Earthquakes are not randomly distributed. Most occur in well defined zones along the boundaries of Earth’s tec-tonic plates. In Canada, earthquakes are most common in coastal British Columbia near the western edge of the North America plate (Rogers and Horner 1991), but they also occur in many other parts of the country, notably southwest and eastern Yukon Territory, southern Ontario and Québec, and some of the Arctic Islands (Fig. 2).

Plate-boundary earthquakes occur along convergent, divergent, and transform plate boundaries. In Canada, plate-boundary earthquakes are restricted to the Pacific coast and are of two types. Subduction earthquakes occur along the thrust fault separating the North America and Juan de Fuca plates (Rogers 1988; Hyndman 1995; Hyndman et al. 1996). Subduction earthquakes occur when the strain energy that has accumulated in rocks adjacent to the convergent plate-boundary fault is suddenly released (Dragert et al. 1994; Hyndman and Wang 1995). The last great subduction earthquake on the Pacific coast, in January 1700, left geological traces in tidal marshes in northern California, Oregon, Washington, and British Columbia (Atwater et al. 1995, 2005; Atwater and Hemphill-Haley 1997; Clague 1996, 1997, 2002; Clague and Bobrowsky 1994a, b, 1999; Nelson et al. 1995, 2006). Its inferred extent, together with written records of the damage caused by its tsunami in Japan (Satake et al. 1996; Atwater et al. 2005), indicate that the quake likely had a magnitude larger than M 9.0.

The second type of plate-boundary earthquake on the west coast involves strike-slip or oblique-slip displacements along the Queen Charlotte Fault, which separates the Pacific and North America plates and extends from the north end of Vancouver Island to the southeast Alaska archipelago. The Queen Charlotte Fault is contiguous to the north with the Fair-weather Fault, and the latter extends along the coast of southeast Alaska to a block of complexly deforming crust (Yakutat Block) located east of the Aleutian Trench. Canada’s largest recorded earthquake, an M 8.1 event, ruptured a portion of the Queen Charlotte Fault in 1949 (Milne et al. 1978).

Intraplate earthquakes occur on faults within, rather than between, plates. Although commonly smaller than subduction earthquakes, some intraplate earthquakes have moment magnitudes in the M 7.5−8.0 range, e.g. the May 2008 quake in southwest China. Moderate and large intraplate earthquakes in densely populated areas are extremely destructive; they can claim tens of thousands of lives and may cause damage assessed in the tens or hundreds of billions of dollars. Because large (M >7) intraplate quakes are infrequent, most people are generally poorly prepared for them and older buildings may not be able to withstand the associated strong shaking. Canada’s largest intraplate earthquake is the M 7.3 Vancouver Island quake in 1946, which was felt as far away as Seattle, Washington, and Calgary, Alberta (Rogers and Hasegawa 1978). Fortunately, the epicentral area in 1946 was sparsely populated, hence damage was minor compared to what a similarly sized earthquake in the same area would cause today. A shallow, M 7 crustal earthquake close to Vancouver or Victoria would now cause catastrophic damage (Munich Reinsurance Company of Canada 1992).

Moderate-size intraplate earthquakes occur frequently in southern Ontario and Québec (Adams and Basham 1989, 2001). Most of these quakes are associated with the ancient rifted edge of the North American continent, which lies along the St. Lawrence River valley from Montreal to Sept Sles, Québec; related rift structures also occur along the Ottawa and Saguenay river valleys. Strain accumulates on these structures due to the westward movement of the North American Plate away from its eastern margin at the Mid-Atlantic Ridge. Earthquakes along the rift involve thrust movements at depths of 5 to 30 km and occur in three clusters (Fig. 2): western Québec, which experienced moderate earthquakes in 1732 (M 5.8), 1935 (M 6.2), and 1944 (M 5.8); Charlevoix, northeast of Québec City (1663, 1791, 1860, 1870 and 1925; M 6.0−7.0); and the lower St. Lawrence region near Baie-Comeau, Québec, comprising a diffuse cluster of mostly small earthquakes (Lamontagne 2002). The most recent and widely felt earthquake in central Canada in recent time occurred in the Saguenay region of Québec in 1988 (M 5.9).

