos 1997). Physics-based criteria based on geometric constraints of the river reach or resistance of the ice cover have also been developed (e.g., Michel & Abdelnour 1975; Ferrick & Mulherin 1989; Beltaos 1997; Nzokou et al. 2009). Although these methods have been tested on break-up field data for some rivers (Beltaos 2008), and implemented in unsteady river ice numerical models (e.g., Jasek et al. 2005; Shen 2005; Ye & She 2021), they often require a comprehensive site-specific calibration that is limited by the inaccuracy and sparsity of field data during break-up.

Locations which promote jamming have been related to river geometry such as at river bends, reduction in slope, and at natural (e.g., islands) or man-made (e.g., bridges) constrictions (Beltaos 2008). Quantitative and qualitative studies have been conducted to identify the relative importance of each of the geometric parameters triggering the formation of a jam (e.g., Kalinin 2008; Morin et al. 2015; Osada et al. 2020). De Munck et al. (2017) developed a geospatial tool to calculate several geometric parameters (presence of islands, narrowing of the channel, sinuosity, presence of a bridge, confluence of rivers, and slope break) over regularly spaced segments of the river and used this tool to predict the locations of ice jam formation. Although such tools are of great interest for ice jam flood forecasting, still their accuracy needs to be assessed with extensive field observations.

In summary, to advance our understanding of dynamic break-up processes as well as improving break-up forecasting models, detailed and comprehensive field data are required. To address this need, a 2-year field monitoring program of the break-up processes on the Chaudière River has been conducted. Field data were collected using optical (trail cameras, drone flights, satellite images), in-stream (water level and hydrometric gauges), and spatiotemporal positioning (GPS trackers) instruments, along with frequent site visits to document ice conditions and reconstruct the details of each break-up event. Using these comprehensive field data sets, the sequence of break-up, jamming locations, as well as the effects of mid-winter events were documented. Finally, empirical criteria for the onset of break-up based on cumulative degree day of thawing (CDDT) and river discharge were assessed independently for each reach.

The Upper Chaudière is about 85.4 km long and extends from the Mégantic Lake (km 188.0) to the Sartigan Dam (km 102.6) and drains 3,075 km² (46% of the watershed) with an average slope of 2.5 m/km. This reach is entirely located in the Appalachian geological region with mild sinuosity and uniform cross sections. Due to the warmer discharges from Mégantic Lake, the ice cover progression is usually delayed at the upstream end of the reach (Biron et al. 2020). At the downstream end, an ice cover forms earlier in the season upstream of the Sartigan Dam, which is an ice control structure with no significant water storage capacity (i.e., the river flow is considered natural in the downstream reach).

The Intermediate Chaudière extends from the Sartigan Dam (km 102.6) to the Town of Scott (km 39.3), has an average milder slope of 0.5 m/km, and is partially entrenched by the valley walls. This reach drains an area of 2,658 km² (40% of the watershed) over only 63.3 km (33% of the length of the river) making it the reach with the largest ratio of drainage area over a river length of 42.0 km²/km (Table 1). The section between the Diable's rapids (km 86.3) and the Town of Scott (km 39.3) is called ‘Eaux Mortes’ (or ‘dead waters’ in English) due to its very mild slope of only 0.14 m/km. This low slope combined with the higher ratio of drainage area over river length makes this reach subject to frequent flooding whether for open water or in the presence of ice (Hamelin 1958; Ouellet et al. 1991; Biron et al. 2020). The ice cover tends to form first at this reach with no recorded historical observations on freeze-up consolidation or jamming.

The Lower Chaudière is characterized by the steepest slope of 3.0 m/km (Figure 1(b)). This reach extends from the Town of Scott (km 39.3) to the outlet of the Chaudière (km 0.0) and drains an area of 962 km² (14% of the entire catchment). The river in this reach is entrenched between rocky valley walls and exhibits successions of meanders, islands, and abrupt constrictions and widening, which results in a random behaviour when it comes to the sequence of freeze-up or break-up. Nevertheless, due to its steep slope, this reach is not prone to flooding problems.

