1 Introduction

Major ice jams have occurred at several locations along the main stem of the Saint John River (SJR) and along several of its tributaries. Ice jams have resulted in ice impact or forces that damaged and destroyed bridges, and in flooding of residential property and highways. Among the communities worse affected by ice-related flood damages is Perth-Andover where devastating flooding occurred in 1976, 1987, 1993, 2009, and 2012. Ice-related flood damages have also been sustained in many other riverside communities including, though not limited to, Dickey and Fort Kent in Maine, and St.-Francis (NB), Clair, Edmundson, Ste-Anne-de-Madawaska, Rivière-Verte, Saint-Léonard, Simonds, Hartland, and Woodstock in New Brunswick.

River ice-related problems are common in New Brunswick throughout the winter and spring seasons. Ice jams frequently result in damage or destruction of bridges, and flooding which causes property damage and other direct and indirect socio-economic damages. For example, in 1970, ice runs and jams destroyed 32 bridges in New Brunswick and resulted in about $27.5 million in damages [estimated in 2018 dollars; these estimates were obtained from the Flood Disaster Database (NBELG 2022; NB Subcommittee on River Ice 2011) and adjusted to 2018 values using the Bank of Canada inflation calculator based on monthly consumer price index]. In 1976, ice jams on the SJR resulted in severe flooding and damages at Perth-Andover and Woodstock that totaled approximately $15.6 million [2018 dollars]. Subsequent ice-jam floods in 1987, 1993, and 2012 at Perth-Andover and elsewhere along the SJR have resulted in over $ 57 million in damages [2018 dollars]. Actual flood damages during these could have been much higher due to undocumented and intangible damages, such as the hardship endured by flood evacuees.

The purpose of the paper is to present an appraisal of ice regimes, and ice jamming along the main stem of the SJR based upon a synthesis of existing information from multiple sources, combined with previously unpublished information and analyses. Background information on the SJR and its basin is provided in the next sections of this paper, followed by a synopsis of the basin’s ice regimes and history of ice jamming. Hydroclimatic and river-ice trends affecting the ice regime are then discussed followed by concluding remarks.

2 The Saint John River basin

2.1 Basin geography

The SJR Basin (SJRB) lies in a broad arc across southeastern Quebec, northern Maine, and western New Brunswick (Fig. 1). The total drainage basin area is 55 200 km2, of which approximately 19 700 km2 (36%) lies in the State of Maine, USA, 7100 km2 (13%) is in Quebec, and 28 400 km2 (51%) is in New Brunswick (Environment Canada 1974). The SJRB drainage area upstream of Woodstock is 37 100 km2 (Acres 1976), which is about 67% of the entire basin area.

Fig. 1
figure 1

Saint John River Basin (SJRB) upstream of Woodstock

The SJRB is a hard rock, recently glaciated, thin-soiled forest area extensively cleared of virgin cover in the “past century and a half” (Hare et al. 1997). The overburden throughout the basin is mainly glacial till composed of silty, gravelly sand with cobbles and boulders deposited in a blanket of varying thickness over the bedrock. Most of the SJR upstream of Woodstock, NB, lies in the Chaleur Uplands, an extension of the Appalachian uplands and highlands of eastern North America.

2.2 Basin climate

New Brunswick generally has a snow-forest climate with no distinct wet or dry season (Environment Canada 1974). The SJRB is considered to have a cold climate (Peel et al. 2007), with most of the basin having a humid continental climate (SJRBB 1975) except near the Bay of Fundy coast. The northern portion of New Brunswick usually experiences colder and longer winters compared to the southern portion nearer the Bay of Fundy, thereby resulting in differences in the timing of ice freeze-up and breakup between the northern and southern regions of the Province.

Climate Normals for selected stations in the SJRB are provided in Table 1. During most of the winter, the basin has a snow cover which during periods of rising temperatures in the spring contributes to increased runoff. The long-term average end of March snowmelt potential is presented in Fig. 2.

Table 1 Climate normals for SJRB climate stations
Fig. 2
figure 2

Long-term end-of-March snow-water-equivalents (SWE), SJR basin (Courtesy of Mélanie LeBlanc, Surface Water Planning Officer, NB Department of Environment and Local Government)

3 The Saint John River

3.1 General

The SJR, also referred to as the Wolastoq River, is approximately 673 km long, falling approximately 482 m from its origin at Little Saint John Lake, Maine, USA, to its mouth at the Bay of Fundy in the city of Saint John, New Brunswick (NB). Figure 3 depicts the bed profile of the river. For this paper, the river is divided into three stretches: the upper Saint John River (USJR) upstream of Grand Falls, the middle Saint John River (MSJR) between Grand Falls and Mactaquac Dam (MD), and the lower Saint John River (LSJR) downstream of Mactaquac to the river mouth.

Fig. 3
figure 3

Bed profile of the Saint John River from Dickey to the City of Saint John. The bed profile is based on the thalweg elevations from cross-sectional surveys done during the past 50 years

The SJR originates in three tree branches: a southwest branch originating in Little Saint John Lake on the Maine–Quebec border near Saint-Zacharie, a branch starting in the Saint John ponds in northwestern Maine, and a northwest branch rising near Lac Frontière near the southeastern Quebec border. These branches combine to form the main river flowing northeastward through the forests of Aroostook County, Maine. Dickey is the furthest upstream settlement in northern Maine, and the site of a devastating ice-jam event in 1991. Below Dickey, the Saint John is joined by the Allagash and Fish Rivers, two major northwardly flowing tributaries that lie entirely in Maine, and the St. Francis River flowing southward. Continuing its northeasterly course, the river flows past Clair (NB) and Fort Kent (Maine), Baker Brook (NB), and then flows past Edmundston (NB). At Edmundston, the Madawaska River enters the SJR, which turns southeast, passing the NB communities of Rivière-Verte, Ste-Anne-de-Madawaska, and Saint-Léonard. These latter communities are located along the headpond created by the barrage at Grand Falls. Near Grand Falls, New Brunswick, the river enters entirely into New Brunswick, and starts to flow south. The Aroostook and the Tobique rivers flow into the SJR upstream of Perth-Andover, a community that has sustained several severe ice-jam floods. Downstream of Perth-Andover, the river passes Florenceville, Bath-Bristol, Hartland, Woodstock, and Nackawic. Along the LSJR lies the capital city of Fredericton and the extensive Maugerville to Lower Jemseg floodplain, before the river empties into the Bay of Funday at the city of Sant John.

