1 Introduction

Ice-jam floods are a threat for residents living along rivers in cold regions. They represent “the greatest hazards related to river ice” (Ashton 1986), and they are often more destructive than open-water floods (Beltaos and Burrell 2002). In 2006 only, their damage assessment reached up to US $250 million in North America (Prowse et al. 2008). The impacts of ice-jam processes cause great damage not only on human infrastructures and navigation, but also on vegetation, aquatic habitats, and particularly on fluvial geomorphology (Beltaos 1995). The Mistassini River (Quebec, Canada) is an archetype of this reality. In May 2011, mechanical ice-jams of high intensity severely damaged 20 riverside houses, as well as vegetation and channel banks (see Fig. 1).

Fig. 1
figure 1

Mechanical ice-jam and its impacts on the Mistassini River (Quebec, Canada), May 2011 a ice-jam release and ice rafts, May 2011 b ice-induced scour on a sandy bank seen post-event, summer 2012 c ice rafts that crashed on the banks, May 2011 d one of the twenty houses destroyed, May 2011 e ice impacts on trees, May 2011 (photographs by MRC Maria-Chapdelaine, May 2011 and S. Morin, June 2012)

Full size image

The scientific community has long-recognized river ice as a geomorphological agent in northern rivers. Among the first authors to report the effects of ice on river morphologies, McPherson (1966) observed ice-scoured banks and sediments pushed up onto the floodplain, immediately after the 1965 spring breakup period of the Red Deer River, in Alberta. Smith (1979) also found an unusual widening of the bankfull width caused by ice scour along 24 rivers in Alberta and hypothesized that ice processes might control hydraulic geometry. In the last decades, several other studies reported ice-impact features on river morphology (see Table 1) and concluded that ice is an important morphological component of northern rivers. However, some authors believe that ice processes, such as punctual ice-jams, do not have an extensive impact on alluvial streams, especially when compared to what is moved over a year to maintain the river channel (Kellerhals and Church 1980). Consequently, the incidence of river ice as a significant geomorphological agent needs to be clarified.

Table 1 Principal ice-impact features reported in the literature
Full size table

The extent and persistence of landforms in fluvial systems most often are related to the characteristics of the geomorphic disturbance regime, namely the frequency (number of events per unit time) and the magnitude (damaging potential) of extreme flood events (Dauphiné 2003). For example, in the case of ice-jams, Boucher et al. (2009) determined that ice-induced landforms on the Necopastic River (Eastern Canada) could be found strictly at sites where ice-jams occurred at least once every 5 years. Similarly, Henoch (1973) and MacKay and MacKay (1977) concluded that ice erosion features on banks of the Mackenzie River (Northwest Territories, Canada) formed during high-intensity events where water levels and ice rafts rose about 15 m above the mean level. However, few other studies cross-analyzed these data and little information exists concerning the frequency and magnitude associated with ice-induced landforms. Moreover, these features were observed in different hydro-climatic and geological settings, precluding an in-depth comparison of linkages between ice-jam regimes and geomorphological signatures. Consequently, such relationships need to be validated in other contexts to provide additional information on frequency/magnitude thresholds required to generate and maintain ice-induced landforms in cold region fluvial landscapes.

To evaluate flooding risks and environmental impacts related to ice jamming, gathering knowledge on the frequency and magnitude of events is of utmost importance. In that context, a correlation between the occurrence of ice-impact features and ice-jam frequency/magnitude characteristics would imply that landforms can point out to particular ice-jam regime properties. In turn, a simple geomorphological characterization of the channel and banks would provide useful information concerning ice regimes, especially where no ice-jam records exist. The localization of those features could ultimately help refine flood hazard maps from a hydro-geomorphological perspective (Demers et al. 2014).

Quantifying river ice regimes, however, remains an important challenge. In most rivers, records of past ice-jam activity is scarce, punctual, and incomplete. Recent development in tree-ring analysis has nevertheless allowed to fill this knowledge gap. Based on the analysis of damages made by ice to the trunks (Taylor et al. 2008; Smith 2003; Boucher et al. 2009), several characteristics of the ice-jam regime can be detected and linked to dominant geomorphological features (Boucher et al. 2009). This paper precisely aims to characterize the spatial variability of ice-jam features and associates it to downstream variations in ice-jam regime characteristics. To achieve this goal, field-based geomorphological surveys will be coupled to extensive ice-jam frequency/magnitude data obtained from tree-ring analysis. In order to do so, riverbanks along four morphodynamic sections of the Mistassini River were scrutinized to establish a typology of ice-impact features. Hydraulic and hydro-geomorphological characteristics were also described to investigate the spatial context of formation such features on the Mistassini River. Dendrochronological analysis of ice scars found on riparian trees was finally used to reconstruct ice-jam location, frequencies, and magnitude on the Mistassini River.