According to the historical records (1999–2018) at Beauceville weather station (ID 7028754, km 83), the average annual air temperature has been 4.8 °C, while precipitation averages 1,108 mm annually. Frost (T_air ≤ 0 °C) occurs on average on 17 November, while thaw (T_air ≥ 0 °C) occurs around 28 March. Rainfall is well distributed throughout the year, although a slight increase is expected from late May to mid-August, when convective systems are strengthened. The historical records (1999–2018) at the St-Lambert station (km 21.7; ID 023402; drainage area of 5,820 km² or 87% of the watershed) indicate that the average discharge at this station is 120 m³/s with minimum discharge during summer reaching values as low as 5 m³/s, while peak values often recorded during spring melt can exceed 2,000 m³/s.

The Chaudière River has a long history of ice jam flooding and has been the focus of several studies. After a devastating ice jam flood in 1912 and open water flood in 1917, the Commission des Eaux Courantes du Québec (CECQ) conducted a study on the cause of these floods. They reported that break-up begins in the Upper Chaudière earlier than in the Intermediate and Lower Chaudière. Several ice jams formed between St-Georges and Beauceville over a reach of approximately 20 km (between km 100 and 80), which led to severe flooding in the region (CECQ 1921). In 1957, Beauceville experienced the worst disaster in its recorded history as a major ice jam formed at station 78.2 km and caused damage amounting to about $10 million (2020 dollars). Consequently, a series of studies were conducted in the 1960s and the ice control structure known as Sartigan Dam was built in 1967 upstream of the municipality of St-Georges (at km 102.6) to retain the ice from the Upper Chaudière. It was claimed that the Sartigan Dam would provide complete ice jam flood protection at St-Georges (km 99), and a reduction of 50–80% in ice jam flood probability in the downstream municipalities (Deslauriers et al. 1965). After a severe spring break-up ice jam flooding in 1991, the engineering firm TECSULT conducted a major study of the Chaudière River which included field observations of the 1993 break-up processes as well as hydraulic modelling of the study reach. The study recommended the construction of an ice control structure at the Diable's rapids (km 86.3) to protect the municipality of Beauceville (TECSULT 1994). This structure has not been built.

In the spring of 2019, a historic flooding event took place over 7 days which combined both ice jam and open water flooding. Following this catastrophic event, an expert committee suggested several solutions including relocating residents away from the floodplain, dredging the low slope intermediate reach, building an ice control structure, weakening of the ice cover prior to break-up and removal of obstacles from the channel (Biron et al. 2020).

A comprehensive 2-year field monitoring program was conducted during the winters of 2018–2019 and 2019–2020 to characterize and quantify the break-up processes on the Chaudière River. Air temperature data were collected from the Environment Canada weather station at Beauceville (ID 7028754) located near km 83, which is located in the centre of the study reach and considered representative of the weather conditions in the entire study reach. Precipitation (both liquid and solid) and snow on ground data were collected from the Ministère de l'Environnement, de la Lutte contre les changements climatiques, and de la Faune et des Parcs (MELCCFP) weather stations located at St-Ludger (ID 7027516) near km 149, at St-Georges (ID 7027283) near km 94, and at Scott (ID 7027840) near km 39. The average precipitation and snow on ground values reported by these three stations were used in this study. The locations of the weather stations are presented in Figure 1(a). Several hydrometric stations are installed on the Chaudière River at several locations by the Centre d'Expertise Hydrique de Quebec (CEHQ). In this study, the daily winter discharges from the Sartigan Dam (km 102.6; ID 023429) and the Famine River (a major tributary draining near km 97.5; ID 023422) stations were used to estimate the discharge associated with different break-up processes.

Over half a dozen field trips were conducted each winter to document ice conditions along the entire river. During these trips, photos, videos, and field notes were collected. An ice thickness measurements campaign was conducted each winter (from 28 February 2019 to 4 March 2019, and from 19 February 2020 to 5 March 2020) in the Intermediate Chaudière reach (approximately from Station 39 to station 100 km) prior to spring break-up.