Hydroelectric dams were constructed along the main stem of the SJR at Grand Falls (1925), Beechwood (1957), and Mactaquac Dam (1968–1969), and along several tributaries including the Madawaska, Aroostook, and Tobique Rivers (see CRI 2011). These dams have affected the hydrology, ice regime, and ecology of the SJR. Mactaquac dam, approximately 80 km downstream of Woodstock, has eliminated the devastating ice-jam floods that once occurred along the downstream tidal portion of the SJR For that reason, the river upstream of Mactaquac Dam is the focus of this paper, especially the section of river from Dickey to Woodstock.

3.2 Hydrology

Discharge and water level records exist for several United States Geological Survey (USGS) and Water Survey of Canada (WSC) hydrometric stations along the SJR. An important station for this study is the Saint John River at Fort Kent (station no. 01AD002). The mean annual flow on Saint John River at Fort Kent (station no. 01AD002, natural flow, upstream basin area = 14 700 km2) is 279 m3/s (based on 96 years of record, 1927–2020); the average winter (JFM) flow is 97 m3/s (ECCC 2022). Hydrology is also considered later in this paper when discussing hydroclimatic change (Sect. 7).

4 River-ice regimes

4.1 Existing typical ice regimes

An ice cover starts to form along the SJR usually by late November and remains until mid-March to mid-April. The spring breakup typically occurs in April, but mid-winter breakups are also known to occur on occasion. Systematic and detailed field observations and measurements were performed between Dickey and Grand Falls during the years 1993 to 1997 under a joint research program (International Saint John River Ice and Sediment study or ISJRISS) that was carried out jointly by NB Environment, NB Power, and Environment Canada.

Between Dickey and Grand Falls, ice-cover formation begins at the headpond above the Grand Falls dam and quickly propagates upstream within the relatively tranquil reach between the dam and Edmundston. At this site, warm effluents from local pulp and paper mills suppress ice formation throughout the winter. Farther upstream, increasing amounts of slush are generated in the relatively rapid upper reaches, leading to formation of ice pans, which eventually lodge at a few congestion-prone sites, e.g., near Saint-Hilaire, a few km upstream of Edmundston. Ice covers then propagate upstream from these sites and one often encounters alternating open-water and ice-covered segments during the freeze-up period. Once established, the ice cover thickens at various rates, depending on weather conditions as well as on the time of lodgment and local flow conditions.

Freeze-up ice jamming along the SJR typically occurs during low flows, and generally does not result in serious ice-related flooding. Therefore, information on freeze-up ice jamming along the SJR is rare. However, freeze-up ice jams are known to affect flow and ice formation in downstream reaches, and thereby indirectly have the potential to affect MWB and spring breakup.

Spring breakups generally follow a common pattern along the USJR upstream of Edmundston. As temperatures rise and approach 0 °C, several river segments between Dickey and Fort Kent break up and small jams form. With increasing flows, the small jams release, leading to formation of major runs that can easily dislodge the remaining sheet ice cover between St. Hilaire and Edmundston. As these ice runs encounter the ice cover below Edmundston close to the community of Ste-Anne, the relative thickness and strength of the local sheet ice cover and the very low slope of the water surface impede ice movement and can cause major ice-related floods, that can extend upstream to Edmundston (e.g., 1991, 1993). When this jam releases, the ice eventually runs past Grand Falls, potentially enhancing the length of downstream ice jams.

The section of the SJR from Grand Falls to Beechwood can be divided, from an ice formation perspective, into two sub-reaches. The reach between Beechwood and Aroostook River is influenced by the backwater created by the Beechwood dam, which results in low flow velocity. Ice formation commences as a skim ice sheet over the headpond that thickens thermally under negative air temperatures. Between the Aroostook River and Grand Falls, the ice cover forms later, in a fashion typical of what occurs in uncontrolled river segments and is partly governed by local hydraulic conditions and channel geometry. (Uncontrolled river reaches in the context of this paper refers to areas of river that are not along reservoirs or head ponds, and therefore ice processes more resemble those of an unregulated river.)

The ice cover in the three headponds of the SJR normally consists of blue ice resulting from thermal growth of the underside of the ice cover and white ice resulting from the freezing of snow cover due to flooding of the ice cover or rain. The ratio of blue ice compared to total ice varies from year to year depending upon antecedent winter weather conditions. In 1976 and 1987 (years of significant ice-jam flooding at Perth-Andover), the percentages of blue ice to total ice thickness were approximately 95 and 60%, respectively (Acres 1988).

Acres (1976) presents plots of end-of-winter blue ice thickness and total ice thickness against probability of non-exceedance for the three headponds. The average total and blue ice thicknesses corresponding to probability of non-exceedance of 10, 50, and 0% are presented in Table 2 (as estimated from Figs. 4.1, 4.2 and 4.3 of Acres 1976).

Table 2 Winter blue ice thickness and total ice thickness for three headponds along the SJR

Measurements between Feb 26 and Mar 5, 1997 revealed considerable longitudinal variability in the laterally averaged sheet-ice thickness: thickest near Dickey and Grand Falls, thinnest between Edmundston and Fort Kent, except for a short segment near St. Hilaire, where an ice cover forms relatively early. The range of measured thickness extended from ~ 0.3 to ~ 0.7 m (Beltaos et al. 2003). This pattern can be significantly modified by mid-winter breakup and ice jamming.

Generally, the spring breakup process can be mechanical or thermal. Mechanical breakup results from rapidly rising flow and water level, dislodging, mobilizing, and breaking up the ice cover, which may be arrested at several locations between Grand Falls and Perth-Andover. Mechanical breakup is more frequent in the reach upstream of the influence of the dam’s backwater. Thermal breakup takes place when the transition from winter conditions to spring conditions is gradual, allowing the ice cover to deteriorate and melt in place. This is more common in the reach affected by the backwater upstream of the Beechwood dam. However, if the change from winter to spring conditions occurs rapidly, the ice cover downstream of Perth-Andover could breakup while it is still competent and strong. This combined with high river flows, natural or as augmented by the release of an upstream jam, could result in severe jamming within the Beechwood headpond.