2 Study area

The Mistassini River is a large semi-alluvial river flowing north to south from the 50th parallel, at the east of the Mistassini Lake, to the 48th where it ends in Lake Saint-Jean, Quebec (Canada). Regional climate is mainly continental, but temperature varies from north to south, ranging from mild to cold subpolar temperatures (OBV Lac St-Jean 2012). The Mistassini River basin occupies 29 % of Lake Saint-Jean watershed with a surface area of 21 814 km2, a length of 298 km, and a discharge up to 800 m3/sec in spring (OBV Lac St-Jean 2012). The principal tributaries are the Samaqua, Ousiemsca, Rats, and Mistassibi rivers. The river flows in a wide range of geomorphological settings, from the granitic highlands of the Laurentian Plateau to the sedimentary lowlands composed of shale and limestone (see Fig. 2). The post-glacial past of the region has left in place various glacial and post-glacial deposits whose composition depends of the topography and the deglaciation phase. For example, in the highlands, the retreat of glacier front has left in place glacial, glaciolacustrine, and glaciofluvial deposits. This retreat has generated considerable mass of water in the south and a isostasic lowering, which correspond to the Laflamme Sea’s marine invasion phase. Glaciomarine, glaciolacustrine, and deltaic deposits are therefore found in the in the lowlands (Daigneault et al. 2011).

Fig. 2
figure 2

Location of the four morphodynamic sections along the Mistassini River

Full size image

The Mistassini region has a proportion of residential occupation of 6.94 buildings/km2, with a higher concentration in the downstream part of the basin. Industry and agriculture also occupy an important proportion of the southernmost part of the drainage area. Forest harvesting is omnipresent in the north and in the south.

Hydro-geomorphological properties allow for the identification of river sections that are morphodynamically homogeneous (Demers and Buffin-Bélanger 2011). Four sections were selected according to their entrenchment ratio, bank deposits, and river sinuosity (see Fig. 2). This method helped optimize the field campaign by allocating efforts in the four sections that were likely to present various morphological responses to ice-jam events.

Section 1 is part of the granitic highlands where the stream channel meanders deeply through glaciofluvial and glaciolacustrine deposits, but also, reworked alluvia (Daigneault et al. 2011). Well-developed floodplains and intense erosive areas alternate as a result of active meandering. Few human infrastructures are found in this section.

Section 2 is part of the fluvial transition between the highlands and lowlands. Its important slope combined with the fine grain alluvia creates a slightly entrenched linear stream. Glaciolacustrine deposits and reworked alluvia cover the bed and banks (Daigneault et al. 2011). Several sandy and rocky islands are found in the channel, along with bars and dune beds.

Section 3 differs from the previous one, mostly by the quasi-absence of islands and because of the coarse-grained deposits covering the bed and banks. In fact, the sudden appearance of glaciofluvial ice-contact deposits (Daigneault et al. 2011) causes an important change in grain size, which brings along gravels and pebbles, linked by a loose sandy matrix. Running water is found in a few narrow segments.

Section 4 differs considerably from upstream sections, since the river is deeply entrenched in fine-grained glaciomarine deposits (clays and deltaic sands) of the lowlands (Daigneault et al. 2011). Moreover, 20-m-high rocky cascades characterize the beginning and the end of this section. In between, the gentle slope leads to a shallow and sinuous channel with several sandy islands and well-developed floodplains. The 2011 high-magnitude event that destroyed 20 riverside houses occurred in this section where most infrastructures are found. Its banks are often modified by residents; some are even free of vegetation.

3 Methods

3.1 Geomorphological description of riverbanks

Natural riverbanks on both sides of the river were firstly identified using only the principal geomorphological processes (erosion, accumulation, ice impacts, etc.). Each bank was described and georeferenced. Seven qualitative and quantitative variables (see Table 2) were used to describe 178 bank segments all across the four sections using their geomorphological and ecological characteristics.