Historical large earthquakes in Canada have caused little damage by California standards. The M 8.1 Queen Charlotte Islands earthquake in 1949 was larger than the 1906 San Francisco earthquake, yet caused little damage and no loss of life because no more than a few thousand people lived in the area of strong ground shaking at the time of the quake. Similarly, the epicentre of the 1946 Vancouver Island earthquake (M 7.3) was far enough from towns and other infrastructure that the damage was limited. Today, the same area supports a much larger population and far more economic wealth than in 1946, so if the same event were to occur now, financial damage would likely be in the hundreds of millions of dollars, if not more.

The intensity of seismic shaking is commonly expressed as the ratio of ground acceleration to the acceleration of gravity (Bolt 2003). Strong ground shaking can be especially damaging to a building if the horizontal component of motion is strong or if the frequency of the shaking matches the natural vibration frequency of the building, a phenomenon called ‘resonance’. In general, high-frequency, short-period seismic waves are most damaging to low structures, whereas low-frequency, long-period waves tend to damage tall buildings or long structures such as bridges.

The characteristics of local earth materials strongly influence the frequency distribution of wave forms and ground-motion attenuation during an earthquake (Bolt 2003). The dense granitic and metamorphic rocks of the Canadian Shield can transmit earthquake energy very efficiently, and even moderate earthquakes can therefore cause damage over large areas. In contrast, seismic energy and associated ground shaking attenuates rapidly away from the epicentres of most earthquakes in the extremely heterogeneous, folded, and faulted crust of western North America. Seismic waves also move much more slowly through loose sediments than through bedrock, especially if the sediments have high water content. For example, seismic waves typically slow down as they move from bedrock to alluvial sands and gravels. As P (primary) and S (secondary or shear) waves decelerate, some of their forward-directed energy is transferred to surface waves. This effect, known as seismic amplification, increases the amount of ground motion experienced in an earthquake.

Intense seismic shaking can elevate intergranular pore-water pressure in loose, water-saturated sediments and cause them to liquefy. Liquefaction commonly takes place at shallow depths and is most common in loose sandy and silty sediments (Jefferies and Been 2006). Liquefaction may be accompanied by lateral shifts of more cohesive earth materials on one or more layers of fluidized sediment, causing much damage to roads and buildings at the surface and to gas, water, and sewer lines buried at shallow depth. Watery sand and silt may also flow, under pressure, upward along fissures in overlying solid materials and onto the surface to form ‘sand blows’ or ‘sand volcanoes’. Liquefaction causes the land surface to settle irregularly, which, in turn, may damage foundations of buildings, water and sewer lines, and other structures.

Some earthquakes raise or lower the land over large areas. The Cascadia subduction earthquake in 1700 caused parts of the Pacific coasts of Vancouver Island, Washington, and Oregon to subside up to 2 m, inundating coastal marshes and forests with sea water (Peterson et al. 1997). A large crustal earthquake near Seattle about 1000 years ago produced co-seismic uplift of up to 7 m (Bucknam et al. 1992).

Large earthquakes may also trigger hundreds or thousands of landslides. The 1970 Ancash earthquake in Peru (M 7.9) triggered a huge landslide that destroyed the towns of Yungay and Ranrahirca, killing thousands of people (Plafker and Ericksen 1978). A large (M 7+) earthquake in southern Québec in 1663 triggered ‘quick clay’ failures in glacial marine sediments in the St. Lawrence River valley (Filion et al. 1991), and the 1946 Vancouver Island earthquake triggered hundreds of mainly small landslides in the mountains of central Vancouver Island (Mathews 1979).

Ground shaking and surface rupture can sever power, telecommunication, and gas lines, starting fires. Such fires may be difficult to suppress because fire-fighting equipment may be damaged; streets, roads, and bridges blocked; and water mains broken. Appliances such as gas water heaters may topple when shaken, producing gas leaks that can ignite. The San Francisco earthquake of 1906 has often been referred to as the ‘San Francisco Fire,’ because most of the damage and loss of life were caused by a firestorm that ravaged the city for several days following the quake. Such a firestorm is unlikely to happen in a Canadian city today, but localized fires can be expected in the event of a large (M 7+) event close to an urban centre.

