Original Article Landslides DOI 10.1007/s10346-006-0039-y Received: 21 December 2005 Accepted: 13 March 2006 © Springer-Verlag 2006 Marten Geertsema . John J. Clague 1,000-year record of landslide dams at Halden Creek, northeastern British Columbia Abstract Large, rapid, low-gradient landslides are common in clay-rich glacial sediments in northeastern British Columbia. Many of the landslides create upstream impoundments that may persist for years in small watersheds in the region. We have documented such events in the Halden Creek watershed, 60 km southeast of Fort Nelson. The events are recorded geologically in two ways. First, trees are drowned in lakes dammed by the landslides and subsequently buried by deltaic sediments, where they are protected from decay. Bank erosion later exhumes the drowned trees. Second, landslide deposits with entrained wood are exposed along stream banks. We have reconstructed the recent history of landslide damming at Halden Creek by performing radiocarbon dating on exhumed trees and wood in and beneath landslide deposits at 13 sites in the watershed. Drowned trees range in age from 169±59 to 274±49 14C year BP. Wood in and below landslide deposits yielded radiocarbon ages ranging from modern to 965±49 14 C year BP. Keywords Landslide dam . Drowned forest . Landslide . Radiocarbon dating . Canada . British Columbia . Halden Creek . Fort Nelson Introduction Large, low-gradient landslides are common in clay-rich glacial sediments in northeastern British Columbia (BC), Canada. They commonly dam small streams, forming lakes that may persist for years (Geertsema et al. 2006). These landslides are members of a suite of slope failures that block drainage and impound upstream lakes in mountainous areas throughout the world (Swanson et al. 1986; Costa and Schuster 1988, 1991; Clague and Evans 1994; Hejun et al. 1998; Casagli and Ermini 1999; Korup 2002, 2004, 2005). Halden Creek (Fig. 1) in northeastern BC is an example of a stream subject to recurrent landslide damming. In this paper, we document a recent landslide dam in the Halden Creek watershed and examine geological evidence for past landslides and landslide dams in the basin. Geological and geomorphological setting The Halden Creek watershed (156 km2) is located on the Alberta Plateau (Holland 1976), 60 km southeast of Fort Nelson, within the Black and White Spruce biogeoclimatic zone (Meidinger and Pojar 1991) and the zone of discontinuous permafrost (Brown 1973). Bedrock comprises a flat-lying to gently dipping Cretaceous clastic sequence including, from youngest to oldest, Dunvegan sandstone, Sully marine shale, and Sikanni sandstone (MacIntyre et al. 1998). The Dunvegan sandstone underlies an old plantation surface and caps remnant steep-walled mesas and buttes, bordered by steep slopes (Edwards and Scafe 1996). The mesas have arcuate embayments formed by landslides and perhaps glacial erosion. The mesas are covered by till deposited by the former Laurentide Ice Sheet, with distinctive crystalline erratics derived from the Canadian Shield hundreds of kilometers to the northeast (Mathews 1980). Most of the surface glacial sediments in the Halden Creek watershed have been displaced by earth slides and flows and are therefore designated as colluvium (Fig. 2). Halden Creek drains to the east from a bedrock-walled amphitheater into Tenaka Creek, a tributary of Prophet River. The waters eventually reach Beaufort Sea via Mackenzie River. Materials and methods The study involved examining a recent landslide, its dam and lake, as well as evidence of older landslides and landslide-dammed lakes. Landslides and terrain types were mapped from aerial photographs and through helicopter overview flights. They were verified through field examination of exposures. Landslides were classified according to the system of Cruden and Varnes (1996). The extent and volume of the most recent landslide that dammed Halden Creek in 1996 (Fig. 3) were estimated from digital elevation models of pre- and post-failure surfaces obtained from Province of British Columbia aerial photographs (30BC79176: 122–124 and 30BCC97102: 97–98, respectively). Two sediment samples were collected and analyzed for particle size, Atterberg limits, elemental chemistry, and clay mineralogy using standard techniques. Crosssectional discs were collected from tree