Table 2 Description of the methods used to sample the geomorphological variables of each bank
Full size table

A cluster analysis was then performed to determine whether some banks presented similarities in their geomorphological and ecological characteristics that can regroup them into clusters. The first step of this analysis was to incorporate the geomorphological and ecological data into a dissimilarity matrix, which can be defined as a matrix that expresses the degree of dissimilarity between sets (in this case, the bank types; Kaufman and Rousseeuw 1990). Dissimilarity corresponded to the Euclidian distance and was routinely calculated by the DAISY function (according to Kaufman and Rousseeuw 1990) of the cluster package (version 1.15.2) in R open-source statistic software. Then, the pairwise dissimilarities (Euclidean distance) between the bank types of the data sets were computed, which resulted in the identification of groups with similar Euclidean distance. To visualize those groupings, the hclust function was run to create a dendrogram, which classified the banks into groups based on similarity (Murtagh 1985). Finally, we compared the groups made from the cluster analysis to the groups made from our field-based classification. The purity of each group was plotted on histograms using bank ID.

In the field, riverbank cross-sections were performed using a Leica NA720 stadia for a series of representative banks. Height variation, deposit type, vegetation cover, and ice scar presence were integrated into a diagram to schematize the shape and composition of the bank. Bank cross-sections were useful to determine important flood stages, such as the bankfull level. The bankfull level was identified according to the procedure described by Williams (1978), who defines it as the height of the relatively flat depositional surface adjacent to the river, corresponding to the elevation of the active floodplain (Wolman and Leopold 1957; Williams 1978). This level was used as a reference point when comparing the riverbank types. These data were used to create a scaled and highly detailed topology of each riverbank type.

3.2 Channel characterization

The channel characterization of the hydraulics and hydro-geomorphology (see Table 3) associated with ice-impact features was performed in three steps. The first step was to recreate the fluvial environment of the Mistassini River in ArcMap Geographical Information System. Therefore, river polygons, road polylines, and a DEM (cell = 20 cm × 20 cm) were generated using BDTQ (1:20 000) topographic data.

Table 3 Description of the methods used to sample the hydraulic and hydro-geomorphological variables
Full size table

Subsequently, the river polygon was segmented using the riverbank groups identified from the cluster analysis. To represent each side of the channel, the Mistassini River polygon was split in half by a centerline. The points georeferenced during fieldwork were then plotted on the river polygon as spatial indicators of the beginning and ending of each bank. Those points were used to construct smaller polygons. Finally, for each riverbank polygon within a section, hydraulic and hydro-geomorphological channel characteristics were extracted from the ArcMap environment created previously and 2012 orthophotographs. The channel characteristics and their acquisition methods are all shown in Table 3.

3.3 Ice cover characteristics

In addition to channel characteristics, ice thickness data were collected. During the February 2013 fieldwork campaign, 80 holes were drilled across the four river sections, with a Kovacs® ice drill to get information about thickness and ice/depth ratio. A kriging interpolation was run in ArcMap with those points to get more information about the ice cover thickness variation along the four sections. A possible limitation of our study is that these data are punctual in time and related to the hydro-climatic conditions of winter 2012–2013. Thus, despite a possible year-to-year variability, we assumed that spatial variations in ice thickness remain proportional between the sections.

3.4 Tree-ring reconstruction of ice-jam frequency and magnitude

Sixty-three coniferous trees (spread across the four sections) were sampled. Trees were sampled when they presented at least two ice scars from different events. Trunks were cut at the scars’ level. Samples were sanded, scars were dated, and for each section, an ice-jam chronology was generated and allowed the intercomparison of river sections based on ice-jam frequencies. Ice scar heights were also used as an indicator of the relative magnitude reached by an ice-jam (Henoch 1973; Smith and Reynolds 1983; Smith 2003). It is assumed that the scars’ height is an indicator of the level reached by water and ice rafts during ice-jam events. In this study, maximum scar heights were measured from the bankfull level with a TruPulse 360B Laser Rangefinder. Ice-scarred trees were all georeferenced, and those GPS measurements were reported on a map to be compared with the geomorphological features identified at Sect. 3.1.

4 Results

4.1 Riverbank classification

During fieldwork (see Fig. 3) and with the cluster analysis (see Fig. 4), two classes of banks could be distinguished using morphological and ecological characteristics: depositional (A–B) and erosional banks (C–D–E–F).