The west coast of Canada is located along the Pacific ‘Ring of Fire’ and is vulnerable to tsunami generated by subduction earthquakes that occur beneath the Pacific Ocean (Murty 1977; Clague et al. 2000, 2003). The largest tsunami results from great earthquakes that rupture the Cascadia plate boundary where the oceanic Juan de Fuca Plate moves beneath North America (Fig. 3). The British Columbia coast is also affected by tsunami of more distant Pacific earthquakes. In fact, the largest tsunami to strike British Columbia in historical time was generated by the M 9.2 Alaska earthquake in March 1964 (Plafker 1969; Lander 1996). A series of waves radiated outward from the epicentre of the earthquake off south-central Alaska and, within a few hours, reached the outer coast of British Columbia. The tsunami caused about $10 million damage (1964 dollars), mainly to the Vancouver Island communities of Port Alberni, Hot Springs Cove, and Zeballos (Wigen and White 1964; Murty and Boilard 1970; Murty 1977; Thomson 1981). Damage was greatest at Port Alberni, located at the head of Alberni Inlet, where 260 homes in the town were damaged, 60 of them extensively (Thomson 1981).

Tsunami of more local extent, triggered by landslides and shallow crustal earthquakes, also pose a hazard to Canada’s west coast, especially in steep-walled fiords and inlets that indent the coast (Clague 2001). A submarine landslide in April 1975 at the head of an inlet on the northern British Columbia coast produced waves up to 9 m high that damaged wharfs and other shoreline infrastructure at Kitimat (Murty 1979; Prior et al. 1982, 1984; Johns et al. 1986; Skvortsov and Bornhold 2007). More than 10 million m3 of fiord-wall and deltaic sediments failed and flowed into the inlet, mobilizing an additional 25 million m³ of fiord-bottom sediments. The subaqueous debris flow travelled over 5 km from the delta front on slopes inclined less than 0.4° to a water depth of 210 m (Prior et al. 1984). Similar unstable sediments are present in most fiords on the British Columbia coast; local, landslide-triggered tsunamis are therefore also possible in many areas other than Kitimat Inlet.

Tsunami can also be triggered by rockfalls and rockslides from the walls of British Columbia fiords. Born-hold et al. (2007) proposed that the First Nations village of Kwalate in Knight Inlet was destroyed by a local tsunami triggered by the catastrophic failure of a nearby steep rock wall about 500 years ago.

Fortunately, most of British Columbia’s coastal residents live along the Strait of Georgia, an area of low tsunami risk (Fig. 4). The total population in communities at higher risk, on western Vancouver Island (e.g. Tofino, Ucluelet, Port Alberni) and at fiord heads (e.g. Kitimat, Bella Coola), is less than 10 000, and small in comparison to the population of tsunami-prone areas on the Pacific coasts of Oregon and Washington.

A large submarine landslide can also generate a tsunami, as happened at the edge of the Canadian Atlantic continental shelf on November 18, 1929. On that day, an M 7.2 earthquake occurred about 20 km beneath the seafloor at the southern edge of the Grand Banks, some 250 km south of Newfoundland (Heezen and Ewing 1952; Piper et al. 1988). Earthquake shaking triggered a huge submarine slump, which, in turn, set off a tsunami that propagated across the Atlantic Ocean (Fine et al. 2005). The tsunami damaged more than 40 coastal communities on Burin Peninsula and claimed 27 lives in Newfoundland and one in Nova Scotia (Whelan 1994). The tsunami arrived near the peak of a high tide, thus increasing the damage.

Landslides may also generate destructive displacement waves and seiches in lakes and rivers bordered by steep, unstable slopes and in lakes containing actively growing deltas. Such events can cause much damage and loss of life. For example, in 1880, a large mass of late Pleistocene glacial marine sediments slid into the Fraser River at Haney, east of Vancouver, producing a displacement wave that travelled several tens of kilometres up and down the river, causing considerable damage to boats, wharves, and buildings at the shore (Evans 2001). And in August 1905, a landslide in Pleistocene glacial lacustrine sediments crossed the Thompson River near Spences Bridge in southwestern British Columbia, producing a 5-m-high displacement wave that rushed more than 1.5 km upstream, destroying twenty buildings and drowning 15 people (Evans 2001). Three years later, in 1908, a landslide on the Liève River in western Québec generated a displacement wave that inundated part of the village of Notre-Dame-de-la Salette, killing 27 people (Evans 2001).