stems in landslide debris and drowned trees in growth position. Samples taken from the discs were AMS radiocarbon dated at the NSF Arizona AMS Facility at the University of Arizona. Radiocarbon ages were calibrated using the program OxCal version 3.10 (Bronk Ramsey 2005). Ages obtained from outer rings of trees killed by the same event were combined statistically. Radiocarbon ages on different sets of annual rings from the same tree were wiggle-matched (Bronk Ramsey et al. 2001) to more tightly constrain kill dates of the trees. Ages of inner and outer wood from single trees, as well as from different trees of the same cohort, were also compared statistically. Landslides The steep mesa walls that border the basin consist of competent Dunvegan sandstone overlying deformable Sully shale. These slopes have been sites of rock fall, rotational bedrock landslides, and translational rockslides, many of which have low gradients. Slopes below the mesa walls are completely covered by landslide deposits. In at least one instance, a bedrock landslide triggered an earth flow in unconsolidated sediment. Lateral ridges that extend Landslides Original Article Fig. 3 Orthophotograph of lower Halden Creek watershed, showing landslide scarps, exhumed drowned forests, dated landslide deposits, and the 1996 landslide dam and lake. The lettered sites are those in Table 3 till plain, and bedrock is exposed in places along the stream. All slopes along this reach of the stream have been affected by landslides of different ages. Most of the landslides are wide, translational slides and spreads with transverse horst-and-graben morphology and minor evidence of true flow. Smaller earth slides (Fig. 4) and flows occur frequently as reactivations of parts of larger landslides. Fig. 1 Shaded relief digital elevation model (DEM) of the region around Halden Creek, 60 km southeast of Fort Nelson. The Halden Creek watershed is outlined in white downslope from a displaced bedrock ridge mark the lateral margins of this earth flow. Similar earth flows triggered by rockslides are common elsewhere in northeastern BC (Geertsema et al. 2006). The landslides that formed the dams discussed in this paper happened on a till plain in the lower part of the watershed (Fig. 3). The till is a cohesive diamicton, approximately 10 m thick, overlying flat-lying shale bedrock. Halden Creek is incised into the Fig. 2 Shaded relief DEM of Halden Creek watershed, showing locations of study sites (see Fig. 3 for site numbers). The site outside the watershed boundary (dashed line) is a landslide on Tenaka Creek Landslides Fig. 4 a Earth slide on a gently sloping surface underlain by till and incised by Halden Creek. The landslide occurred sometime between July 22, 1967 and August 16, 1993. Reaction wood in conifers suggests the slide occurred in 1985. Halden Creek flows from right to left; view to the southwest. b Recent reactivation of part of a prehistoric landslide deposit at sites 14 and 15 (Fig. 8) Fig. 6 Geologic section of the 1996 landslide, showing diamicton (till and colluvium) overlying horizontally bedded shale. The movement was mainly translational. Note the vertical exaggeration in the sketch deposits cover 47 ha and have a volume of more than 3 million cubic meters. Shale is exposed along Halden Creek below the landslide deposit. Fig. 5 a Photographic stereopair showing the 1996 landslide and impoundment (Province of British Columbia aerial photographs 30BCC97102: 97–98). b Drowned forest upstream of the 1996 landslide (September 2000). The shoreline of the lake is marked by a dashed blue line. Arrows indicate direction of movement of the landslide that dammed the creek The most recent large landslide (Fig. 5) occurred some time between August 16, 1993, which is the date of a LANDSAT 7 image on which the slide is absent, and August 6, 1997, the date of an aerial photograph that shows the slide. Tree-ring dating suggests that the failure occurred after the 1996 growing season. Precipitation at the nearest climate station (Fort Nelson) was above average in 1996 (Egginton 2005), perhaps contributing to the instability of the slope. Movement was translational along a nearly horizontal rupture surface, accompanied by spreading (Fig. 6). The landslide occurred in clayey till with less than 15% stones. Its Material properties of the 1996 