An investigation of subaqueous failures triggered by the 1935 CE M_w 6.1 Témiscaming earthquake, Quebec, Canada, as an analogue for a paleoseismic study

Studies of lake sub-bottom deposits in areas that experienced significant modern or historical earthquakes recognized that widespread mass movement and/or turbidite deposits (collectively referred to as mass transport deposits (MTDs)) can be stratigraphic evidence of significant seismic events (e.g., Shilts 1984; Siegenthaler et al. 1987; Shilts and Clague 1992; Chapron et al. 1999; Nomade et al. 2005). The compilation of event horizons by mapping and dating MTDs occurring at specific stratigraphic levels within a lake sub-bottom allows disturbance events to be evaluated and compared both within, and possibly between nearby lake basins, thereby supporting the interpretation of a seismic or aseismic trigger for each event horizon (see Schnellmann et al. 2002, 2006; Strasser et al. 2013; Van Daele et al. 2015). The recognition of paleoearthquakes and their attributes by numerous investigations of prehistoric MTD event horizons in lakes through reflection seismic surveys, core sample collection, and application of chronological tools have considerably augmented historical earthquake records at locations over a range of global neotectonic settings (e.g., Beck et al. 1996; Karlin and Abella 1996; Schnellmann et al. 2002, 2006; Karlin et al. 2004; Becker et al. 2005; Strasser at al. 2006, 2013; Moernaut et al. 2007, 2009, 2018; Bertrand et al. 2008; Anselmetti et al. 2009; Maloney et al. 2013; Smith et al. 2013; Kremer et al. 2015; Praet et al. 2017, 2022; Oswald et al. 2021; Daxer et al. 2022; Bellwald et al. 2024; Vervoort et al. 2024).

In eastern, central, and northern Canada, only one historical earthquake has generated a known surface rupture, the M_s 6.3 1989 CE Ungava earthquake located in northern Quebec far from the zones of concentrated seismic activity in southern Quebec–southeastern Ontario (Adams et al. 1991). There are also not yet any confirmed examples of early postglacial faults that may have been generated from rebound stress associated with the thinning and retreat of the Laurentide Ice Sheet (LIS; see Brooks and Adams 2020; Adams and Brooks 2022). The identification of clusters of commonly-aged subaqueous and submarine MTDs, however, has provided geomorphic or stratigraphic evidence of some historical and prehistorical earthquakes within or proximal to Charlevoix–Kamouraska and Western Quebec seismic zones (WQSZs; see Shilts 1984; Shilts and Clague 1992; Shilts et al. 1992; Syvitski and Schafer 1996; St-Onge et al. 2004; Doughty et al. 2010a; Normandeau et al. 2013; Lajeunesse et al. 2017; Trottier et al. 2019; Mérindol et al. 2022; Brooks and Perret 2023). The investigation of subaqueous MTDs in lakes, in particular, has widespread potential for expanding the record of paleoearthquakes in central Canada because lakes represent a continuous or near-continuous record of postglacial sedimentation, are common within the region, and are especially ubiquitous on the Canadian Shield (Brooks 2015).

On 1 November 1935 CE, an M_w 6.1 earthquake occurred near the Town of Témiscaming, Quebec, that is the largest instrumentally-recorded event in the WQSZ (Figs. 1 and 2; Lamontagne et al. 2018; Earthquakes Canada 2024). Hodgson (1935, 1936, 1945) visited the area shortly after the earthquake documenting rockfall, damaged chimneys, cracked brick walls, shifted heavy objects (e.g., cabin, spare train-track rails), and cracks in shoreline sediments, but no surface rupture was identified. However, he also reported that the normally clear waters of Tee Lake and parts of Kipawa Lake became discoloured after the main shock (Fig. 2). Described as café au lait, white or milky (see also Drouin 1989), this discolouration remained for at least a month. An explanation for the discolouration came from Shilts (1984), who mapped numerous, hummocky slump deposits (i.e., MTDs) within a sub-bottom acoustic profile (SAP) survey of the western basin and east arm of Tee Lake, which is located proximal to the earthquake epicentre (Fig. 3). Shilts (1984) and Shilts and Clague (1992) both attributed the MTDs and the observed discolouration of lake waters to subaqueous failures of gyttja and underlying clastic sediments triggered by shaking from the 1935 CE earthquake. They also mention that similar hummocky deposits are present within the nearby bays of the Kipawa Lake, located several kilometres to the east of Tee Lake (Fig. 3). Related to the discolouration of Tee Lake, Doig (1991) described a silt layer, 1–2 cm thick, overlain by a chaotic zone of organic material with silt fragments contained near the top of three sediment cores recovered from the western basin of the lake. He attributed this sequence to the redeposition of sediments dispersed into the water column by the earthquake. A subsequent SAP survey by Doughty et al. (2010a) recognized that late Holocene-aged, subaqueous MTDs are much more widely distributed in the Kipawa Lake basin than was initially recognized. They mapped hummocky MTDs up to about 21 km northeast and 23 km north of the epicentre (Fig. 3). They also attributed these MTDs to failures triggered by the 1935 CE earthquake.