Fig. 3
figure 3

Riverbank photographs separated into two classes: depositional banks composed of two groups and erosional banks composed of four groups (photos by S. Morin, June 2012–2013)

Full size image
Fig. 4
figure 4

a Dendrogram of bank types according to the cluster analysis, which is based on the dissimilarity between the geomorphological and ecological variables of the banks. The more similar the banks are, the closer they are in the dendrogram. The number of groups in the cluster analysis is based on the number of groups in our first classification, which was determined by geomorphological processes. All groups are described in Sect. 4.1; b Cluster purity. For each group made from the cluster analysis, we compared the number of segments in our first classification using the geomorphological processes. The box represents the groups from the cluster analysis, the X-axis represents the groups from our first classification using the geomorphological processes, and the Y-axis the number of segments from each group

Full size image

Group A banks are predominantly formed by depositional processes. They present a very low and constant slope (mean = 10°, rms = 6.58), where recent alluvia form a sandy or bimodal bar in front of the riverbank. The bank is composed of graded, fine grain elements, such as sand, clay, or silt. Vegetation covers most of the bank (mean = 96 %, rms = 16.71), especially shrubs (mean = 70 %, rms = 18 %) and riparian herbs (mean = 23 %, rms = 12.66). No root exhumation is visible, and there are generally no ice scars on the vegetation. The bankfull level is easy to identify on the bar. The cluster analysis shows that this group is apart from the others, since no node appears on the left branch. The group is considered pure, i.e., that banks found in this cluster are homogenous and composed of a high proportion of bank A and a low proportion of bank D.

Group B presents similarities with group A, but is more heterogeneous. Vegetation cover is high (mean = 89 %, rms = 18.45), particularly shrubs (58 %, rms = 19.4) and herbs (mean = 26 %, rms = 8.11), but few trees are present (mean = 8 %, rms = 7.46). Recent well-graded fine and coarse sands are found at bankfull level. However, by contrast with group A, moderate erosion steepens the bank by creating a low talus with an inclined slope (mean = 19°, rms = 9.80). Bankfull and floodplain levels are separated by about 50 cm. Ice scars are found mostly on shrubs, but also on a few trees. There is little root exhumation (1–30 %).

Group C presents evidence of intense erosion at the bottom and top of the slope. It is composed of a tall and steep slope (mean = 30 %, rms = 9.47), due to the mass movement dynamism and bottom slope destabilization that causes bank failure. Total vegetation cover is low (mean = 73 %, rms = 21.77), especially the tree’s strata (mean = 17 %, rms = 10,2). Few shrubs (mean = 34 %, rms = 17) and herbs (mean = 22 %, rms = 11.53) cover the bank. Root exhumation is very high (35 % to 50 %). Ungraded fine-grained elements, such as sand or clay, are found at bankfull level. In most cases, sand covers a clay slope. No ice scars are found on the few trees. In the cluster analysis, this group is one of the purest.

Group D, like group C, presents clear evidence of intensive erosion. The two-level structure separated by a steep and irregular talus of approximately 2-m-high suggests that severe abrasion occurred at levels higher than bankfull stage. The floodplain level is on average 50 cm higher than the second level. The average bank slope is less steep than in group C (mean = 25°, rms = 4.72). Generally, the bank is composed of graded fine-grained elements, such as sand, clay or silt. Recent fine alluvia accumulations were found on top of the second level. The strength and frequency of those erosive events also impacted the vegetation cover composition. The few trees (9 %, rms = 6.56) found on the upper terrace presented ice scars. Heavily scarred shrubs dominate these banks with 43 % cover (rms = 16.5) and seem to show great resistance to the abrasive material, due to their deep roots and flexible trunk. Total vegetation cover represents 79 % (rms = 16.13) of the bank surface. Also, tree and shrub roots are exhumed at more than 35 % in most of the cases. The evidence of ice erosion is undeniable for this bank type. The presence of ice scars and loose alluvia on the second terrace confirms that an abrasive material, such as ice rafts, could have reached this level. Undercutting is also present below bankfull level. In the cluster analysis, this group presents the most homogenous cluster.

Visually, group E has more similarities with group D than with any other groups. However, the cluster analysis reveals that this group is clearly apart from group D and is the opposite of group A. All the other bank types are found in low proportions in this group, which makes it the least pure. The bank is composed of a low talus with a steep slope (mean = 22 %, rms = 7.39) that separates the floodplain and bankfull levels, as in group D. Trees cover 10 % (rms = 6.55) of the bank surface, shrubs 43 % (rms = 21.88) and herbs 30 % (rms = 12.7). However, few ice scars are found on the trees and shrubs. Root exhumation is low, ranging between 0 and 30 %, but occasionally reaches 30–50 %. Ungraded fine and coarser sands are found at the bankfull level. The general shape of this group differs from group D by the dominance of the undercutting processes that deeply eroded the fine deposits. Sometimes, these banks seem to defy gravity since several trees and shrubs hang in the void, holding on by their roots.