4.2 Non-typical conditions, mid-winter breakup

Uncontrolled river segments of the study area are prone to mid-winter breakup and jamming when a mid-winter thaw occurs over the basin. Such thaws are usually accompanied by rainfall, enhancing melt and resulting in sharply rising flows, which may or may not be large enough to initiate breakup.

De Coste et al. (2022) used a recently released Canadian River Ice Database (CRID; de Rham et al. 2020), containing river-ice data from 196 rivers across Canada obtained from time-series analysis, combined with the Natural Resources Canada (NRCan) gridded daily climate dataset, to identify a list of potential hydrologic and climatic drivers for MWB events. A new threshold for the prediction of MWB initiation based on climatic conditions: 4 melting degree-days and 8 mm of precipitation in the preceding 20 days was proposed (De Coste et al. 2022). Weather data at Edmundston, support this criterion for known MWBs that occurred in 1981, 1995, 1996 (twice) and 2006, but refute it for a mid-winter thaw that occurred in 1979, which did not culminate in a breakup event.

A MWB followed by resumption of cold weather can considerably alter the spatial distribution of ice-cover thickness and overall resistance to dislodgment and breakup, which also depends on freeze-up levels. Where midwinter jams form, both thickness and freeze-up level are greater than what they would be otherwise. Resistance to breakup is therefore enhanced locally and may result in the priming of new jams when the spring flow hydrograph rises and spring breakup is initiated (Beltaos et al. 2003).

A MWB can cause formation of one or more jams along the reach where the ice cover has been dislodged and broken up (Beltaos et al. 2003). Upon return of the cold weather, a new ice cover will eventually form along the open reaches upstream and downstream of the midwinter jams but will not be as thick as it would have been if a midwinter thaw and breakup had not occurred. If the MWB event occurs late in the winter season, it is conceivable that the open reaches may not form a complete ice cover. Along reaches where jams formed during the MWB event, a sheet ice cover will form near the surface by freezing of the interstitial water within the pores of the ice rubble. The freezing process is more rapid in such a medium than in plain water (Calkins 1979; Wazney et al. 2019); therefore, the sheet ice cover over the previously jammed reaches would be expected to be thicker than the ice cover that forms over reaches that remained open during the MWB event. This effect may also result in local sheet ice covers that are thicker than what would have formed if a MWB had not occurred at all, despite briefer exposure to freezing air temperatures (per measurements presented in Beltaos et al. 2003).

5 Past River-ice jams

5.1 Past ice-jam floods

Many recorded flood events along the study stretch of the SJR and its tributaries are the result of ice jams, for example, the major floods that occurred at Dickey in 1991, at Perth-Andover in 1976, 1987, 1993, 2008, and 2012; and at Simonds and Hartland in 1986. Several publications have discussed ice-jam flooding along the middle SJR (esp. at Perth-Andover), including, for example, Humes and Dublin (1988), Acres (1992), Beltaos and Burrell (2002, 2015) and Knack and Shen (2018), and along the upper SJR (upstream of Grand Falls), including Beltaos et al. (1994, 1996), Wuebben et al. (1995), Zufelt et al. (1997), and Beltaos et al. (2003). Reports on specific floods have also been prepared including LeBrun-Salonen (1983, 1985), and Acres (1988). Information on past floods along the SJR can be found in the NB Flood History Database (NBELG 2022). Table 3 presents peak water levels of past floods (i.e., flood stage equals or exceeds a defined flood level at which significant damages start to occur) for selected NB communities along the ice-jam-affected SJR upstream of Mactaquac Dam.

Table 3 Water levels during past floods at selected NB communities, m geodetic

Several bridges along the SJR have been severely damaged or destroyed by ice jams and ice runs, including railway bridges at Fredericton (March 1936), Perth-Andover (April 1987) and Newburg Junction (April 1987), and highway bridges at Grand Falls (May 1887), Hartland (April 1920), and Dickey (April 1991). Substantially more bridges were damaged or destroyed by ice-related events along tributary streams during these and other ice-related events (NBELG 2022).

Historic information is often spatially and temporally sporadic, with bias coverage given to larger communities and more significant events. Government agencies and other groups may not have consistently collected, collated, and summarized information on past events, therefore resulting in gaps or varying level of detail within the historic record. Such may be the case for the period from 1935 to 1955, in which few ice-related events along the middle SJR are reported in the flood-history database (NBELG 2022). In addition, the number of flood events along the more remote, less populated sections of the river may be under-represented compared to less remote, more populated areas. Nonetheless, examination of historic information found in the flood history database and various reports provides an understanding of the occurrence of past ice jamming and ice-related flooding along the SJR that otherwise could not be obtained. The following three sections discuss ice-jam flooding along the upper, middle, and lower stretches of the SJR.

5.2 Upper SJR ice jamming

Many communities along the USJR (i.e., the SJR upstream of Grand Falls) have been affected by ice-related floods. These communities include Dickey (Maine), Allagash (Maine), Fort Kent (Maine), Clair, Connors, Edmundston, Rivière-Verte, and Sainte-Anne-de-Madawaska. The locations of noteworthy past ice-jams along the upper SJR are shown in Fig. 4.

Fig. 4
figure 4

Locations of past ice jams along the upper stretch of the Saint John River. Channel width is exaggerated and not to scale

Ice jams are a frequent occurrence in the Big Rapids to Allagash stretch of the SJR in northern Maine, USA. Two noteworthy events were the 1974 and 1991 events at Dickey. During the 1974 event, the stage at Dickey reached an elevation of 188.9 m, with the top of the ice reported to be 189.4 m at the bridge, resulting in the superstructure being moved 1.3 m. During the 1991 ice jam, the maximum ice backwater stage was 191.7 m compared to a free-flow stage of 186.5 m (Wuebben et al. 1995). The highway bridge along with several hundred metres of riverbank and road was destroyed (Fig. 5). Shear wall heights of 7.5 m were reported by Wuebben et al. (1995).

Fig. 5
figure 5

Source Photo taken by S. Beltaos

Remnants of 1991 Ice jam at Dickey, Maine, USA, looking upstream towards piers of the Dickey bridge. The superstructure was broken and carried off by ice, while a portion of the bank and adjacent highway were eroded away. Note grounded ice and high shear walls across the river.

Two significant ice-related flood events at Fort Kent and Clair occurred during late-April in 1973 and 1974, when flows increased due to rain-on-snow events during periods with air temperatures over 15 °C (Wuebben et al. 1995). In mid-May 1969, serious flooding occurred at Fort Kent, Maine, when the SJR rose more than 4.3 m. Several families were forced to evacuate their homes and at least 15 business establishments reported flood damages.