Trail cameras have been widely used to study various ice processes in rivers (e.g., Vuyovich et al. 2009; Ansari et al. 2019). A total of 18 and 42 trail cameras (Brinno BCC100^®, Spypoint LINK-S^®, Enlaps Tikee^®), taking images at 15 min intervals, were installed along the river in the 2018–2019 and 2019–2020 winters, respectively. In addition, the CEHQ and the Comité de Bassin de la Rivière Chaudière (COBARIC; https://cobaric.qc.ca/) operate time-lapse cameras along their stations. These images were also available in the current study. Helicopter and drone observational flights have been widely used in the literature to document the spatial characteristics of the ice cover (e.g., Beltaos & Burrell 1991; Jasek 2003; Wazney et al. 2018; Ehrman et al. 2021). During the 2018–2019 winter, five helicopter flights (17 November, 8 December, 5, 16, 21 April) were conducted, and for the 2019–2020 winter, over 20 drone flights (using several drone models: DJI Mavic Air^®, DJI Mavic Mini^® and DJI Matrice 200^®) were conducted which made it possible to interpolate the information between the camera locations.

Satellite imagery is a very useful tool for documenting the spatial ice cover characteristics of a large watercourse (e.g., Beaton et al. 2019; Li et al. 2020; Altena & Kääb 2021). In this study, public optical satellite images of the Landsat-8 (every 16 days) and Sentinel-2 (every 5 days) constellations offered through the Sentinel-HUB^® portal (available at https://www.sentinel-hub.com/) were gathered. Cloudy conditions affected the number (and extent) of satellite images available for the analysis. Therefore, a total of 19 and 16 satellite images were used for the 2018–2019 and 2019–2020 winters, respectively.

Information on the ice movements (e.g., dislodging, jamming, and ice runs) was collected continuously using GPS trackers (Tracksolid^® and Tracki^®). Similar tools have been applied to the study of iceberg movements and ice sheet retreats (e.g., Rose et al. 2013; Yulmetov et al. 2016; Muckenhuber & Sagen 2018; Beltaos & Carter 2021). The GPS trackers were fixed on the ice cover before the break-up started and were used to create a spatiotemporal map of ice movements during that period. A total of 21 GPS trackers were deployed each winter.

Finally, interviews with municipal authorities (public works and fire departments) and information from local newspapers (L'Éclaireur-Progrès, Le Beauce Média, and BeauceTV) were used to confirm the timings and locations of specific events. More details on the specifics of each instrument (make, measurement interval, exact location, etc.) are available in Pelchat (2022). Table 2 provides a reach-by-reach summary of the instruments and gauging stations used to monitor river ice conditions during each season. Note that not all the data from all instruments/gauges listed in Table 2 were used in the current study.

The spatiotemporal sequence of the dynamic break-up was developed based on an integrated analysis of the data from all sources of information collected during this study. This analysis was conducted manually by observing every source of information and building time series of ice conditions at every observation station. The temporal frequency of documented ice conditions ranged from 15 to 60 min, based on the speed of change in ice conditions at the observation station. At each station and each time step, the ice was given one of the following descriptions: open water, partial ice cover, complete ice cover, mid-winter ice jam, and spring break-up ice jam. An ice jam was defined in this study as any accumulation of broken ice pieces due to the sudden stoppage of an ice run. This definition embraces all scales of jam sizes in length (axial), width (partial or bank to bank), or in amplitude (vertical), whether it caused flooding. The length of a jam was defined as the final length before it was released. This definition would consider the cascade of jam formation and release along the river since they are not separate independent entities in the time-space domain. Finally, the ice condition between fixed observation stations was interpolated using flights (helicopter or drones), GPS trackers, and satellite images.