Group F presents erosion at its final stage, where lateral migration impinges on the valley’s bedrock wall. Consequently, the shape of these banks is less influenced by the hydro-geomorphological process than the other banks, since it depends on the valley rock composition and geometry. These banks are mainly composed of rock, which is visually different from all the other bank types. However, the cluster analysis found similarities with group B proprieties. Bank slope is approximately 20 % on average (rms = 8.39). Shrubs (mean = 48 %, rms = 22.25) and herbs (mean = 21 %, rms = 12.51) are the most common. The few trees (mean = 7 %, rms = 4.01) that have conquered the rocky substratum, generally present ice scars. There is no root exhumation, and the bankfull level is not clearly visible. The cluster purity is moderate, since three bank types composed it.

4.2 Spatial distribution of riverbank types along the four sections

The proportion of a bank type in a given section can be expressed as a ratio of the bank’s length to the section length. Some groups change proportions within each section or from upstream to downstream, while others are almost absent (see Fig. 5). Group A occupies a high proportion of S1’s total length (47 %) and generally decreases in the downstream direction. Group B’s presence is constant within each section, though in low proportions, but it is almost absent in S2. Group C is present in high proportion in S1 (23 %), S2 (24 %), and S4 (19 %), but is almost absent in S3. The two-level banks (group D) increase in the downstream direction, but occupy a similar proportion in S3 (27 %) and S4 (28 %). Proportion of bank D is significantly higher in S3 than it is in S4 (z test, Z c = 4.79 > Z α/2 = −1.96). Group E increases in the downstream portion of section S1 to S3, but is almost absent in S4 (5 %). Finally, group F occupies a high proportion in S2 (27 %), S3 (18 %) and S4 (19 %), but is sporadic in S1.

Fig. 5
figure 5

Length occupation proportion of each bank type within a section. The length (m) occupied by each bank type across a whole section is first specified. The occupation percentage is in bracket below. Total length (m) is the sum of the lengths of both river banks in one section. Dashed lines represent the group D bank type

Full size image

Bank types are distributed according to the dominant geomorphological process that takes place in the channel (see Fig. 6, 7). When sinuosity is high, group A usually occupies the convex bank, in opposition to group C or E, and occasionally D. In linear sections, group C and E alternate with group D. The same bank types can also face each another. This situation is mainly observed for group D at the upstream end of an island.

Fig. 6
figure 6

Spatial distribution of riverbank types along the S1 and S2. The dashed segments represent the group D bank type

Full size image
Fig. 7
figure 7

Spatial distribution of riverbank types along S3 and S4. The dashed segments represent the group D bank type

Full size image

4.3 Channel characteristics

The four sections present important variations in their hydraulic, hydro-geomorphological, and ice cover characteristics (see Fig. 8). Group D’s spatial distribution, representing the main ice-impact features in the Mistassini River, was added to Fig. 8 to illustrate which channel variables might explain their presence. As mentioned in the study area, sinuosity and slope are the variables that distinguish the four sections most from each other. S1 is visually more sinuous than S2 and S3, while S4 is meander-like. The slope decreases from upstream to downstream, while it passes from the highlands to the lowlands. A constant and low slope characterizes S1 (a = −0.0008) and S4 (a = −0.0017), while it gets steeper in S2 (a = −0.0082) and S3 (a = −0.0084).

Fig. 8
figure 8

Downstream evolution of the hydraulic, hydro-geomorphological and ice cover variables sampled along the four sections. The dashed line represents the linear extrapolation of the mean in unsampled sections. Spatial distribution of group D banks is also represented in this figure; elevation refers to the thalweg; (a) represents the mean channel slope

Full size image

Depth and width show important variations between sections and within a section. There is no clear proportional relationship between mean flow depth and distance downstream, whereas width definitely increases with the mean width in S4 (W = 296.0 m) being three time larger than the mean width in S1(W = 116.3 m). Depth varies, however, twice as much in S4 (rms = 1.0 m) than in S2 (rms = 0.4).

Channel entrenchment ratio does not vary linearly in the downstream direction. S1 (mean = 4.89, SD = 1.58) and S4 (mean = 6.23, SD = 4.63) are more entrenched than S2 (mean = 2.44, SD = 1.42) and S3 (mean = 1.88, SD = 0.68). Width/depth ratio and flow area increase from upstream to downstream. S4 presents important standard deviation for both of these variables (W/D ratio SD = 154.42 and flow area SD = 274.13). As for the ice cover thickness, it presents a slight increase in the downstream direction and is more variable in S4 than in the other sections (F test, F = 1.63E−6, p value = 0.70, α = 0.05).