In Madawaska County, NB, bridges have been lost at Caron Brook and Baker Brook, and highways in the vicinity has been flooded, during past flood events. Caron Brook is a confluence on a river bend; Baker Brook has an island at the downstream end of the community.

Flood events on the SJR at Edmundston have been recorded since the mid-1800s (NBELG 2022). Most of these flood events were caused by a combination of ice jams, snowmelt, and rainfall. Ice jams at Edmundston have caused localized flooding. Ice jams in the SJR downstream of Edmundston such as the 1991 Ste.-Anne-de-Madawaska ice jam, the 1993 Siegas ice jam, and the 1996 St.-Basile ice jam can also influence water levels at Edmundston. The effect of downstream ice jams on water levels at Edmundston depends upon the size and location of the ice jam and the discharge. Maximum recorded stages during spring freshets on the SJR at Edmundston are 143.191 and 143.100 m in April 1991 and May 2008, respectively (ECCC 2022). High-water events on the SJR have generally caused only minor damage to residential, business, and public properties, and infrastructure in the Edmundston area.

Acres (1976) provides a description of antecedent conditions, a detailed chronology, and an evaluation of the 1976 breakup and jamming events along the SJR from Fort Kent to Grand Falls. The formation of the ice jams occurred when the Grand Falls headpond level was high and the discharge was relatively low (Acres 1976).

Flood damages occurred in Rivière-Verte during 1969, 1986, 1987, 1989, and 2009. Flooding in the town of Rivière-Verte often resulted from overbank flow of the Green River due to ice jamming near the mouth of the Green River. These ice jams formed due to high water levels and (or) ice in the SJR. The blockage of tributary ice causing ice jamming near tributary confluences also occurs elsewhere along the SJR.

Beltaos et al. (1994) provides quantitative information on an ice-jam event that occurred at Sainte-Anne-de-Madawaska during April 1993. Major jamming had started on the upper SJR on April 11 after rain caused a rapid increase in flows. An ice jam upstream of Fort Kent released on the evening of April 11 and ran for 14 h resulting in an ice jam near Sainte-Anne-de-Madawaska (Beltaos et al. 1994). After releasing and reforming a few kilometres downstream, the Sainte-Anne-de-Madawaska ice jam remained in place for two more days allowing water levels profiles to be measured along the jam on April 13 and 14. Ice-jam thickness was measured using an under-ice profiler and shear wall heights were obtained as an approximate indicator of jam thickness shortly after the jam released (Beltaos et al. 1994).

Javes (ice-jam-release waves) on the upper SJR have been studied by field measurements/mathematical analysis (Beltaos and Burrell 2005a, 2005b; Beltaos et al. 2012) and by numerical modelling applications (Hicks et al 1997; de Coste et al. 2017). Field observations and measurements of a jave created upon the release on April 13, 2002 of an ice jam on the SJR at Ledges are reported in Beltaos and Burrell (2005a). A rise in water levels was detected at the surge meter, placed on the bank approximately 900 m downstream of the ice-jam toe, of approximately 0.85 m peaking with a second peak of approximately 1.5 m about 16 h after the initial crest (Beltaos and Burrell 2005a). The celerity of the leading edge and peak of the jave between the surge meter and the Fort Kent gauge (distance of 8.2 km) was 5.5 and 2.5 m/s, respectively (Beltaos and Burrell 2005a). Several javes were measured remotely in 2009 along the upper stretch of the SJR using pre-installed pressure loggers (Beltaos et al 2012). She and Hicks (2005) used RIVER1-D models of a 1993 ice-jam release event on the SJR (documented by Beltaos et al. 1994). The results suggest that natural channel geometry and inclusion of ice effects improve model results, especially when close to the release point of the ice jam.

5.3 Middle SJR ice jamming

Severe damaging ice runs and ice-related flood events have occurred along the middle SJR from Grand Falls to Mactaquac. Among the communities affected were Perth-Andover, Florenceville-Bristol (and surrounding area), Simonds, Hartland, and Woodstock. The locations of noteworthy past ice-jams along the middle stretch of the SJR are shown in Fig. 6.

Fig. 6
figure 6

Locations of past ice jams along the middle stretch of the Saint John River. Channel width is exaggerated and not to scale

The community most substantially affected by ice-related flood events has been the Village of Perth-Andover. Major ice-jam floods occurred along the SJR at Perth-Andover, New Brunswick, in 1976, 1987, 1993, 2009, and 2012 (see Fig. 7). These floods have been devastating and resulted in relocation or demolition of many buildings in affected areas of the community. The location of ice jams that affected flooding in Perth-Andover are given in Table 4. Substantial ice-related damages have not been reported within the village before 1976.

Fig. 7
figure 7

2012 Ice-Jam Flooding at Perth-Andover: a taken upstream of bridge from Perth looking across at Andover (right bank facing downstream). b taken looking upstream from Andover side. Photos courtesy of the NB Department of Environment and Local Government

Table 4 Ice-jam locations along the Beechwood Headpond that affect Perth-Andover flooding

When an ice jam releases, a jave (jam-release wave) can occur downstream causing sudden inundation of low-lying areas. Such an event on April 16, 1994 resulted in the drowning death of two men by floodwaters that suddenly inundated their vehicle on the Tinker Road along the Aroostook River (NBELG 2022).

Several ice-related floods have occurred in the Florenceville-Bristol and surrounding area. Ice jams were reported to have formed at SJR Bridge crossings at Florenceville during 1932 and 1987. Ice jams have also been reported to have occurred at Bath, Buckwheat Bridge, the Big Presque Isle Stream confluence, and Stickney, approximately 9 km upstream, 4.6 km downstream, 5.4 km downstream, and 8.6 km downstream of the TCH bridge at Florenceville. These locations exhibit channel features, such as gravel bars or small islands, that could impede ice movement.