landslide Representative sediment samples collected from exposures of the 1996 landslide deposit along Halden Creek were analyzed for physical, chemical, and mineralogical properties (Tables 1 and 2). The diamicton consists mainly of clayey silt with about 20% sand. The sediment plots in the CL range of the standard plasticity diagram, indicating that it is a low-plasticity inorganic clay. It has a low activity (less than 0.65) and relatively high salinity and pH, perhaps due to its derivation from marine shale. The clay-size fraction has abundant smectite, illite, and kaolinite, and a moderate amount of quartz (Table 2). One of the samples contains traces of chlorite. The silt-size fraction contains abundant quartz and common feldspar. Plagioclase is more abundant than orthoclase. These properties are similar to those of other translational landslide deposits in northeastern BC (Geertsema et al. 2006). Landslide dams and lakes Clague and Evans (1994) recognized four types of dam-forming landslides in the Canadian Cordillera on the basis of source material: (1) bedrock failures, (2) failures in dissected Quaternary fills, (3) failures of thin sediments overlying bedrock, and (4) Table 1 Physical and chemical properties of sediment associated with the 1996 Halden Creek landslide WLa 36.6 40.5 Sample Halden Halden WPa 17.5 19.8 I Pa 19.1 20.7 Activitya 0.63 0.65 Sand (%) 22.0 21.0 Silt (%) 47.5 47.0 Clay (%) 30.5 32.0 Salinity (g/l) 1.63 1.40 pH 8.2 7.9 a Atterberg limits Analyses done by J.K. Torrance, Carleton University, Ottawa Table 2 Clay mineralogy of sediment associated with the 1996 Halden Creek landslide Sample Halden 1 Halden 1 Halden 2 Halden 2 clay silt clay silt Smectite Abundant Illite Abundant Abundant Abundant Chlorite Trace Kaolinite Abundant Abundant Quartz Common Abundant Common Abundant Feldspar Common Common X-ray diffraction was performed by J.K. Torrance, Carleton University, Ottawa Landslides Original Article Fig. 7 Sketches of stratigraphy at site 11 (see Fig. 4a for location), showing radiocarbon ages and sample numbers. Note fluvial sand and silt separating modern and prehistoric landslide debris from failure of thick Quaternary sediments in northeastern BC and adjacent Alberta have been described by Cruden et al. (1993, 1997), Evans et al. (1996), Lu et al. (1999), Fletcher et al. (2002), Miller and Cruden (2002), and Geertsema et al. (2006). The Halden Creek landslide dams fall within the type II category of Costa and Schuster (1988)—they span the entire valley floor and commonly extend onto the opposite slope. The longevity of such dams depends on the rate of inflow to the impoundment and the size, shape, and geotechnical properties of the dam itself (Costa and Schuster 1988). Geertsema et al. (2006) found that many landslide dams consisting of cohesive diamicton and impounding small streams in northeastern BC are stable. Although the dams are incised, the lakes may remain in place for decades, likely due to armoring of overflow channels by clasts in landslide debris. The 1996 landslide dam is substantially incised, but it still impounds a lake. The lake was originally 1,100 m long and had a surface area of 13.5 ha. About 3.8 m of deltaic sediment had accumulated on the pre-dam floodplain when we visited the site in September 2000. The original high-water level, evident from detritus in trees, was an additional 7 m above the delta surface, indicating an original lake depth of about 11 m. Projecting this water level to the dam gives an original overflow height of 13–15 m. failures of mine tailings. The Halden Creek landslides belong to the second group, as they involve relatively thick Quaternary sediments incised by Halden Creek. Other landslides resulting Fig. 8 Stratigraphy at sites 14 and 15 (see Fig. 4b for location), where a recent, small landslide has occurred within an older, larger deposit. The graph of calendric ages shows a wiggle match (black area) for four outer ring and one pith radiocarbon ages (area delineated by line) Landslides Fig. 9 Landslide debris at site 12 overlies fluvial sand, silt, and pebble gravel. Radiocarbon ages were obtained on two roots in the fluvial sediment and on the pith and outer rings of the cottonwood (Populus balsamifera var. tricocarpa) stem projecting from the stream bank. The graph illustrates the increased precision