None of the subaerial impacts observed by Hodgson represent a long-term record of the earthquake that might be an analogue for a paleoseismic investigation to focus upon. Shilts (1984), Shilts and Clague (1992), and Doughty et al. (2010a), however, demonstrated that a relatively widespread geomorphic record is preserved consisting of subaqueous MTDs within the sub-bottom of local lakes. This paper presents a case study of the MTD signature generated by the moderately-strong 1935 CE Témiscaming earthquake to provide insights into applying such results to paleoseismic studies in a Canadian Shield setting. The specific goals are to

• 1) expand and constrain the spatial distribution of subaqueous late Holocene MTDs in lakes within the Témiscaming area, Quebec–Ontario,

• 2) confirm the connection of the MTDs to the 1935 CE Témiscaming earthquake,

• 3) characterize the failed deposits, and

• 4) utilize the spatial MTD dataset to estimate the minimum magnitude of the triggering earthquake using empirical data, as would be done in a paleoseismic study, and test the results against the known instrumental earthquake magnitude.

The study area is located near Témiscaming, Quebec, and straddles the Ontario–Quebec border (Figs. 1 and 2). Most of the area is forested terrain that is sparsely developed and with road access of variable density and quality. The area contains literally hundreds of lakes, although most are small and less than 1 km² (Fig. 2). Most are also oligotrophic, lack major contributing streams, and receive a negligible to minimal clastic sediment load. Reflecting this, the postglacial lake bottom deposits typically are composed of organic-rich gyttja sediment. Kipawa Lake (about 300 km²) is the largest in the area and is up to 60 km across from the most distant locations, locally up to about 100 m deep, and exhibits an irregular shape, imparted by many elongated bays and channels, and numerous islands. Modern lake level is regulated by dams originally built in the early 20th century and ranges from 269.75 to 267.45 m asl between navigational and winter seasons, respectively (ORRPB 2023). Other lakes relevant to this study range in size from West Beauchêne (16.7 km²) and Bois Franc (15.0 km²) down to Serene (1.5 km²) and Spring (0.5 km²) lakes (Fig. 2). Included in the study area is the narrow (1–1.5 km wide), southern-most portion of Lake Timiskaming, which feeds into the Ottawa River (Fig. 2).

The study area is underlain by Canadian Shield bedrock of the Grenville Province, composed predominately of gneiss with lesser areas of migmatite, granite, and granodiorite of the southwest–northeast-oriented Central Gneiss Belt (Baer et al. 1977). It straddles the Timiskaming graben, a northwest extension of the Ottawa-Bonnechere graben, which is a failed arm of Iapetan rifting that branches from the St. Lawrence Valley rift system (Fig. 1; Kumarapeli and Saull 1966; Kumarapeli 1985; Bleeker et al. 2011). A narrow, linear trench, oriented approximately northwest–southeast, truncates the area and contains the southern arm of Lake Timiskaming and the northern end of the Ottawa River (Fig. 2). East and west of the trench, rolling bedrock hills, up to about 460 m asl, dominate the landscape. Numerous faults, brittle lineaments, and ductile deformation structures are present within the bedrock, especially to the east of the trench (see Lamontagne et al. 2020). The topographic expression of these features influences lake morphologies, as is apparent by, for example, the elongated linear bays and channels of Kipawa Lake as well as linear to curvilinear chains of narrow lakes on the east side of the study area (Fig. 2).

Reflecting the former presence of the Laurentide Ice Sheet, surficial geology in the study area consists generally of patches of bare bedrock, and discontinuous drift veneer or a continuous drift cover commonly composed of nearshore and beach glaciolacustrine deposits, or eskers and glaciofluvial deposits (see Veillette and Daigneault 1987). To the west of Lake Timiscaming-Ottawa River, the surfical geology is generally similar, but a large area is covered by proglacial fluvial deposits of the McConnell moraine (Gartner 1979a, 1979b; Veillette and Daigneault 1987). Many small lakes located upon this moraine occupy kettle depressions. Aspects of the late Quaternary deposits preserved within the northern portion of the Lake Timiskaming basin are reported by Shilts (1984), Shilts and Clague (1992), Doughty et al. (2010b, 2013, 2014), and Brooks and Pugin (2019).