Tsunami height and run-up depend on the distance from the source, possible resonance amplification in inlets and bays, tidal height at the time of the tsunami, shoreline configuration, and the gradient of the seafloor directly off the coast. Numerical modelling by Chasse et al. (1993) has shown that the maximum amplitude of a tsunami in the St. Lawrence Estuary would be about 1.5 m. Such a tsunami, however, could reach up to 4.5 m above sea level (asl) at the highest tide, or only 0.9 m asl at the lowest tide. Also, the tsunami would travel much faster along the north side of the estuary, due to the greater water depth, than along the south side. Similar dependence of speed on propagation direction was noted during the 1929 Grand Banks tsunami − it took much longer for the tsunami to reach Nova Scotia than Newfoundland, even though Nova Scotia was closer to the tsunami source (Berninghausen 1968).

Sand sheets deposited by prehistoric tsunami have been found in mud and peat deposits in protected tidal marshes at many sites on western Vancouver Island, Washington, and Oregon (Clague and Bobrowsky 1994a, b, 1999; Clague 1997, 2002; Peters et al. 2007). The sand sheets become thinner and finer in a landward direction, and commonly contain marine microfossils, indicating that they were deposited by strong landward surges of water. Some of the sand sheets directly overlie buried peaty or forest soils that subsided during great earthquakes at the Cascadia subduction zone (Atwater et al. 1995; Clague and Bobrowsky 1994a, b, 1999; Clague 1997, 2002). Analogues for these prehistoric subduction earthquakes include the great earthquakes in Chile in 1960, Alaska in 1964, and Indonesia in 2004, each of which produced widespread crustal subsidence in their source areas and destructive fartravelled tsunami.

Tsunami deposits also have been found in nearshore coastal lakes in Oregon and on western Vancouver Island (Hutchinson et al. 1997, 2000; Clague et al. 1999; Kelsey et al. 2005). Tsunamis enter the lakes by surging up outlet streams or by crossing sand dunes that lie between some of the lakes and the sea. The most complete record found to date is from Bradley Lake in Oregon, where there are 14 tsunami layers in a sequence spanning the last 7500 years (Fig. 5; Kelsey et al. 2005).

So far, evidence for a prehistoric tsunami has been found only on Canada’s west coast, but similar evidence likely exists in Atlantic Canada. A widespread layer of 8000-year-old sand, which has been found in coastal areas of eastern Scotland, western Norway, and Iceland, was deposited by a large tsunami triggered by a submarine landslide in the North Atlantic Ocean (Dawson et al. 1988; Bondevik et al. 1997). On the basis of the known extent and character of the landslide deposit, the tsunami may have been large enough to impact and leave a trace along the Atlantic coast of Canada.

Landslides occur in all provinces and territories in Canada, as well as on the seafloor off the Pacific, Atlantic, and Arctic coasts, but their incidence is highest in the mountainous parts of the country – the Cordillera of British Columbia, Yukon, and Alberta, and the Appalachian provinces of Québec and New Brunswick (Mollard 1977; Evans 2001). Areas at risk from landslides in other parts of the country are localized, notably in parts of the St. Lawrence Valley underlain by Leda Clay, and along glacial meltwater valleys eroded into shales and clays on the Prairies. Leda Clay was deposited in an arm of the sea that occupied the St. Lawrence Valley at the end of the last glaciation. Its constituent clay minerals are weakly bound to one another; an apt metaphor for the sediment is a ‘house of cards’. When disturbed, grains may collapse and become suspended by water in the pores of the sediment, leading to liquefaction and landslides.

Some of the most damaging landslides in Canada are debris flows – watery slurries of silt, sand, gravel, and woody debris. They range in volume from several hundred cubic metres to about 100 000 m³, and therefore are small in comparison to many other types of landslides. Most debris flows are funnelled along steep stream channels and ravines and are triggered by intense rainstorms or rain-on-snow events (VanDine 1985; Jakob and Weatherly 2003). They are common in the mountains of western British Columbia, but are by no means restricted to that area.

Events following establishment of the community of Lions Bay, located along Howe Sound between Vancouver and Squamish, illustrate the problems that debris flows can cause. Development at Lions Bay began in 1957 during construction of Highway 99. Most of the community is located on a late Pleistocene fan and on adjacent rock slopes, but about 63 houses are situated on the coalesced Holocene fans of Harvey and Alberta creeks. On February 11, 1983, a disastrous debris flow swept down Alberta Creek into Lions Bay (Fig. 10; Thurber Consultants Limited 1983; Jackson et al. 1985). All of the bridges and culverts along Alberta Creek, with the exception of the British Columbia Railway bridge, were destroyed. About 10 000 m³ of debris were deposited below the highway; in addition, some sediment flowed into Howe Sound. Within the community of Lions Bay, destructive tongues of logs and boulders were forced out of the channel at corners and obstructions, one of which killed two people in a trailer.