Several onsets of break-up criteria, either empirical (threshold) based or physics based, have been suggested in the literature (Ye & She 2021). In the current study, two empirical criteria were investigated to quantify a threshold for ice mobilization (onset of break-up) at each reach: the CDDT was an indicator for the ice resistance (or weakening) and the discharge (Q) as an indicator of hydrodynamic forcing. The CDDT was calculated using air temperatures at the Beauceville station (ID 7028754), located in the centre of the watershed. Following Bilello (1980), the CDDT was calculated as the cumulative sum of absolute daily air temperatures from when the air temperature started to be above −5 °C. This base temperature of −5 °C was suggested by Bilello (1980) to account for the increased effects of solar radiation on the ice cover during spring break-up. To estimate the discharge associated with ice mobilization, a virtual station at the Town of Saint-George, called ‘0234sg’, was established by summing the discharge from the Sartigan Dam (ID 023429) and the Famine River (ID 023422) stations. The discharge at this virtual station corresponds to a drainage area of 3,781 km², or 56% of the watershed, and it is this discharge that was associated with the various ice mobilization events (and ice jam flooding) rather than elsewhere since most of the ice jams floods occur in the Intermediate Chaudière. The CEHQ publishes only daily discharges during winter at these stations. It was then important to estimate the instantaneous discharges since ice movement (jamming and release) can occur in a fraction of a day. To do this, the first instantaneous water levels (15 min) were converted to the corresponding open water discharge using published rating curves for each station. Then the instantaneous open water discharges were daily averaged and divided by the published daily winter discharge to estimate the daily backwater coefficient applied by the CEHQ. This coefficient was applied to the instantaneous open water discharge to estimate an instantaneous winter discharge for stations 023429 and 023422. Finally, the instantaneous discharges from both stations were simply added to each other to estimate the instantaneous discharge at the virtual station 0234sg.

The continuous monitoring of the river ice conditions made it possible to characterise mid-winter break-up events occurring during both winter seasons. These events caused partial break-up of the river, and as a result, ice pieces jammed at various locations and froze in place. These frozen jams later influenced the evacuation of ice in the spring. The following sections describe the mid-winter break-ups and their influences for each winter season.

During this season, the ice cover started forming on the river between 11 and 12 November 2019. Three warm spells accompanied by rainfall precipitations have been observed on 9 and 10 December 2019, 14 and 15 December 2019, and 11 January 2020, with CDDF reaching values between 180 and 370 °C-day. During each warm period, the discharge on the Chaudière increased to about 160–200 m³/s, and the ice cover was mobilized, mainly in the Intermediate Chaudière. Consequently, three ice jams have formed during each of the events at (a) 67.0–71.0 km, (b) from 71.0 to 76.0 km, and (c) from 89.0 to 92.3 km. It is worth noting that the two first warming events resulted in two back-to-back ice jams totalling 9 km long between 67.0 and 76.0 km. Similar to the first winter, an ice thickness campaign was conducted between 19 February 2020 and 5 March 2020, covering nearly 48 km between 38 km (Town of Scott) and the Diable's rapids at km 86.3. During this campaign, more than 1,200 ice thickness measurements were conducted at cross sections spaced every 400 m with 8–12 measurements per section. The results of this campaign are presented in Figure 3(b). The two mid-winter ice jams between 67 and 76 km resulted in ice thicknesses approaching 1.4 m near their toes. The average ice thickness along the river was 0.67 and 0.55 m upstream and downstream of the two jams, respectively.

Based on the integration of all sources of field data, it was possible to reconstruct the spatial and temporal chronology of the spring break-up on the Chaudière River during the two monitoring seasons.

The second period lasted for 14 days from 28 March to 10 April (Figures 4 and 5). During this period, the average air temperature was hovering around 0 °C and reached a maximum of 10 and 13 °C on 28 and 31 March, respectively. The CDDT increased from 25 to 90 °C-day, and the snow depth decreased from 45 to 20 cm before increasing again to 40 cm on 9 April due to a late winter snowstorm. The discharge increased slowly at the start of this period and peaked on 1 April to ∼175 m³/s right after the rain event on 30 and 31 March. Then it decreased to below 100 m³/s by 10 April. This increase in discharge resulted in a complete break-up in the Upper Chaudière. Consequently, on 2 April, two temporary ice jams were observed at stations 145 and 122 km, and a third ice jam at 108 km was observed forming upstream of the stable ice cover retained by the Sartigan Dam. This latter jam eroded in place over the following few days (Figure 5). In the Intermediate Chaudière, the river lost its ice cover downstream of the Sartigan Dam and downstream of the Diable's rapids (km 86.3). Also, several open leads near confluences with tributaries began to widen and eventually became completely ice free. In the Lower Chaudière, the steeper sections, already partially open, began to lengthen and completely lost their ice cover.