obtained by from wiggle matching the calibrated inner and outer ring ages (black area) compared to the calibrated outer wood age alone (area delineated by line) Table 3 Radiocarbon ages of wood in landslide debris Site 5 5 5 6 7 Sample, laboratory number 1, AA40525 2, AA40526 3, AA40527 1, AA40529 1, AA40530 2, AA40531 6,7 Radiocarbon age (14 C year BP) 112±34 Weighted mean radiocarbon age (year BP) (32%) 1670–1780 AD (64%) 1800–1940 AD (27%) 1640–700 205±34 121±39 175±39 AD (18%) 1920–1960 AD (36%) 1670–1780 AD (59%) 1790–1950 AD 231±28 AD (37%) 1620–1700 AD (39%) 1720–1820 AD (13%) 1920–1960 AD (78%) 1650–1890 AD (18%) 1910–1960 AD (9%) 1520–1590 233±39 AD (39%) 1620–1690 AD (37%) 1720–1810 AD (12%) 1920–1960 AD (47%) 1630–1690 AD (28%) 1730–1810 AD (11%) 1930–1960 AD 8 1, AA40532 139±37 (95%) 1660–1950 AD 11 1, AA40537 post-bomb (95%) 1950–1955 AD 11 2, AA40538 231±43 (10%) 1510–1600 AD (36%) 1610–1700 AD (37%) 1720–1820 AD (12%) 1910–1960 AD 11 3, AA40539 post-bomb (95%) 1950–1955 AD 11 4, AA40540 post-bomb (95%) 1950–1955 AD 11 5, AA40541 184±51 (95%) 1640–1960 AD 11 6, AA40542 381±43 (95%) 1440–1640 AD 11 7, AA40543 1,251±51 (95%) 660–890 Kill date from wiggle matching 2 sigma (95.4%) AD (50%) 1720–1820 (6%) 1520–1580 229±39 6-1 and 7-2 Calendar age range AD Landslides Original Article Table 3 (continued) Site 11 Sample, laboratory number 8, AA40544 12 1, AA40545 1 pith, AA46360 Radiocarbon age (14 C year BP) post-bomb Weighted mean radiocarbon age (year BP) Calendar age range (95%) 1950–1955 AD 644±46 (95%) 1270–1410 AD 724±52 (79%) 1200–1330 AD (16%) 1340–1400 AD 12 Wiggle, 103 rings 644±−46 14 1, AA40549 867±41 (95%) 1040–1260 14 1 pith, AA46361 965±52 (95%) 980–1210 14 2, AA40550 922±49 (95%) 1020–1220 AD 15 1, AA42070 929±47 (95%) 1020–1210 AD 15 3, AA42071 938±59 (95%) 990–1220 14, 15 14-1, 14-2, 15-1, 15-3 908±24 14-1I 14,15 Wiggle, 95 rings 1, AA40518 3757±42 Tenaka 1 pith, AA46362 3660±57 Tenaka Wiggle 120 rings Inversion 3757±42 The dam is about 900 m wide and has a volume of about 1 million cubic meters, much less than the more than 3 million cubic meters of material involved in the landslide, most of which remained on the slope. The size of the contributing catchment is 120 km2. Precise dimensions of prehistoric large landslide dams on Halden Creek are unknown, but the widths of the dams (width = distance along long axis of stream) can be determined from the widths of the landslides. Large dams were long-lived, as were the lakes they impounded. Deltaic sediments deposited in these lakes partially bury drowned forests that grew on the pre-slide floodplain. In contrast, small landslides produced much smaller dams, which were probably short-lived. Although trees were likely killed by inundation, they are unlikely to have been preserved by burial in deltaic sediment. Landslide and dam ages Large and small landslides dam Halden Creek. Trees and other organic debris are preserved in and under the deposits of both large and small landslides and are visible in exposures along Halden Creek. Only the large landslides produce long-lived lakes in which deltas are formed. Over time, the drowned trees disappear, by decay and stand-destroying wildfires with an average return period of about 100 years (British Columbia Ministry of Forests 1995). Thus, only the buried portions of the trees are preserved. These buried remnants may later be exposed by bank erosion. Two types of datable landslide records are thus preserved in the study Landslides (95%) 1365–1410 AD (95%) 1080–1220 AD (16%) 2180–2120 BC (79%) 2110–1970 BC AD AD AD (95%) 1030–1210 AD 908±24 Tenaka Kill date from wiggle matching 2 sigma (95.4%) (95%) 2290–2030 BC (95%) 2200–1890 BC area: (1) plant debris within and under landslide deposits and (2) remnants of rooted drowned trees upstream of landslide dams. Both records are amenable to radiocarbon and tree-ring dating. Landslide record No portion of the studied section of Halden Creek has escaped landsliding (Fig. 3). Therefore, new landslides involve the reactivation of older landslide deposits. Seven exposures of landslide debris with entrained or subjacent wood were found and examined along Halden Creek, and two exposures were examined along Tenaka Creek just upstream of its confluence with Halden Creek (Fig. 2). The landslide deposits at these sites consist of massive clayey