Measures taken to protect people and property in Lions Bay and elsewhere along Howe Sound from future debris flows include the construction of debris catch basins and new highway bridges, lining of stream channels, and channel clearing. The catch basins trap debris before it reaches the highway, rail line, and homes. The other protective works allow debris flows to travel unimpeded along channels through communities. The total cost of these works has been estimated at $12–17 million (1985 dollars; Jackson et al. 1985).

Slab avalanches, which are the most dangerous type, require a buried weak layer and an overlying stronger mass of snow. Weather determines the evolution of the snowpack and the formation of weak, unstable layers within the snow. The most important contributing factors are heating by solar radiation, radiative cooling, temperature gradients in the snowpack, snowfall amounts and type, and wind (McClung and Schaerer 1993; Jamieson 2001). Weak layers in the snowpack can form in three main ways (Jamieson 2006). First, wind can contribute to the rapid build-up of snow (wind slabs) on sheltered lee slopes. If winds become stronger during a snowstorm, a dense layer of broken and packed snow crystals (hoar) may accumulate on a delicate layer of unbroken crystals that had fallen earlier in the storm. In this manner, the layer of unbroken crystals becomes a buried weak surface along which failure may subsequently occur. Second, surface hoar, also called hoar frost, may form on cold clear nights by sublimation. Ice crystals produced in this manner form a weak layer that changes only slowly once buried. In contrast, the overlying and underlying snow layers commonly gain strength over a period of days to weeks, leaving the buried surface hoar as a potential failure plane. Third, weak layers of hoar can develop inside the snowpack.

During the first few seconds of a slab avalanche, the failed snow mass comprises large fractured blocks of snow sliding downslope. The slab rapidly gathers speed and, commonly within a few tens of metres, disinte-grates into smaller fragments and, in many cases, individual grains of snow. At velocities above about 35 km/hr, dry avalanches commonly generate a cloud of powdered snow that billows above the main mass of flowing snow. Wet avalanches have higher densities than dry ones (300–400 kg/m³ vs. 50–150 kg/m³), but do not attain the extreme velocities of some large dry avalanches (Jamieson 2001). Large snow avalanches may have sufficient momentum to climb tens of metres or more up opposing slopes and destroy the forest cover. They may also displace air, producing a damaging air blast that arrives seconds before the avalanche.

Most large snow avalanches are released from slopes with gradients of 30−45°, although some occur on slopes as steep as 60° (Abromelt et al. 2004). Slopes in the 30–45° range are also popular with skiers, snowboarders, and snowmobilers, which is why recreationists often trigger avalanches. Fewer than 5% of dry snow avalanches begin on slopes of less than 30°, but wet snow slides can occur on slopes under 25° (Abromelt et al. 2004). Higher elevations commonly have a higher avalanche hazard than lower ones because they receive more snow, have fewer trees to impede avalanches, and commonly are windier. Gullies or ravines funnel snow avalanches, increasing their destructive force and making escape difficult.

Over 600 people have died in snow avalanches in Canada since the earliest reported accidents in the middle nineteenth century (Stethem et al. 2003; Abromelt et al. 2004). Until the early twentieth century, most avalanche fatalities were people engaged in construction of railways and roads, or working at mine sites (Fig. 11). The Canadian Pacific Railway line, completed in 1886, was the first rail line built across Canada. During construction of the rail line and in the early years of its operation, ‘white death’ snow avalanches claimed approximately 250 lives in Rogers Pass in Glacier National Park, British Columbia (Woods 1983). The number of industrial and transportation fatalities from snow avalanches, however, is now low, mostly because pre-emptive measures to reduce the risk are routinely taken to protect road and rail lines. In contrast, the number of recreational accidents has increased dramatically since the 1930s. Indeed, an average of 12 people have died, annually, in snow avalanches in Canada over the past decade, and nearly all of them were recreationists who were buried in an avalanche that either they or a member of their group triggered.