At Simonds, the 1986 flood was the most damaging. On January 29, a midwinter ice jam formed at Lower Becaguimec Island near Hartland. Especially mild weather prevailed from March 26 to April 3, with rain on March 27. About noon on April 2, an ice jam formed at the upstream end of Upper Becaguimec Island (UBI), approx. 2 km downstream of Simonds. At 18:22 h, April 3, the water level at Simonds peaked at 51.14 m, causing major flooding in the Simonds area. It forced the evacuation of at least six families from their homes; some evacuations were carried out using boats and a helicopter (NBELG 2022). Between 21:00 h and 22:00 h on April 3, the ice jam at UBI released, causing the water level at Simonds to drop by 2.3 m in less than an hour. Subsequent analysis revealed that, although ice jams frequently lodge at Upper Becaguimec Island, they do not usually remain in place as long as in 1986. It is likely that the backwater created by the ice jam at Lower Becaguimec Island extended upstream, allowing the ice jam at UBI to stay in place longer than normal, thus requiring a greater-than-normal river flow to dislodge it, and greatly aggravating flooding in the area (NBELG 2022).

The provincial flood database (NBELG 2022) contains 20 occurrences of ice jamming or ice-related flooding at Hartland since 1900; with three during January (1995, 1996, 2006), two in February (1981, 1996), six in March (1902, 1933, 1968, 1976, 1998, 2001), and the remainder in April (1920, 1921, 1922, 1932, 1989, 1991, 1993, 1994, and 2009). The events exemplify the greater likelihood of midwinter and early spring events along the section of the SJR from Beechwood to Woodstock. Many ice-related events resulted in only minor flooding or inundation of low-lying land, but buildings along the town’s Main Street were affected during the 1920, 1968, 1976, 1994, and 2009 events. Several ice jams are reported to have occurred near or just downstream of the town limits, or more specifically at Lower Becaguimec Island, an indisputable location of ice-jam re-occurrence.

The provincial flood database (NBELG 2022) contains over 15 occurrences of SJR ice runs, ice jamming or ice-related flooding at Woodstock and nearby locations since 1887. Several of these events followed release of ice jams that formed upstream at Simonds and Hartland. The most common consequence of the ice-related events were road closures, although some events have caused damage to buildings and infrastructure at Woodstock and severe flooding of the Grafton area in 1976 caused subsequent closure or relocation of buildings. The Meduxnekeag River confluence with the SJR is at downtown Woodstock, and the confluence area appears to be a location of past ice jam occurrence in both rivers. Other reported locations of ice-jam formation include Newburg Junction (6 km upstream), Grafton bridge (2.1 km upstream), Island Park (near the mouth of the Meduxnekeag River), Lower Woodstock (3.5 km downstream), Bulls Creek (5.5 km downstream), and Sullivan Creek (approx. 28 km downstream). The presence of islands and channel constrictions (narrowing of river width) contribute to ice-jam formation at these locations.

Since the construction of Mactaquac dam and the creation of the upstream headpond (1968), any ice jamming downstream of Sullivan Creek has been largely remnants of ice covers/ ice jams from upstream. These accumulations of broken ice and the remaining ice sheet on the headpond usually melt in place.

5.4 Lower SJR ice jamming

Severe ice-related flood events historically occurred along the 148-km tidal stretch of SJR (downstream of Mactaquac Dam) at Fredericton, Maugerville, Sheffield, and other communities. A March 1936 ice jam at the provincial capital of Fredericton resulted in a water level of 8.9 m (approximately 0.3 m above the maximum open-water stage and 7.6 m above typical summer water levels) and caused inundation of three-quarters of the business district and several residential streets of Fredericton. The ice jam was lodged at the Fredericton railway bridge, which was destroyed during the event. Following release of the ice jam at Fredericton, another ice jam that formed downstream grounded against Ox and Gilbert islands at Sheffield and flooded over 240 km2 in the Maugerville area (LeBrun-Salonen 1983). However, damaging ice-related flood events no longer occur along the main stem of the lower SJR since completion of Mactaquac Dam in 1968. This dam does not alter the downstream flow hydrograph (run-of-the-river operation), but retains all of the incoming ice during the breakup event, which eventually melts in place. This feature could, at least in part, explain the absence of significant jamming in the downstream portion of the river. An additional contributing factor could be accelerated melt caused by open-water conditions that typically persist throughout the winter immediately downstream of, and for a significant distance from, a dam (Huokuna et al. 2022).

5.5 Deductions from past ice-jam floods

Based on the review of historic information, the following can be deduced.

  1. 1.

    Ice jams in the SJR often occur against a solid downstream ice cover, but commonly re-occur at specific channel constrictions, bends, bridge piers, and islands. Ice jams also occur where the slope decreases abruptly, as can happen naturally or by the construction of dams and associated headponds. The low headpond velocities also lead to thicker ice covers so jams above Grand Falls, and above Beechwood, occur frequently.

  2. 2.

    Ice jams occur because of increasing runoff, which in the SJRB often occurs due to rain-on-snow events or a sudden rise in air temperatures.

  3. 3.

    Spring ice jams along the SJR are often of short duration, generally lasting from a few hours to a few days (although ice accumulations in headponds can last longer as they melt in place).

  4. 4.

    The rate of backwater rise behind a forming ice jam can be several metres in a few hours. Caution: this rate can be greatly exceeded during the passage of waves that result from ice-jam releases, posing an additional hazard to people and animals that may happen to be near river banks (see details in Sect. 7.3).

  5. 5.

    Midwinter events have historically been rare events in the USJR but the number of midwinter events seems to be increasing since 1990. Midwinter events have been much more common in the MSJR, especially for the section of river below Beechwood.

  6. 6.

    Although historically long and devastating flood events have occurred in the Fredericton and Maugerville-Sheffield areas, ice-related flooding along the main stem of the LSJR has ceased to be a problem since the construction of Mactaquac dam.

Beltaos (2008) stated that certain morphological features are more conducive to jamming than others but jams are known to form anywhere, so long as there is competent ice cover to impede the downstream passage of ice blocks. Preferred jamming sites include sharp bends, islands, constrictions, shallows, reductions in river slope, bridge piers and abutments, thick and strong ice cover, and high-water levels during the preceding freeze-up season. Most of these features are not evident on the maps of Figs. 4 and 6 and we are not aware of any attempts to identify the causes of jamming in each of the depicted jams. This question could form the topic of a future research project.