4.4 Dendrochronological analysis of the ice-jam regime across the four sections

The spatial distribution of ice scars reveals that ice-jams occur in all four sections, but at different frequencies (see Fig. 9). One of the most striking aspect is that ice-jam frequencies decrease in the downstream direction, passing from 0.4 event/yr in S1 to about 0.2 event/yr in S4. Moreover, only two events are common to all sites (2011 and 1990).

Fig. 9
figure 9

Spatial distribution of ice-jam events in the four sections using a dendrochronological analysis. The frequency was calculated by dividing the total number of events by the 43-year time period

Full size image

Ice scar heights (see Fig. 10) also vary considerably across the four sections, and by contrast to ice-jam frequency, they tend to increase in the downstream direction. For example, S1, S2 and S3 present the lowest ice scars (about 200 cm), while S4 exhibits significantly higher damages (about 600 cm; ANOVA, F = 14.67, p value = 0.000378). This relation is clearly perceptible in both years 1990 and 2011 (see Fig. 10b, c). Ice-jam heights are also much more variable in the downstream section (F test, F = 0.0002, p value = 0.34).

Fig. 10
figure 10

a Spatial distribution of ice scar heights above bankfull level along the four sections in all the years sampled, b spatial distribution of 1990 ice scar heights above bankfull level along the four sections, c spatial distribution of 2011 ice scar heights above bankfull level along the four sections. The box plots present the variation of heights, the second and third quartile, the maximum and minimum and the median

Full size image

5 Discussion

This study aimed to spatialize the morphological signature of river ice-jams along Mistassini River. Specific landforms and ice-scarred trees helped to shed light upon the interrelationships between the ice-jam regime and the impacts on the fluvial geomorphology of northern rivers. Ice processes, such as ice erosion and ice push, are well known to affect the bed and banks of a river, as well as fish habitats (Beltaos 1995). In turn, river morphology also influences the frequency and magnitude of those processes. Several features, such as wide and shallow channels, highly sinuous meanders, and steep slope channels (Beltaos 1995; Beltaos and Burrell 2003; Taylor et al. 2008), can be considered factors aggravating ice processes. Therefore, our study helped to determine the links between the spatial variability of such ice-induced landforms, taking into accounts the frequency and magnitude of ice-jam events and channel features.

5.1 Two-level banks: a morphological imprint of recurrent ice-jams

Using simple morphological and ecological characteristics, our study provides a typology of riverbanks that can help distinguish and locate ice-induced landforms in a complex and highly variable fluvial landscape or, at least, determine where ice-induced erosional processes are morphogenetically dominant (see Fig. 11). The identification and spatialization of ice-induced landforms may be useful for river management programs and could help authorities retrieve information on local ice regimes (ice-jam frequency and magnitude), in problematic, ice-affected rivers such as the Mistassini River.

Fig. 11
figure 11

Bank typology established from the geomorphological and ecological descriptions and the mapping of river banks. Only four banks on six are presented here due to the inconsistency of the morphology of group B and F

Full size image

According to our topology, many bank types may have been affected by river ice in one way or another, but ice-induced erosional processes seem to be essential in the formation of two-level banks (bank type D). For example, riverbanks C and F (see Fig. 11) both result from a combination of erosional processes, but the role of ice as a formative agent appears unclear, if not negligible, as evidenced by the low density of ice-scouring marks on the vegetation and the banks’ sediments. In this context, erosional processes, such as channel degradation, undercutting (bank F) and active lateral meandering are more likely to form maintained such banks in the long term. Low-magnitude ice-jams may accelerate the pace at which fluvial erosion operates, but do not seem to be indispensable in generating landforms and maintaining them over long periods of time and over vast portions of landscape. Concretely, in our study, when we compared only the bank groups from the first classification (determined by geomorphological processes) with groups from the cluster analysis, the polygenic banks identified by the first method could lead to misclassification in the cluster analysis if other processes affect their morphology and ecological composition. The banks from S3 are the most affected by this situation, with five misclassified banks. Misclassification here refers to a situation where the first identification determined by just the geomorphological processes and the cluster analysis that was determined by both geomorphological and ecological bank characteristics did not concur.