Snow avalanches also cause traffic delays, with significant economic losses (Morrall and Abdelwahab, 1992). For example, closures of the Trans-Canada Highway associated with snow avalanches in Rogers Pass, BC, average about 100 hours per winter. A typical two-hour closure of this highway can result in monetary losses of between $50 000 and $90 000, depending on the number of trucks that are delayed (Morrall and Abdelwahab 1992); the average annual cost of snow avalanches at Rogers Pass alone due solely to traffic delays is therefore several million dollars. This sum does not include the costs of preventing avalanches and clearing the highway during closures. Taking into account the many other highways in British Columbia that are affected by snow avalanches, the total annual cost of traffic delays in that province alone likely exceeds $5 million. This figure is conservative because it does not include railway delays or the effect of closures on timely deliveries, lost business, and other indirect costs.

Property damage caused by snow avalanches in Canada in most years is less than $500 000, although much damage is likely unreported. In comparison, damage from snow avalanches in the European Alps in some winters can be an order of magnitude larger than in Canada. Typical infrastructure damaged in Europe includes residential and commercial buildings in ski resorts, ski lifts, vehicles, and communication and power lines. Snow avalanches also damage forests by uprooting, breaking, and injuring trees.

Most snow avalanches in Canada occur in remote or uninhabited areas during fall, winter, and spring, and therefore go unnoticed. They are extraordinarily common – by one estimate, about 1.5 million snow avalanches large enough to bury a person occur annually in western Canada (Jamieson 2001). Yet, only about 100 avalanche accidents are reported in this region each year. Most snow avalanches occur in the mountains of Alberta, British Columbia, Yukon, and Northwest Territories (Fig. 12). They are also common in parts of Labrador, in the Gaspé region of Québec, along the north shore of the St. Lawrence River, in parts of Newfoundland, and along the eastern margin of the Arctic Islands. Snow avalanches are uncommon, except in localized areas, in the Prairies and Arctic, and in central and eastern Canada.

Flooding occurs at different times of the year in Canada depending on the geographic location of the watershed and the size of the stream or river involved. Large Canadian rivers, e.g. the Fraser and the Red, always flood in late spring due to a combination of melt of heavy snowpacks, prolonged warm weather, and heavy rainfall. The basins of these rivers are too large to be influenced by thunderstorms or by a single cyclonic storm. Mid-sized rivers may also crest during the late spring, although those in temperate areas such as the Pacific coast commonly flood in the fall during periods of heavy rain after snow has accumulated on the ground. Small streams can flood at any time of the year − during spring when snow rapidly melts or rain falls on still-frozen ground; in mid-water thaws; or during summer rainstorms. These streams can reach flood stage very quickly. Ice-jam floods occur in most parts of Canada, but are especially common in the north. They happen when rivers ice-over in the fall or during ice break-up in late spring (Brooks et al. 2001). Ice cover along upstream reaches of a river may break up and then jam against downstream river ice. River water rises against the downstream obstruction and spills out of its channel, inundating the adjacent low-lying floodplains. Flooding may also occur when the ice jam breaks, releasing the impounded water behind the dam.

One of the areas in Canada at greatest risk of flooding is southern Manitoba (Brooks et al. 2001, 2003). In the spring of 1997, rapid snowmelt and heavy rain on water-saturated and frozen soils produced widespread and prolonged flooding along the Red River Valley in North Dakota, Minnesota, and southern Manitoba. Over 7000 military personnel were mobilized for 36 days to assist in strengthening dikes and in relocating nearly 25 000 evacuees along the Red River Valley. The floodwaters inundated the broad valley floor to depths of 12 m and created a vast temporary lake some 40 km across and 1840 km² in area (Fig. 15). Many communities were under water or cut off from surrounding areas; three people drowned; and damage totalled $815 million (Brooks et al. 2001). Winnipeg was spared because the Red River Floodway, constructed at considerable cost after an earlier flood, channelled peak flows around the city (Fig. 16). The floodway has been used in 27 of the 40 years since its completion. In contrast, in the spring of 1951, a flood similar to that of 1997 and resulting from similar conditions, lasted 51 days, forcing the evacuation of 60 000 people in Manitoba; nearly 10 percent of Winnipeg was inundated during this event. Since 1900, other damaging Red River floods occurred in 1904, 1916, 1922, 1923, 1960, 1966, 1969, 1970, 1974, 1979, 1984, 1986 (Andrews 1993; Brooks et al. 2003), and 2009. Given this history, the Red River Floodway is being expanded, with an expected completion date in 2010, further reducing the flood risk to the city.