6 Flood damage reduction

6.1 Flood mitigation strategies

Flooding and ice damage can be mitigated by structural and non-structural measures as identified in Burrell (1995). Ice-jam mitigation includes both structural and non-structural techniques that can be divided into three broad categories: ice-jam prevention, ice-jam breaching and removal, and flood-damage reduction. The effectiveness and cost of different techniques vary with the characteristics of river-ice regimes and ice-jam flooding (e.g., timing and duration), the physical features of the section of river (e.g. river dimensions, orientation, and obstructions), and the type of infrastructure (e.g., bridges, roads, buildings) to be protected. Non-structural approaches that have been effective in reducing flood damages and the risk to human life along the SJR has been the issuance of ice-jam flood advisories and warnings (sub Sect. 6.2) and flood delineation and education (Sect. 6.3).

6.2 Flow/flood forecasting

Flood and flow forecasting in the SJR Basin is part of the mandate of the provincial Department of Environment and Local Government (NBELG). Water level forecasts are done when extreme weather events make it necessary for public safety and during spring freshets, which typically runs from mid-March to mid-May. Flow forecasts are also done during other times of the year for NB Power to assist with decisions on maintenance scheduling. Field monitoring of ice conditions in the study is carried out annually by local observers, commissioned by NBELG and NB Power, with the primary objective of informing flow/flood forecasters of potential flood threats.

After flow, water level, precipitation, temperature, and snow data are collected and checked for quality, NBELG’s hydrologic (SSARR, Raven) and hydraulic (HEC-RAS) models can be updated historically. Each morning during the spring season, NBELG receives a weather briefing and forecast files from the Atlantic Storm Prediction Centre (ASPC). This information is input to the hydrologic models (SSARR and Raven), which are run to produce a flow forecast that is sent to NB Power and the provincial Emergency Measures Organization (EMO). The flow forecast is used by NB Power to determine a 5-day discharge schedule at the Mactaquac Dam. Once NB Power returns a 5 day discharge schedule at Mactaquac to NBELG, a hydrodynamic HEC-RAS 1D model is run to obtain water-level forecasts for the lower SJR, and these water-level forecasts are posted online and distributed to government agencies.

The flow forecast is also used by NB Power to assess the risk for potential ice movement. A potential for ice movement along the SJR is considered to exist if the flow exceeds 20,000 cfs (approx. 565 m3/s), 30,000 cfs (850 m3/s) and 40,000 cfs (approx. 1130 m3/s) at Fort Kent, Grand Falls, and Beechwood, respectively. This river-ice assessment is supplemented with ice observations made by NB Power staff and NBELG contracted river observers. When deemed necessary, NB Power will have observers monitor jams around the clock.

Flow forecasts and field observations can also be used to predict severity of mid-winter ice jam flooding. Recently-published research from the University of Saskatchewan on ice-jam flood forecasting uses a combination of hydrological (MESH) and hydraulic model (RIVICE) simulations to develop a stochastic framework that can be used to forecast mid-winter breakup severity in an operational context (Das et al. 2022). This can help to make appropriate management decisions and emergency measures before the breakup. The stochastic outcomes allow the assessment of hundreds of possible scenarios that could be used in decision-making during emergency situations. The percentile results from these hundreds of forecasts could be used to determine the level of ice-jam flood risk for the study area (Das et al. 2022). A stochastic framework, a combination of hydrological and hydraulic models and multiple data sources, was also proposed to forecast the severity of mid-winter ice-jam flooding along the transborder Saint John River (Das et al. 2023).

Other notable efforts to model water levels generated by ice-jams, given the anticipated discharge magnitude, include a calibrated HEC-RAS ice-jam model along the upper and middle SJR, using measurements of ice jams obtained during the breakups of 1993 to 1997 and 2000 (Beltaos et al. 2012; Beltaos and Burrell 2015).

6.3 Flood mapping

With New Brunswick’s long history of flooding, it is important to have the right tools and information to mitigate flood damage. Flood mapping provides decision-makers and the general public with information by which flood damages can be avoided or mitigated. Different types of flood mapping include:

  • Inundation mapping of a past or occurring flood,

  • Flood-hazard maps depicting the potential extent of flooding, usually based on hydraulic and (or) hydrodynamic investigations,

  • Flood-risk maps that associate flood hazard with the potential negative social, economic, and environmental consequences of flooding, and

  • Public awareness (public information) maps that depicts a flood-hazard area with a narrative on the potential for future flooding and the associated risks.

Flood maps are important because they can be used by communities and individuals to better understand and plan for avoidance of flood hazard, and by governments and communities to make decisions about the location and design of infrastructure and emergency planning. Flood-hazard maps that identify areas prone to flooding can be used for engineering purposes and as regulatory maps for controlling land-use in flood-hazard areas.

New Brunswick first created flood maps in the 1980s under the Canada-New Brunswick Flood Damage Reduction Program (Burrell and Keefe 1989). These flood-hazard maps (which were termed flood-risk maps at that time) depicted hypothetical 20-year and 100-year open-water flood-hazard areas based on hydrotechnical studies, or in the case of rural areas, a flood-hazard area based on topographic delineation of an envelope curve of past flood levels, including ice-related events (Burrell et al. 1989). In addition, flood information maps were prepared that depicted the inundation extents of the 1976 and 1987 ice-jam floods at Perth-Andover, as well as the 1973 flood along the LSJR.

The information and technology used to create the earlier flood-hazard maps are now outdated and the maps do not include the future impacts of climate change. In January 2022, the New Brunswick Department of Environment and Local Government (NBELG), with support from the Government of New Brunswick and Public Safety Canada’s National Disaster Mitigation Program, released updated flood-hazard mapping for New Brunswick. These maps cover New Brunswick’s many rivers that are prone to flooding, and also incorporate projections regarding the impacts of climate change (https://flooding-inondations-geonb.hub.arcgis.com/). In general, these maps pertain to open-water floods and work remains to produce flood-hazard maps for communities along the SJR prone to ice-related floods.

Different types of models and approaches (e.g., empirical, statistical, and process-based) can be used to delineate an ice-jam flood hazard. These models and approaches rely on information about the river channel (shape, depth, slope, hydraulic roughness of riverbed) and flood plains, and information about ice-jam properties such as porosity, thickness, ice volume, and hydraulic roughness of jam underside. Flood-hazard delineation and mapping of areas affected by ice jams has been customarily performed by engineering consultants for specific vulnerable sites, under contracts with municipal, provincial, territorial, and federal authorities. Respective engineering reports are unpublished and not readily available. Gerard (1989) reviewed early flood-hazard delineation methodology. More recently, this topic was discussed in several journal papers (Kovachis et al. 2017; Lindenschmidt et al. 2018; Das et al. 2020, Lindenschmidt et al. 2023, Rokaya et al. 2022).