diamicton with low percentages of clasts. They overlie fluvial sand and gravel or bedrock. Logs, oriented in the direction of movement, protrude from the base of the landslide deposits, just above the basal shear surfaces (Figs. 7, 8, and 9). More than one layer of logs were found in landslide debris at three sites (sites 5, 7, and 11). Twenty-two radiocarbon ages were obtained on buried wood within and below the landslide deposits from the nine exposures. Some of the radiocarbon ages pertain to smaller landslides that have reactivated within a much larger landslide (sites 6, 7, and 8; Fig. 3). Correlative landslide deposits are exposed at sites 6 and 7; two logs from these sites yielded a weighted mean radiocarbon age of 231±28 14C year BP (Table 3). Table 4 Inferred dates of drowning of forest Site 1 Sample/laboratory number 1, AA4021 Radiocarbon age (14 C year BP) 258±35 Weighted mean radiocarbon age (14 C year BP) Calendar age range (76%) 1510–1680 AD (15%) 1760–1800 AD (4%) 1930–1960 5a 1, AA40528 1, 5a AD (84%) 1460–1680 274±9 263±28 AD (9%) 1720–1810 AD (3%) 1930–1960 AD (31%) 1520–1600 AD (54%) 1610–1670 AD (10%) 1780–1800 AD (1%) 1940–1960 AD (72%) 1640–1670 AD 6-1, (22%) 1780–1800 AD 7-2 (1%) 1940–1960 1, 5a 247±20 2, AA40533 169±59 (95%) 1640–1960 10 1 pith, A40534 226±46 (9%) 1510–1600 1-2, AA40535 210±43 AD AD (32%) 1610–1700 AD (38%) 1720–1820 AD AD (13%) 1910–1960 10 (1%) 1520–1550 Combined from Tables 3 and 4 AD 8 (2%) 1830–1880 Kill date from wiggle matching AD AD (28%) 1630–1700 AD (45%) 1720–1820 AD (4%) 1830–1880 AD (16%) 1910–1960 AD 10 3 pith, AA40536 268±43 (82%) 1480–1680 AD (10%) 1760–1810 AD (3%) 1930–1960 AD 10 2 pith, AA46359 201±50 (95%) 1630–1960 AD 13 1, AA40548 196±43 (71%) 1640–1820 AD (7%) 1830–1890 AD (17%) 1910–1960 AD Landslides Original Article Table 4 (continued) Site Sample/laboratory number 10, 13 10 pith, 1,2,3 Radiocarbon age (14 C year BP) Weighted mean radiocarbon age (14 C year BP) 203±30 235±27 Calendar age range (27%) 1640–1690 AD (51%) 1720–1810 AD (18%) 1920–1960 AD (53%) 1630–1690 AD (34%) 1760–1810 AD (9%) 1930–1960 10, 13 Wiggle, 157 rings 203±30 Sites 5, 7, and 11 contain woody debris at several levels within and below landslide deposits. Multiple layers of logs in landslide debris do not necessarily indicate multiple events. Radiocarbon ages from site 5 are inverted, but they overlap statistically. Thus, the logs may be of similar age or they may represent reworked material. The ages at site 7 are in proper order, but they too overlap statistically. Site 11 (Figs. 4a and 7), however, does have landslide units of different ages. Based on aerial photographs and a satellite image, the most recent slide occurred between 1967 and 1983. Four of the five uppermost radiocarbon samples yielded modern (postbomb) ages, consistent with a post-1967 landslide (Fig. 7). A root at the base of a sand–silt unit underlying the landslide deposit also yielded a modern date. However, a twig in a 6-cm thick peat bed near the top of a sand–silt unit gave a radiocarbon age of 381±43 14C year BP, which is significantly older than ages from two lower wood samples. This inversion can be explained by fluvial erosion, transport, and deposition of the twig. A log underlying a second, lower landslide deposit gave a radiocarbon age of 231±43 14C year BP, and a twig from near the base of an underlying silty clay unit yielded an age of 184±51 14C year BP. Although inverted, these two ages are statistically equivalent. The oldest age in the sequence at site 11 came from the lowest sample. Twigs and roots from a 2-cm thick peat bed between fluvial gravel units gave an age of 1,251±51 14 C year BP. Precise dating of the Halden Creek landslides is challenging, given the radiocarbon age inversions, statistical overlap in calendar dates, and the possibility of reworking of older material. Greater precision can be obtained by combining radiocarbon ages, where there is reasonable certainty of a common kill date, and by wiggle matching. The weighted means of radiocarbon ages of the outer rings of logs at sites 6 and 7 and at sites 14 and 15 were calculated (Fig. 8; Table 4) to reduce the age uncertainty. The combined kill date at sites 14 and 15 was then wiggle-matched with a pith date for an even tighter constraint. Wiggle matching also reduced the age uncertainty at site 12 (Fig. 9) and at Tenaka Creek (Fig. 2). At least eight landslides are indicated by the radiocarbon ages. All of the dated landslide deposits along Halden Creek yielded calendar ages within the last millennium (Table 3). The landslide Landslides Kill date from wiggle matching AD (40%) 1640–1680 AD (43%) 1760–1800 AD (12%) 1930–1960 AD deposit at Tenaka Creek, however, gave a much older age, approximately 4,000 calendar years BP. Landslide dam record Five exposures of deltaic sediments deposited in landslidedammed lakes were found along lower Halden Creek (Figs. 2 and 3). Rooted trees truncated at or near the modern ground surface were preserved in the sediments and were later exposed by bank erosion (Figs. 5 and 10). Figures 11 and 12 show remnants of drowned forests exposed in stream banks composed of silt and sand at sites 1 and 13, respectively. The fossil trees at site 1 are truncated below the sediment surface, whereas those at site 13 protrude above the surface. The difference may relate to the younger age of the forest at site 13 (Table 4). Alternatively, an upper sediment unit may have been removed by erosion at that site. Nine radiocarbon ages were obtained on outer and pith rings of drowned trees at five sites. As in the case of the landslide ages, greater precision was achieved by combining radiocarbon ages and by wiggle-matching ages on inner and outer rings (Table 4). Fig. 10 Deltaic sediments partially burying drowned forest in the 1996 landslidedammed lake (September 2000). About 3.8 m of sediment had accumulated above the pre-slide floodplain when this photograph was taken Fig. 11 a Stratigraphy and photograph of an exhumed forest in deltaic sediment deposited in a former landslide-dammed lake (site 1). b Graph showing calibrated radiocarbon age of drowned tree at site 1. c Combined calibrated ages of outer rings of drowned trees at sites 1 and 5a. d Increased precision is achieved by combining the ages of drowned trees at sites 1 and 5a, with ages of wood in the landslide deposit at sites 6 and 7 Kill dates of the drowned trees are less than 300 14C year BP (Table 4) and, thus, span only the most recent third of the dated Halden Creek landslide period. The radiocarbon ages and locations of the samples suggest that trees at four of the five sites constitute two populations. Based on proximity and geomorphic setting, trees at sites 1 and 5a form one group, and those at sites 10 and 13 form a second group. The 1996 landslide-dammed lake, which is more than 1 km long, is a modern analogue for large prehistoric lakes in the basin. This analogue suggests that the grouped sites are well within the limits of possible impoundments. Only one dated landslide possibly corresponds to a drowned forest, suggesting that our event record is incomplete. The ages on the landslide at sites 6 and 7 are statistically equivalent to those of the upstream, drowned forests at sites 1 and 5a (Fig. 3). Assuming Fig. 12 a Stratigraphy and photograph of exposure at site 13. A drowned tree extends above the modern ground surface. b Graph showing calibrated radiocarbon age of drowned tree at site 13. c Combined calibrated ages of outer rings of drowned trees at sites 10 and 13. d Graph showing a wiggle match of the three combined pith ages at site 10 and the combined outer wood age shown in c (Table 4) Landslides Original Article Fig. 13 Halden Creek landscape evolution model. a. Pre-slide condition with stream bordered by floodplain forest. b. Landslide blocks stream and drowns the floodplain forest. Deltaic sedimentation begins, partially burying the drowned forest. c,d Progressive incision of landslide dam; continuing deltaic sedimentation. e Exposed portion of buried forest either decays or is burned. f Halden Creek migrates, exhuming drowned forest the four radiocarbon ages at the four sites record the same event, we obtain a calibrated age range of 1640–1670 AD (72% probability) for the landslide (Table 4). This 30-year interval is much shorter than the window of 320 years (1640–1960 AD) provided by the single sample at site 8. Discussion and conclusions Wood within and beneath landslide deposits is commonly used to date mass movements (e.g., Lang et al. 1999). However, to our knowledge, our study is the first to use drowned trees preserved in deltaic sediments of landslide-dammed lakes to date these