However, two-level banks represent unique landforms that unequivocally reveal the impacts of ice processes in northern fluvial landscapes. Our cluster analysis revealed that the geomorphological and ecological properties of this bank type could be isolated from few simple field-based metrics. These banks present an easily recognizable two-level structure separated by a steep (average slope = 25 %) and irregular talus of about 2 m high, overlaid by freshly deposited alluvia. These banks are covered by an abundant stratum of shrubs and few trees, both marked by ice scars. This bank type is associated to intense erosive events that occurred well above the bankfull level (see Fig. 11d). The ice-related origin of these landforms has been recognized by Boucher et al. (2009), who described similar erosional features on the Necopastic River (Quebec). They hypothesized that ice run increases shear stress and induces mechanical abrasion, when the channel bankfull capacity is exceeded during ice-jams. Consistent with this idea, Turcotte et al. (2011) reported that during an ice-jam release, important quantities of sediments are transported, due to turbulence and ice rafts rubbing on the channel banks and bed. Such sediments may be deposited on top of ice flood terraces during ice-jams. McPherson (1966) also observed the push of recent alluvia on the top of the banks, creating a long and discontinuous levee on the upper terrace, just as we observed on the Mistassini River.

These results support the hypothesis that river ice is an important morphological component of northern river landscape. Their own, unique, morphological signature was found in a northern temperate climate with a post-glacial past, which is further south than what is reported in the literature. The other studies that described such landforms were mostly realized in a glacial subarctic environment (MacKay and MacKay 1977; Smith 1979) or in a high-boreal continental climate with a post-glacial history (Boucher et al. 2009). However, more studies on other rivers need to be done in order to see whether those forms vary in different hydro-climatic and geomorphological contexts.

5.2 Ice-jam occurrences and morphological forcing

The spatial distribution of the two-level banks combined with a qualitative description of very simple hydro-geomorphical measurements in the four sections allowed us to highlight two recurrent scenarios where those ice-induced landforms are found. Yet, in all cases, the decrease in the channel capacity and slope stands out as the most important variable associated with the occurrence of group D banks (see Fig. 12).

Fig. 12
figure 12

Most recurrent scenarios where the two-level banks are found. These scenarios present visual examples of how ice affects water and ice evacuation, which could lead to the formation of two-level banks. Scenario 1 presents how water and ice travel in a channel cross-section, which is mild, wide, shallow, and covered by a thick layer of ice. Scenario 2 presents how the water and ice travel in a bird’s-eye view of an entrenched narrowing channel. Light gray color and snowflakes represent the ice cover, whereas the dark grey color portrays water. The white forms represent ice rafts. Black lines represent the elevation curves

Full size image

As the proportion of ice-induced banks increases in the downstream direction, channel slope becomes more gentle. River bed is also wider and shallower, which is reported as the most common congestion situation associated with recurrent ice-jams (Beltaos 1995). When a jam occurs in such a channel, water and ice rise to a level that depends on the duration of the event (Beltaos 1995). It generally reaches higher levels than those reached during open-water floods (Prowse and Beltaos 2002). Therefore, the shear stress induced by the action of ice on the banks probably enhances scouring at those higher levels, which can probably create the steep and irregular talus of the two-level banks. This scenario is mostly found in S4.

Scenario 2 presents a case where an entrenched channel becomes narrower in the downstream direction, which possibly limits the channel capacity to evacuate important ice volumes. Width decrease associated with a natural constriction, island or bar is a well-known congestion factor that favors ice-jamming processes (Pariset et al. 1966; Michel 1972; Beltaos 1995). This scenario is found in S2, S3 and S4.

The spatial distribution of the two-level banks across the four sections showed greater proportions in S3 and S4. S3 presents several sites where channel is suddenly narrowing, due to natural constrictions, which decrease the channel capacity to transport high ice discharge and could lead to channel congestion. S4 presents multiple factors that can lead to the two-level banks appearance: (1) W/D ratio and flow area present the highest mean values, but also the most variable; (2) the channel slope is milder than S2 and S3; (3) the presence of islands and sand bars in the middle of the stream induce channel narrowing at several entrenched sites. Furthermore, S4 presents a moderate sinuosity. As mentioned by Beltaos (1995), a highly sinuous channel is known to aggravate ice processes. However, it does not seem to affect the bank morphology of S1, as it does in the downstream sections. Also, the presence of high cascades at the beginning of S4 can produce a high quantity of frazil ice, which could lead to frazil accumulation in the flatter river segments and eventually congestion. Frazil production is linked to the supercooling process induced by turbulent flows and very low temperatures (Ye and Doring 2004). Therefore, when the air temperature falls below the freezing point of water, a high quantity of frazil ice can be generated by the 20-m-high cascades in the upstream segments of S4. Frazil ice possibly accumulates in the downstream direction, where it increases the ice cover thickness and decreases the ice transport capacity of the channel.