6.4 Relocation

Sometimes avoidance of a flood hazard requires building relocation or structural alteration. Following serious ice-jam flooding (1976, 1987, 1993, and 2012) in the community of the Village of Perth-Andover, a redevelopment plan was implemented to make the community more resilient to major flooding. Several downtown businesses were relocated to higher elevations and many flood-damaged/flood-susceptible homes were either demolished, relocated, or flood-proofed.

7 Hydroclimatic trends and a changing climate

Winter is an important period for the functioning of forested ecosystems of the SJR Basin. As average winter temperatures increase due to anthropogenic climate change, less snowfall (due to more precipitation falling as rain), and less snow accumulation is anticipated. Climate change is expected to increase open-water flood flows in New Brunswick (Turkkan et al. 2011) with the increase in low-return-period floods about 30% and for higher return-period floods about 15% (Turkkan et al. 2011). Prediction of the effects of climate change on ice-related flooding is less clear, as the timing of both rainfall events and snowmelt may change.

Average annual temperature in the State of Maine has increased 1.8 °C in the last 124 years; the rate of warming has increased most notably since 1960, with the six warmest years on record having occurred since 1998 (Fernandez et al. 2020a, b). Throughout the State of Maine, warming winter temperatures mean that more precipitation falls as rain instead of snow. Statewide average annual snowfall is estimated to have decreased by about 17 percent over the past century, however, since 1960, the downward trend has been only 3% due to partially attributed to increased moisture availability associated with warmer air (Fernandez et al. 2020a, b). However, portions of northern Maine have diverged from the trend. Since the mid-1990s, there has been considerable variability, for example, low snowfall occurred during the winters of 2004 and 2010 and high snowfall occurred during the winters of 2008 and 2019 (Fernandez et al. 2020a, b).

For New Brunswick, climatic data have been obtained from archives of four Edmundston climate stations as well as stations near Aroostook and Woodstock. Data since 1950 generally indicate slight, non-significant, cooling or warming trends for the winter and spring months (Table 5). The only statistically significant trend was found for November at the Woodstock station. These results do not presage climate-model projections of considerable mid- and late-century winter warming over the SJR basin (Budhathoki et al. 2022). If the projections are accurate, a perceptible warming trend can be expected to commence in the next 10 to 20 years.

Table 5 Statistics of mean air temperature for winter and spring months, 1950–2022

According to climate-model projections, the average temperature in the SJRB during the winter months may increase by 4 and 6.2 °C in the 2041–2070 and 2071–2100 periods, respectively (Budhathoki et al. 2022). Considering increased rainfall as well, one can hypothesize that (a) the SJR will be normally ice free during most of April by the year 2070, (b) ice breakup by 2070 would most likely occur during March in the USJR, and in February or March along the MSJR, and (c) that midwinter breakups will be common along sections of the upper and middle SJR by the end of the century, becoming the norm along the MSJR by the end of the century. It is indeed conceivable, given current climatic projections, that competent ice covers may not form most years or be of short duration along sections of the SJR by the end of the century.

Shifts in the seasonal and annual flow regimes are also sensitive to interannual temperature variability and warming trends. The Tropical/Northern Hemisphere (TNH) pattern, which affects atmospheric air currents, was found to have a statistically significant correlation with the local surface temperature in the northeastern United States, thereby affecting the pattern and the frequency of warm episodic events (Kim et al. 2021). The change of the TNH pattern in response to the warm phase of the El Niño-Southern Oscillation (ENSO, a recurring climate pattern involving changes in central and eastern tropical Pacific Ocean water temperature) is one factor thought to contribute to wintertime hydroclimatic variability in the SJRB, with unusually warm episodes occurring during the winter and spring inducing changes in the flow regime (Kim et al. 2021).

8 River-ice trends and a changing climate

8.1 Trends in ice phenology

Trends in ice phenology and breakup flooding potential are assessed using WSC’s hydrometric station records for the Saint John River at Fort Kent (station 01AD002, drainage basin area = 14 700 km2). WSC’s B flag in the hydrometric records signifies that the recorded water level was influenced by ice (but not necessarily the presence of an ice cover at the gauge site). Though flow data begin in 1926, B flags were not assigned until late in 1951; they continued through the spring of 1952; no flags for the 1952–1953 ice season, then flags start again in late 1953 and continue to the present time. The first day with a B flag in the fall and early winter of each year indicates that ice is affecting the water levels experienced at the gauge site. In the fall and early winter, sequences of a few or several B-days can occur, until the B-sequence becomes uninterrupted until the following spring. Hereinafter, the latter is termed the period of stable ice cover. The dates of initial ice effect and stable ice-cover formation often coincide, but can also be separated by several, or even many, days. Both the first day of B and the first day of stable ice cover along the SJR tend to occur earlier in the year (Fig. 8). It is difficult to explain the trend towards earlier freeze-up. Earlier ice-cover formation is promoted by lower discharges, but this is not the case for December, the month when the river typically freezes up. Mean December temperatures at nearby Edmundston suggest a slight, albeit insignificant (at the 5% level) warming rather than cooling trend since 1949 (see Sect. 7). [Note throughout this article, statistical significance refers to the 5% level]. The last day each year of ice effect at the hydrometric station (the last day with a B flag) shows a very slight, statistically insignificant decrease. These findings suggest the timing of the spring breakup has not essentially changed since the 1950s.

Fig. 8
figure 8

Freeze-up initiation dates. Saint John River at Fort Kent (WSC station 01AD002). Statistics for first day of stable ice cover: PMK = 0.008; Sen’s slope = − 0.200 days/year (nonparametric estimate of trend slope (Sen 1968), not to be confused with the parametric slope indicated by linear regression)

A more detailed method (Beltaos 2002) considers the onset of spring breakup, based on exceedance of a critical discharge (570 m3/s for station 01AD002) estimated using typical ice thickness and degree of pre-breakup thermal deterioration. This approach reveals a slight, but significant, trend towards earlier spring breakups (Fig. 9), as has also been noted by others (Arisz et al. 2011; Lacroix et al. 2005). In the majority of temperate ice-affected rivers, climate change has contributed to a decrease in ice-cover duration due primarily to earlier breakup rather than later formation (Fuks 2023).