events. The Halden Creek watershed is subject to frequent landslides and upstream inundation due to blockage of the drainage. The landslides are translational, moving on nearly horizontal surfaces and involving spreading. Their widths are typically 300 to 1,000 m, and the volumes of the larger failures range from 1 to 5 million cubic meters. These large landslides produce relatively long-lived dams. Smaller slides and flows occur due to reactivation of older, larger landslide deposits and may expose sediments left by the older events. The smaller landslides are less than 100 m wide and involve 10,000–50,000 m3 of sediment. They are too small to produce long-lived dams. A model of basin evolution involving landsliding, lake development, delta formation, and subsequent incision is illustrated in Fig. 13. A stream flowing along a forested floodplain is dammed by Landslides a large landslide. The resulting lake drowns the forest, and deltaic sediments begin to accumulate in the lake. Over time, the lake lowers as the dam is partially incised. Exposed portions of trees disappear through decay and fire, but buried portions are preserved in the deltaic sediment. Eventually, the landslide dam is entirely breached and the stream channel is reestablished at its pre-slide base level. The stream erodes the banks of the delta and exposes the drowned pre-slide forest. The documented landslide record of lower Halden Creek spans the last millennium. From this, one might conclude that the landslides are a response to climate change over the last thousand years. However, the modern climate regime, which triggered the expansion of muskeg in northeastern BC, was achieved about 5,000–6000 years ago (MacDonald and McLeod 1996). A more likely explanation is that the lower Halden Creek system is so active that landslide deposits older than about 1,000 years are removed from the basin or are buried by younger landslides. The ca. 3,800-year event recorded at Tenaka Creek, a much larger watershed than that of Halden Creek, indicates that landslides have been occurring for more than just the last thousand years. The age range of the drowned forests spans only the last 300 years (500 years at the extreme limits of 2-sigma age ranges), in contrast to the 1,000-year record of landslides. The drowned forests date landslides, but the record is shorter. This difference may relate to the greater erodibility of the deltaic sand and gravel in comparison to the cohesive landslide diamicton. In addition, the drowned forests are located on the Halden Creek floodplain, whereas the landslide deposits occur on adjacent slopes. The drowned forests are thus subject to more frequent fluvial erosion and are removed from the sedimentary record sooner than the landslide deposits. Implications for forest management Episodic landslides and impoundments in lower Halden Creek place forests in the watershed at risk. Our data indicate a minimum of eight landslides in the last millenium and three landslideinduced forest drownings in the last 300 years over a 6-km reach of Halden Creek. Furthermore, all slopes in the study area have been modified by landslides and are unstable. With an 80- to 100-year forest-harvesting rotation period, it is likely that some forests will not reach maturity due to landsliding or impoundment. Thus, the probability of landslides and landslide damming are factors that should be considered in economic models evaluating the hazards of harvestable working forests. Acknowledgement The project was funded by the British Columbia Ministry of Forests and Range and Forest Renewal British Columbia. We thank Ljiljana Knezevic, Chad Seigel, and Barbara Wohlfarth for assistance in the field. Theo van Asch, Rens van Beek, and Steven de Jong reviewed an early draft of this manuscript. Mauri McSaveney and an anonymous referee reviewed the paper for the journal. Sean Barry, Marlene Flannery, and Richard Franklin drafted the figures. References British Columbia Ministry of Forests (1995) Biodiversity guidebook. Forest practices code of British Columbia. British Columbia Ministry of Forests, Victoria, BC, p 99 Bronk Ramsey C (2005) OxCal Version 3.10 Bronk Ramsey C, van der Plicht J, Weninger B (2001) ‘Wiggle matching’ radiocarbon dates. Radiocarbon 43:381–389 Brown RJE (1973) Permafrost. Canada Department of Energy, Mines and