The third period lasted 7 days from 11–17 April. The average air temperature was around 4 °C and reached a peak of 16 °C on 13 April (Figure 4). On 14 April, a rainstorm of nearly 20 mm fell on the watershed. By the end of this period, the CDDT reached 150 °C-day, which led to the rapid melting of the 30 cm of snow cover. The discharge continued to increase until it reached a peak of ∼705 m³/s on 16 April. During this period, the ice cover in the Upper Chaudière upstream of the Sartigan Dam continued its thermal deterioration. In Intermediate Chaudière, the river lost its ice cover downstream of 55 km, but the mid-winter jam at 69.3 km has persisted and supported the upstream ice cover, thus blocking any further evacuation of the ice pieces from upstream (Figure 5). As a result, a significant ice jam accumulation near the Town of Beauceville (km 80) formed on the morning of 16 April. This ice jam resulted in overbank water levels and a devastating flood event in the Town. In the Lower Chaudière, all the ice was evacuated to the St. Lawrence River by 14 April with some temporary jams that did not cause any flooding for local residents.

The last period of the break-up lasted 3 days from 18 to 20 April. The average air temperature was around 4 °C and more than 60 mm of rain fell over 3 days. Consequently, the snow cover melted completely, and the discharge reached a peak of ∼1,845 m³/s on 20 April. This high discharge evacuated the last standing jam on 19 April, and the river was completely ice free in the early hours of 20 April.

The second period lasted 17 days from 21 March to 6 April. For the first 3 days (21–24 March), the average daily air temperature dropped to −8 °C and subsequently the CDDT dropped slightly (from 50 to 40 °C-day) and the depth of the snow on the ground decreased only by 2 cm (from 19 to 17 cm). Following this cold spell, the air temperature followed an increasing trend with an average daily temperature of 2 °C and a CDDT reaching 145 °C-day by 6 April. Three rain events occurred during this period. The first two events on 26 and 29 March were relatively mild with a rain depth of 3–4 mm/day. The third event on 1 and 2 April had a total depth of 17 mm over the 2 days. Subsequently, the snow on the ground melted completely by 3 April. The discharge increased dramatically on 21 March, reaching a peak of 660 m³/s as a response to the weather conditions. Following this peak, the flow stabilized around 150 m³/s before it peaked again at 780 m³/s on 3 April in response to the significant rain event on 2 April as well as the continuous snow melt. Regarding the ice conditions, the break-up was relatively progressive from 21 March until 3 April. During this period, two ice jams started to form in the Intermediate Chaudière, one near km 97 (Town of St-Georges) and another near km 87 (Town of Beauceville). Also, the ice cover started to weaken with several open water sections in the Lower Chaudière. Following the peak in discharge on 3 April, almost all the ice was evacuated from the intermediate and Lower Chaudière except for the jam that formed on 4 April around km 77. This jam has lengthened because of the release of the upstream mid-winter jam and remained in place until the early hours of 6 April (Figure 7).