More studies are needed to better circumscribe the hydro-geomorphological variables that influence the occurrence of ice-jams. Those findings could help prevent infrastructure destruction and human loss and could be integrated in hazard mapping. For example, better-suited flood map techniques should be used at locations where ice-jams are likely to occur (Demers et al. 2014).

5.3 Ice-jam regime

The tree-ring ice-jam regime reconstruction illustrated how the occurrence of erosional, ice-induced features may be linked to ice-jam magnitude and frequency along the Mistassini River. Our results suggest that a strong, inverse relationship (in the downstream direction) exists between the frequency of ice-jam events and the occurrence of two-level banks—implying that these features do not necessarily tend to localize where ice-jams are the most frequent. It is interesting to contrast these results with those of Boucher et al. (2009) who estimated that one event every 5 years (0.2 event/yr) is needed for the maintenance of such banks on the Necopastic River—a value that is very close to what we calculated in section S4 (0,19 events/yr). This frequency is also consistent with the recurrence calculated in other ice-affected rivers, for example one event every 4.9 years in the Peace-Athabasca Rivers, Alberta (Smith 2003) or one event every 6 years in the Missouri River, North Dakota (Wuebben and Gagnon 1995). However, even if the upstream section of the Mistassini River presents much higher frequencies (e.g., one event every 2 years in the S1 section), two-level banks do not seem to be as abundant, and evidence of ice as a dominant, channel-forming agent become attenuated as ice-jam frequency increases in the Mistassini River. Obviously, a minimum frequency is an essential condition for ice-induced banks to form and maintain in the fluvial landscape, but our study suggests that frequency alone is not sufficient to explain the occurrence of ice-induced banks.

We hypothesize that not only the frequency, but also the magnitude of these events needs to be taken into account to adequately characterize ice-jam regimes and determine how they impact riverbanks. In the Mistassini River, two-level banks are found in greatest proportions in the downstream sections, where ice scars are the highest. It is therefore reasonable to assume that, even if ice-jam floods vary in magnitude between years, they tend to be proportionally higher in lower parts of the basin. Magnitude is associated with the levels reached by water and ice during ice-jams but might also reflect the amount of shear transferred to the banks during the ice flood. In turn, magnitude might be strongly influenced by local bank heights and floodplain morphologies (Beltaos 1995). Indeed, the downstream sections (S3–S4) of the Mistassini watershed present multiple aggravating factors, as described in the previous section. Therefore, channel morphology in these sections enables ice-jam flooding to reach dangerously high levels, which result in significantly higher ice scars on trees, and a higher proportion of ice-induced erosional banks.

6 Conclusion

Ice-jams are characteristic phenomena of cold region\ fluvial environments. The aim of this study was to use the morphological signature of ice-jams along the Mistassini River to document their spatial extent, magnitude, and frequency. On the Mistassini River, their impacts on morphology and vegetation were successfully identified and used as indicators of the ice-jam regime. One particular form, the two-level banks, clearly results from the effect of ice scouring on banks. Those landforms present a two-level structure separated by a steep and irregular talus (average slope = 25 %) of approximately 2 m high, overlaid by freshly deposited alluvia, as well as little and scarred vegetation. This bank type is associated to an intense erosive event that occurred higher than the bankfull level. Such landforms are found mostly downstream in two dominant contexts, related to the channel capacity to transport high ice discharge: (1) wide and shallow channel with a mild slope; (2) when there is a sudden channel narrowing. Dendrochronological analysis of ice-scarred trees and the spatial distribution of the two-level banks allowed us to allocate a recurrence rate of 1–1.5 events every 5 years and a high-magnitude level to those forms, which are mostly present in the downstream sections.

Our study provides new insights into the incidence of river ice as a geomorphological agent of the fluvial landscape. However, it also brings new concerns about the role of fluvial morphology components on ice dynamics. Therefore, the spatial prediction of problematic areas along a stream remains a challenge. Yet, no predictive model has been adequately developed to predict potential ice-jamming sites. Statistical and determinist models remain imprecise, since they can not take into account all spatial and temporal scales, neither the morphological and hydro-climatic components of fluvial hydro-system (White 2003). Still, predicting the sites where ice-jam probabilities are increased could help prevent infrastructure destruction and human loss. Mitigation measures can also be taken at these sites to reduce the effects of ice-jam floods, such as a surveillance system for ice cover melting placed at strategic points (“Vigilance” program of OSCQ), reservoirs, eco-friendly ice-control structures (ICS) or adequate designs for bridges, and hydrological structures (Beltaos 2008). In combination with well-known hydro-climatic factors, a more effective and global perspective of ice-jam dynamics among a river could be achieved.