Resources, National Atlas of Canada. Plate 11–12, scale 1:15,000,000 Casagli N, Ermini L (1999) Geomorphic analysis of landslide dams in the Northern Apennine. Chikei 20:219–249 Clague JJ, Evans SG (1994) Formation and failure of natural dams in the Canadian Cordillera. Geol Surv Can Bull 464:35 Costa JE, Schuster RL (1988) The formation and failure of natural dams. Geol Soc Amer Bull 100:1054–1068 Costa JE, Schuster RL (1991) Documented historical landslide dams from around the World. U.S. Geological Survey, Open-File Report 91–239, p 486 Cruden DM, Varnes, DJ (1996) Landslide types and processes. In: Turner AK, Schuster RL (eds) Landslides investigation and mitigation. National Research Council, Transportation Research Board Special Report 247:36–75 Cruden DM, Keegan TR, Thomson S (1993) The landslide dam on the Saddle River near Rycroft, Alberta. Can Geotech J 30:1003–1015 Cruden DM, Lu Z-Y, Thomson S (1997) The 1939 Montagneuse River landslide, Alberta. Can Geotech J 34:799–810 Edwards WAD, Scafe D (1996) Mapping and resource evaluation of the tertiary and preglacial sand and gravel formations of Alberta. Alberta Geological Survey, Open File Report 1994–06 Egginton VN (2005) Historical climate variability from the instrumental record in Northern British Columbia and Its iinfluence on slope stability. Unpublished M.Sc. Thesis, Simon Fraser University, Burnaby, B.C., p 144 Evans SG, Hu XQ, Enegren EG (1996) The 1973 Attachie slide, Peace River Valley, near Fort St. John, British Columbia, Canada: a landslide with a high-velocity flowslide component in Pleistocene sediments. In: Proceedings of the 7th International Symposium on Landslides, vol. 2. A.A. Balkema, Rotterdam, pp. 715–720 Fletcher L, Hungr O, Evans SG (2002) Contrasting failure behaviour of two large landslides in clay and silt. Can Geotech J 39:46–62 Geertsema M, Clague JJ, Schwab JW, Evans SG (2006) An overview of recent large catastrophic landslides in northern British Columbia, Canada. Eng Geol 83:120–143 Hejun C, Hanchao L, Zhuoyuan Z (1998) Study on the categories of landslide-damming of rivers and their characteristics. Journal of Chengdu Institute of Technology 25:411–416 Holland SS (1976) Landforms of British Columbia: a physiographic outline. British Columbia Department of Energy, Mines and Petroleum Resources, Bulletin 48, p 138 Korup O (2002) Recent research on landslide dams—a literature review with special attention to New Zealand. Prog Phys Geogr 26:206–235 Korup O (2004) Geomorphic characteristics of New Zealand landslide dams. Eng Geol 73:13–35 Korup O (2005) Geomorphic hazard assessment of landslide dams in South Westland, New Zealand: fundamental problems and approaches. Geomorphology 66:167–188 Lang A, Moya J, Corominas J, Schrott L, Dikau R (1999) Classic and new dating methods for assessing the temporal occurrence of mass movements. Geomorphology 30:33– 52 Lu Z-Y, Cruden DM, Thomson S (1999) Landslides and preglacial channels in the western Peace River lowland, Alberta. Proceedings, 51st Canadian Geotechnical Conference 1:267–274 MacDonald GM, McLeod K (1996) The Holocene closing of the ‘Ice-Free Corridor’: a biogeographical perspective. Quat Int 32:87–95 MacIntyre DG, Okulitch AV, Taylor GC, Cullen B, Massey N, Bellefontaine K (compilers) (1998) Geology, Fort Nelson, British Columbia; Central Foreland Map NO-10-G. Geological Survey of Canada, Open File 3604, map, scale 1:500 000 Mathews WH (1980) Retreat of the last ice sheets in Northeastern British Columbia and adjacent Alberta. Geological Survey of Canada, Bulletin 331, p 21 Meidinger DV, Pojar JJ (1991) Ecosystems of British Columbia. British Columbia Ministry of Forests, Special Report Series 6, p 330 Miller BGN, Cruden DM (2002) The Eureka river landslide and dam, Peace River Lowlands, Alberta. Can Geotech J 39:863–878 Swanson FJ, Oyagi N, Tominaga M (1986) Landslide dams in Japan. In: Schuster RL (eds) Landslide dams: process, risk and mitigation. American Society of Civil Engineers, Geotechnical Special Publication 3:131–145 M. Geertsema ()) Research Geomorphologist, B.C. Ministry of Forests and Range, 1011 4th Ave., Prince George, BC V2L 3H9, Canada e-mail: Marten.Geertsema@gov.bc.ca Tel.: +1-250-5656923 Fax: +1-250-5654349 J. J. Clague Department of Earth Sciences, Simon Fraser University, 8888 University Dr., Burnaby, BC V5A 1S6, Canada Landslides