Based on the 2 years of field monitoring of the spatial-temporal sequence of spring break-up, a generalization of the break-up sequence has been developed. Although the hydrometeorological conditions were different from one winter to the other, the behaviour of the observed break-ups retained a relatively similar pattern. For this analysis, a break-up sequence over the different sections of the Chaudière River was developed based on a time scale (steps) from 1 to 8 (1 being the first reach to lose its ice cover, and 8 being the last). These ‘steps’ are independent of the break-up periods defined for each season. A section was defined as a part of a reach of the river along which similar hydro-morphological conditions exist, such as discharge, depth, width, and slope. Based on this definition, 27 sections were identified: 3 in the Upper Chaudière, 9 in the Intermediate Chaudière, and 15 in the Lower Chaudière. The results of this analysis are presented in Table 3. The first section to lose its ice (Step 1) was the 1 U in the Upper Chaudière due to the continuous supply of warmer water from Mégantic Lake. At the same time, sections 7L, 10L, 12L, and 14L of the Lower Chaudière became open due to their steep slope and the presence of partial ice cover during winter. The second section to lose its ice is the remainder of the Upper Chaudière, down until km 107.8 (downstream of Stafford rapids), where the ice is retained by the Sartigan Dam until the very end of the break-up season (section 3 U in Step 8). Step 3 is then triggered by the first increase in spring discharge. At this point, the ice cover is mobilized at section 1I (immediately downstream of the Sartigan Dam) and 4I (at the Diable's rapids) in the Intermediate Chaudière, as well as in the other steeper reaches in the Lower Chaudière (2L, 4L, 6L, and 9L). The ice released from these sections usually forms short (100–500 m) accumulations against the downstream ice cover still in place. Once the increase in discharge became significant, the ice in section 2I (around the Town of St-Georges) in the Intermediate Chaudière started to move (Step 4). Ice jams of about 2 km in length usually formed in this reach between km 99.3 and 97.8. Under the effects of increased discharge, these jams then are released and added to the ice accumulations at section 3I.

Based on the detailed field measurements, it was possible to build an inventory of observed ice jams. A total of 51 ice jams (26 for 2018–2019 and 25 for 2019–2020) were documented during the two observation seasons (mid-winter and spring break-ups) ranging in length from 200 m to 5.7 km. These ice jams are presented in a chronological order in Tables 4 and 5 for the winters of 2018–2019 and 2019–2020, respectively. For each jam, the date of the jam formation and the final location of the head and the toe was identified (and therefore the length of the jam). Also, the geomorphological and/or anthropogenic river features most likely (to the best of the author's knowledge) to impede ice and initiate the formation of an ice jam as suggested in the literature (e.g., Beltaos 2008; De Munck et al. 2017) were identified. These include slope change (ΔSlope), river meanders, narrowing, bends, accumulation bars, islands, confluences with tributaries, and bridges.

Of a total of 51 ice jams observed in two observation seasons, 16 are associated with (at least in part caused by) the presence of one or more islands. If the deposition bars are considered in this category, the count rises to 19, which equates to more than 37% of ice jam sites that would be triggered by the presence of islands. Both the channel meandering and the slope changes (Δ slope) are responsible for 17 accumulation sites, or 33% of the jams. Confluences with tributaries, channel width reduction, and bridges were found to be the reasons for 22% of the remaining ice jams. It should be noted that for four ice jams (8%), it was not possible to identify geomorphological and/or anthropogenic settings that would have contributed to the formation of these ice jams.

Figure 10(a) presents the recorded CDDT at the onset of break-up per river section for the two monitored spring break-up seasons. As shown in the figure, the two spring break-ups have similar behaviour and trend in terms of the CDDT. The Upper Chaudière tended to lose most of its ice during the first spring warm spell when the CDDT reached between 20 and 80 °C-day. Consequently, the rest of the ice in the Upper Chaudière migrated to the Sartigan Dam reservoir and melted on site at a CDDT of around 135–155 °C-day. In the Intermediate Chaudière, an open water lead began to appear at the confluence of the Chaudière with the Nadeau and Lessard Creeks (km 56.6) as early as the CDDT reached 20 °C-day. The first ice cover movement began at around a CDDT of 50 °C-day at the Diable's rapids (km 86.3) and just downstream of the Sartigan Dam (km 102.6). However, the majority of the ice movements generally occurred between a CDDT of 110 and 150 °C-day. The section between the towns of Saint-Joseph (km 67) and Beauceville (km 83) is the last section to evacuate its ice cover at a CDDT between 140 and 180 °C-day. In the Lower Chaudière, several steep sections did not freeze completely during the winter, and therefore, CDDT at the onset of break-up is 0 °C-day for these sections. The remaining sections of the Lower Chaudière lost their ice cover at a fairly consistent pattern and around the same CDDT of ∼125 °C-day over the two break-up seasons.