Geology plays an important role in Red River flooding in two ways. First, the Laurentide Ice Sheet depressed the crust several hundred metres during the late Pleistocene. Because the centre of ice loading was located to the north, the magnitude of depression decreased progressively to the south into the United States. As a result, isostatic rebound has been, and continues to be, greater at the northern end of the Red River Valley than in the south; the gradient of the Red River is therefore decreasing, making it progressively more difficult for the river channel to contain flood waters (Brooks et al. 2001). Second, most of the Red River Valley is underlain by low-permeability clay deposited in glacial Lake Agassiz, a huge late Pleistocene lake impounded by the retreating Laurentide Ice Sheet.

Another notable Canadian flood disaster was the Fraser River flood in early June 1948. A late, but wet spring and the rapid onset of warm temperatures triggered rapid melting of a heavy snowpack in late May (Sewell 1969). The river level rose and overtopped dikes in the Fraser Lowland beginning on 26 May (Fig. 17). Flood waters peaked on 10 June at Mission; by that time, the river had breached over a dozen dikes and flooded about 20 000 ha of land. Much of Agassiz and Rosedale, and parts of Mission were inundated (Andrews 1993). Vancouver was isolated from the rest of Canada, except by air (Sewell 1969). Ten people died, 200 families were left homeless, and 16 000 people were evacuated. Three thousand buildings were destroyed and 82 bridges were washed out. The total cost of the disaster is estimated to have been $133 million (1991 dollars; Andrews 1993).

Severe flooding can also accompany hurricanes that reach Canada. The worst flood disaster in Canadian history occurred in October 1954, when Hurricane Hazel struck Toronto. The hurricane produced record precipitation, with an estimated 210 mm of rain falling over a 36-hour period on the watersheds in the Toronto region (Andrews 1993). The deluge caused the most severe flooding in over 200 years. Because there was considerable urban development on most of the affected flood-plains, flood damage was high, estimated at over $150 million. More than 20 bridges were destroyed, 81 lives were lost, and almost 1900 families were left homeless (Andrews 1993).

Hurricanes are primarily a threat to the Atlantic provinces, although, as mentioned earlier, weakened hurricanes can cause severe flooding in south-central Canada. Damage and loss of life from hurricanes result from winds, storm surges, and flooding caused by heavy rainfall. Storm surges cause the greatest damage and are responsible for 90 percent of all hurricane-related fatalities in the United States (Flanagan 1993). A storm surge is a local rise in sea level that occurs when hurricane winds drive seawater inland. Tropical cyclones commonly generate storm surges of over 3 m, and surges of 12 m or more have been recorded in Bangladesh and Australia (Burt 2004). Higher storm surges develop on broad, shallow coastlines where the forward motion of wind-driven water is impeded by friction. Storm surge and wind damage in the Northern Hemisphere are greatest in the right, forward quadrant of the storm as it makes land-fall−this is the direction the storm is travelling and also rotating, thus the wind speed at the ground is higher there than on the other side of the eye of the hurricane. The height of the surge will also be greater if the hurricane comes onto land at high tide. Hurricane Juan (Fig. 18), which struck Halifax with peak winds of about 160 km/hr on September 29, 2003, was a harsh reminder that Atlantic Canada is vulnerable to tropical storms. It killed eight people and caused more than $200 million in damage in Nova Scotia and Prince Edward Island, including major damage to trees and property in the urban core of Halifax.

Tornadoes are much more common in the midwestern United States and southern Canada than elsewhere in the world, largely because the flat, low topography of central North America allows subtropical and arctic air masses to interact. The southern and central United States, in particular, has just the right combination of weather, topography, and geographic location to make it a perfect spawning ground for tornadoes (Grazulis 2001). Most US and Canadian tornadoes occur in spring and summer in the region extending from Florida to Texas and north to southern Alberta, Saskatchewan, Manitoba, and Ontario (Fig. 19; Environment Canada 2004b).

The most deadly tornado in Canadian history occurred in central Alberta on the afternoon of July 31, 1987. The tornado, rated F4 on the Fujita scale, ripped through Edmonton, killing 27 people and injuring 300 more; property damage amounted to $330 million. The tornado remained on the ground for almost an hour, achieving wind speeds of up to 460 km/hr and cutting a path of destruction 40 km long and up to 1 km wide (Charlton et al. 1997). Other recent deadly tornadoes in Canada include a F3 event that travelled across Pine Lake, Alberta, and into a trailer park in July 2000 (11 people killed, 132 injured); and a F4 tornado that struck Barrie, Ontario, in May 1985 (more than 800 people were left homeless, 8 people died, and 60 were seriously injured).