Fig. 9
figure 9

Flow-based timing of spring breakup: first spring day to exceed 570 m3/s, 1952–2019 inclusive, Saint John River at Fort Kent (station 01AD002). Statistics: PMK = 0.023; Sen’s slope = − 0.119 days/year

8.2 Trends in mid-winter breakup and ice-jam flooding potential

As noted earlier, mid-winter breakup (MWB) may result from substantial runoff that is typically driven by rain-on-snow events occurring after establishment of a stable ice cover. MWBs are becoming increasingly common events in cold-region rivers (de Coste et al. 2022; Fuks 2023).

Brief mid-winter thaws along the SJR are common, but not all of them generate sufficient flow to break up the ice cover. Beltaos (2002) estimated that the ice cover could be mobilized in the vicinity of Fort Kent if the flow (at hydrometric station 01AD002) exceeds 510 or 640 m3/s, for ice-cover thickness values of 0.28 or 0.44 m, which represent low and average thickness for the end of January, respectively. The peak winter flow (Fig. 10) has exceeded the 510 m3/s threshold seven times (up to 2019): twice in 1996 and once in each of 1979, 1981, 1995, 2004 and 2006. Beltaos (2002) indicated that a MWB did not occur in 1979 but did occur in 1981, 1995 and 1996 (twice). Whether the relatively high flows in 2004 and 2006 resulted in breakup near Fort Kent is unknown; breakup did occur in the middle SJR in January 2006 and resulted in at least one ice jam (near Hartland). Flows greater than the threshold also occur on occasion in December, very likely breaking up any ice cover that formed earlier, but are unlikely to cause any significant jamming due to the thin and incomplete ice cover.

Fig. 10
figure 10

Maximum mean daily discharge during the winter period, herein defined as January 1 to March 15 of each year. Saint John River at Fort Kent (station 01AD002), 1952–2019. Statistics: PMK = 0.007; Sen’s slope = 2.3 m3/s per year

8.3 Trends in spring breakup and ice-jam flood potential

As already discussed, ice jams can form at various locations along the study stretch during the spring breakup of the ice cover. They can be a few hundred metres or many kilometres long. Because of their large aggregate thickness and underside roughness, ice jams raise water levels, by several, or many in extreme cases, metres above what would occur under open-water or sheet-ice cover conditions in the same river segment. Specifically, for the SJR, prediction of ice-jam locations and water levels has been successfully carried out using different river-ice models (e.g., Beltaos et al. 1996, Tang and Beltaos 2008, Beltaos et al. 2012, Beltaos and Burrell 2015, Knack and Shen 2018, Kowshal et al. 2019, Jamil 2020).

In addition to channel bathymetry and slope, ice-jam water levels depend on the magnitude of the flow in the river. Consequently, the maximum daily mean spring breakup flow is an indicator of ice-jam flooding potential. Hydrometric data for the Saint John River at Fort Kent (station 01AD002) show large, statistically significant increase (PMK = 3 × 10–5; Sen’s slope = 12.9 m3/s per year; overall series average value ~ 1090 m3/s) in the estimated maximum daily mean spring breakup flow and in mean April flow (Burrell and Beltaos 2022). The spring breakup flow was defined herein as the highest daily mean flow occurring during the last 15 days with a B flag. Besides April, significant increasing trends were found for January, February and March, consistent with the projections of Budhathoki et al. (2022). For April, Budhathoki et al. (2022) project a slight increase in mid-century flow at Grand Falls and a decrease in mid-century flow at Mactaquac, and late-century decreases in flow at both Grand Falls and Mactaquac.

Examination of monthly climatic data at Edmundston and Allagash did not reveal a relationship between total winter (Nov–Mar) snowfall and ice-jam flood occurrences in the SJR. Instead, floods seem to result from episodic large rainfall in late March and/or April and (or) snowmelt from sudden extreme rise of air temperature. The latter occurrence, which brought daytime temperatures up to the mid-twenties, was responsible for the devastating flood of 2012 at Perth-Andover, despite no significant accompanying rainfall. It appears that the end-of-winter snow cover over the basin is, to date, not a limiting factor for generation of large breakup flows if melt is sufficiently rapid.

Measurements and numerical modelling have shown that javes can greatly augment the water surface slope, flow velocity, and tractive force that is exerted on stationary ice covers. The enhanced tractive force is often sufficient to dislodge the winter ice cover, initiating breakup over long distances, and on occasion causing the release of downstream jams. Because the water level can rise sharply during the passage of a jave, observers—and the public in general—need to be vigilant when walking along the river downstream of an ice jam.

9 Research needs

River-specific studies are needed to better document and understand river-ice processes, and their effects on geomorphology, sediment transport, water quality, and ecology. Additional field data and information about river-ice processes would be useful in the calibration and verification of available numerical models to simulate river-ice processes. This includes identification of factors that contribute to the formation of ice jams in different river reaches. Rokaya et al. (2018) assessed the current state-of-the-art of research concerning ice-jam flooding highlighting the progress, gaps, and opportunities in this research area.

Winter-spring hydrology is often a limiting factor in defining ice conditions under existing and future climates (Turcotte et al. 2019). Research with respect to discharge measurement during the ice season remains important, and discharge quantification during freeze-up and breakup when ice conditions are unstable (and measurement potentially unsafe) is especially needed.

Research about the effects of climatic change on the future frequency and magnitude of ice-induced floods is also needed (Turcotte et al. 2019). This research could include interpretation and extrapolation of historical ice-affected water-level trends for a projected future climate, and numerical modelling of hypothetical future ice-jam floods using inputted winter hydrographs and ice parameters based on a projected future climate.

10 Concluding remarks

An overview of past ice-related flooding along the SJR, New Brunswick, has been presented along with a discussion of flood-damage reduction measures and a discussion of how changes in the hydroclimate and ice regime can affect future ice-related flooding.

The changing climate will alter the ice regime of the SJR, with ice covers likely becoming less stable, more intermittent, and of shorter duration by the end of the century. This corresponds to trends and projections on other temperate river basins. The frequency and severity of ice jamming will also change as the ice regime changes. Ice jamming will possibly become more frequent and severe during the next few decades due to increased flows during the breakup period and the possible impediments caused to ice passage by refrozen mid-winter jams. However, by the end of the century, reduced ice-cover thickness and extent will render ice-jamming, if it occurs, less severe.