Based on the aforementioned results and the similarity between the two observed seasons, it is evident that the CDDT (i.e., heat added to the water to melt/weaken the ice cover) was the dominant factor in the onset of break-up at several sections. In the Upper Chaudière, the most noticeable section is upstream of Sartigan Dam where the ice was retained by the structure and melted locally towards the end of the break-up season at a peak CDDT of ∼140 °C-day. In the Intermediate Chaudière, the CDDT was most probably the principal factor responsible for the onset of break-up at three locations: downstream of the Sartigan Dam (km 102.6), at the Diable's rapids (km 86.3), and at the confluence with the Nadeau and Lessard Creeks (km 56.5). These three sections lost their ice before any major increase in discharge and at a CDDT between 20 and 50 °C-day. At the first two locations (downstream of a control structure or rapids), the ice is already weak and would be ready to melt with minimum heat input. The open lead that formed at km 56.5 is clear evidence of the effect of warmer water coming from smaller streams on the break-up of the downstream rivers.

The discharge is the main hydrometric parameter representing the forces exerted by the flow on the ice. Under increased discharge, the ice would lift, break into smaller pieces, and detach from the side. Then the drag forces would move these broken ice sheets downstream with the flow. Figure 10(b) presents the estimated instantaneous discharge at the virtual station ‘0234sg’, above which the ice was mobilized at each river section for the two observation seasons. It is important to note that the discharge threshold for ice mobilization is referring only to the initial ice cover displacement, and not the discharge at which ice jams were released. There is a consistent pattern between the two seasons (Figure 10(b)). In the Upper Chaudière, the reach between km 172 and 108 lost its ice at a discharge between 60 and 200 m³/s, as opposed to a much larger range (from 125 to 750 m³/s) between km 108 and 102, immediately upstream of Sartigan Dam. The relatively narrow range of mobilization discharge in the upper reach indicates a strong dependency of the break-up of this reach on the discharge.

In the Intermediate Chaudière, most of the reach started to mobilize its cover between 350 and 750 m³/s. This relatively smaller range recorded over the 2 years is an indicator of the dominancy of the discharge threshold on the onset of break-up in the Intermediate Chaudière. The exception is during the spring of 2019 when the mid-winter ice jam between the towns of Saint-Joseph (km 67) and Beauceville (km 85) was the last reach to lose its cover at a discharge above 1,000 m³/s. Also, it is important to note that at the three locations in the Intermediate Chaudière at km 56.6, 86.3, and 102.6, the recorded mobilization discharge is relatively much smaller at 125 m³/s. This confirms that the ice at these sections will start moving once its internal strength is reduced (or a specific CDDT is reached) independent of any significant increase in discharge.

In the Lower Chaudière, there is a much higher variability between the two seasons. In spring 2019, most of the ice was mobilized at a discharge between 150 and 375 m³/s, while the upper limit of 365 m³/s seemed to be the dominant limit. For the 2020 spring break-up, the range is much wider from 150 to 710 m³/s, with most of the reaches mobilizing their ice at a discharge between 550 and 710 m³/s. It is important to note that the reaches where a threshold discharge of 0 m³/s was recorded indicate the reaches that never developed an ice cover during the winter season. The fact that there is a much higher variability of mobilization discharge between the two seasons, together with the relatively similar mobilization limit of CDDT, could suggest that the CDDT could be the dominant factor controlling the onset of break-up in this reach.

The values of the ice mobilization discharge are in line with the few quantitative studies on the break-up of the Chaudière River. During the spring break-up of 1993, it was estimated that the ice mobilization upstream of the Town of Beauceville (between km 83.5 and 102.6) occurred at a discharge of 620 m³/s, which is consistent with the observed upper limit in Figure 10 for this reach (TECSULT 1994). Also, during the same 1993 break-up, the equivalent discharge required to evacuate the ice jam between the Town of Beauceville (km 83.5) and the ‘Rocher’ (km 78.2) was 1,035 m³/s, which is consistent with the 1,000 m³/s threshold observed in this study (TECSULT 1994). This consistency demonstrates that this criterion can be potentially used to develop a break-up forecasting system for the Chaudière River. Turcotte et al. (2020) used an inventory of ice jam events recorded between 1947 and 2016 and extending from Saint-Lambert (km 21.7) to Saint-Ludger (km 151), to identify a discharge threshold for ice mobilization. In their study, the entire river was considered as one reach. Also, the Sartigan Dam ice control structure, which was built in 1967, was not taken into consideration in the analysis. Nevertheless, the reported discharge for initiation of ice mobilization ranged from 120 m³/s to 1,225 m³/s. Their results also showed that the majority of recorded spring ice jam flooding was associated with equivalent discharge ranging between 170 and 735 m³/s. This wide range in reported discharge demonstrates the feasibility of the application of the criteria per river reach as presented in the current study.

This study presents the results of an extensive field monitoring program of the spatial-temporal break-up of the Chaudière River over two winter seasons, 2018–2019 and 2019–2020. The results showed that the mid-winter break-up events had a significant effect on the spring break-up. During both monitoring seasons, several mid-winter ice jams have formed in the Intermediate Chaudière and froze in place forming ice accumulations over 1.4 m in thickness. These accumulations resulted in considerable ice jam flooding and were the last reaches to evacuate their ice during the spring break-up. This highlights the importance of monitoring the entire winter season (from freeze-up to spring break-up) when trying to develop an ice jam flood forecasting system. A generalization of the break-up sequence per reach was developed based on the detailed monitoring program. It was shown that in general, the steeper reaches in the Upper Chaudière and the Lower Chaudière were the first to lose their ice followed by the upstream and the downstream extremities of the Intermediate Chaudière. Finally, the middle reach of the Intermediate Chaudière (the reach with a milder slope and where the mid-winter ice jams had formed) was the last reach to evacuate its ice. Concurrently, the ice cover upstream of Sartigan Dam would melt in place showing the efficiency of this structure to retain its ice. An inventory of 51 ice jams was developed in this study including their location, length, time of formation, and the main trigger for the jam formation. It was found that 37% of ice jams formed due to the presence of islands, 33% of the jams due to channel meandering and slope changes, and 22% due to channel confluences with tributaries, channel width reduction, and bridges. There was no clear morphological feature responsible for the remaining 8% of the observed jams.

The CDDT and the discharge were tested independently to assess their applicability to predict the onset of break-up per river reach on the Chaudière River. It was found that ice movements began when the discharge and the CDDT were between 60 and 200 m³/s and 20 and 80 °C-day, for the Upper Chaudière, between 220 and 1,000 m³/s and 110 and 180 °C-day for the Intermediate Chaudière, and between 140 and 710 m³/s and about 125 °C-day for the Lower Chaudière, respectively. Based on a closer analysis of these thresholds, it was possible to identify the dependency of specific reaches on either of the two criteria versus the other. For example, the dependency of the steeper reaches and confluences with smaller streams on the CDDT as opposed to the dependency of the middle section of the Intermediate Chaudière on the discharge threshold.

The data collected and knowledge gained from this study will form the basis of future research. This includes the effects of hydrometeorological conditions on water temperatures and ice processes, the systematic evaluation of several onsets of break-up criteria per river reach, analysis of ice jam formation, consolidation and release, the effects of tributaries on the break-up processes, as well as developing an ice jam flood early warning system.

This research has been funded by the Ministère des Affaires municipals et de l'Habitation (MAMH) under the project ‘Actualisation de la cartographie des zones inondables de la rivière Chaudière (MU122274)’ as well as the Ministère de Sécurité publique (MSP) under the project ‘Compréhension du comportement des rivières en hiver et mesures de gestion des risques liés aux inondations (FLUTEIS; CPS-18-19-26)’ of the Québec gouvernement. This funding is greatly appreciated. The authors also would like to express their gratitude for Jean-Robert Ladouceur, Dany Crépault, Martin Lapointe, Éric Boucher, Christian Juneau, and Catherine Blouin from Laval University for their assistance in the instrumentation, field visits, and data analysis.

Tadros Ghobrial: conceptualization, supervision, writing paper draft, reviewing and editing, funding and resources management. Gabriel Pelchat: fieldwork, formal analysis, writing thesis. Brian Morse: conceptualization, supervision, and funding and resources management.

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

The authors declare there is no conflict.