Portions of the study area along the trench occupied by Lake Timiskaming-Ottawa River and the lower areas proximal to modern Kipawa Lake were inundated during the initial phase of glacial Lake Barlow, which formed at about 11.0 ka cal B.P., as the LIS locally retreated northeast of the Ottawa River (Vincent and Hardy 1979; Veillette 1994; Dyke 2004). Many of the locations investigated in this study were submerged during this phase, including Kipawa Lake (survey areas 18–30), Tee Lake (7), West Beauchêne (8), and probably Sheffield Lake (15; Fig. 4), as indicated by modern lake elevations and the presence/absence of biological indicator species that reveal submergence by a large glacial lake (Dadswell 1974; Veillette 1994). The erosion of morainic deposits underlying the glacial lake outlet caused water levels to fall within the study area and become confined within the trench. Different stages of the glacial lake are termed successively Barlow, Barlow–Ojibway, and Ojibway that correspond to geographical extents of the lake in the southern, southern–northern, and northern portions of the large regional basin, respectively (see Veillette 1994). The glacial lake drained catastrophically northwards into the Hudson Bay basin around 8.2 ka cal B.P. (Brouard et al. 2021). Within the study area, glaciolacustrine deposits relating to the former presence of the glacial lake are preserved in local lake basins.

The M_w 6.1 1935 CE Témiscaming earthquake occurred at the northwestern end of the WQSZ, a band of elevated seismicity that extends southeast–northwest from upstate New York through southwestern Quebec and southeastern Ontario (Fig. 1; see Basham et al. 1982; Adams and Basham 1989, 1991). With an epicentre at 46.739°, −79.024° (Fig. 2; Bondár et al. 2015), the earthquake was widely felt and caused local damage (see Lamontagne 2021). Source mechanism modelling by Bent (1996) indicates that the rupture occurred as a thrust fault at a depth of about 10 km and along an approximately northwest–southeast striking plane that dips at 45°. The clustering of the 1935 CE earthquake and associated aftershocks has been described as the Kipawa seismic zone (Bent et al. 2002; Adams and Vonk 2009;). The “Kipawa Millenium” earthquake (M_n 5.2; 46.88°,−78.92°), occurring on 1 January 2000, and located about 22.5 km northeast of the Témiscaming, is the second largest historical event within the area (see Bent et al. 2002).

SAP surveys were conducted in late July–August during 2019, 2021, 2022, and 2023 on Kipawa Lake and connected basins (15 locations), other lakes (14 locations), and the southern-most end of Lake Timskaming (one location; Fig. 4; see Supplementary Material, Table S1). The surveys used a Knudsen CHIRP Pinger SBP™ echosounder system (Knudsen Engineering Ltd.) with low frequency 15 or 3.5 kHz and high frequency 200 kHz transducers mounted on the side of a 4.9 m (16 ft) aluminum boat powered by a 25 hp motor. Profiling boat speed ranged between 5.5 and 6.9 km/h. A NovAltel Smart6L™ or Smart7™ receiver collected real-time corrected, streamed GPS data along the profiling routes that were recorded in combination with the digital profiling data. The velocity of sound waves in the lake waters was assumed to be 1450 m/s.

The SAP survey data were collected in a zig-zag pattern when reconnoitering a given sub-basin, or a grid pattern when focusing on a particular area of interest; in places the zig-zag pattern evolved into a grid pattern as an MTD(s) was encountered (see Fig. 5 and Supplementary Material, Figs. S1–S31). Grid spacing usually ranged between 40 and 100 m wide, but could be narrower or wider depending on whether features of interest were encountered or not, respectively. As summarized in Table S1, most surveys used exclusively the 15/200 kHz transducers, but some collected in 2022 used the 3.5/200 kHz combination. During 2021, several of the 15 kHz surveys were supplemented with 3.5 kHz profiles that targeted MTDs of interest. Also, SAP equipment problems led to the 200 kHz transducer being turned off for four surveys on the recommendation of the manufacturer (see Table S1). Major equipment problems at the end of the 2022 field season yielded uninterpretable results from surveys at Audoin Lake, northwest of Grindstone Lake.

Within a given survey area, where MTDs were identified, the lake bottom surface and the MTD surface and bottom reflections were picked using the horizon picking tool within IHS Markit Kingdom™ v.8.8 software. From the 2019, 2021, and 2022 surveys, the x-y-z coordinates of points along all three picked horizons from each survey area were exported separately into and gridded using Golden Software Surfer™ v.11, and then mapped and edited using Blue Marble Geographics GlobalMapper™ v.20 and Adobe Illustrator™ CS6, respectively. This yielded a bathymetric map(s) for each of the survey areas and, where MTDs are present, a map representing MTD thickness (see Table S2). Project time constraints precluded compiling similar maps from the picked horizons in the 2023 data.

Mapping subaqueous failures is done from boat surveys using geophysical equipment and is time-consuming; hence, it was not practical to survey every single lake or every part of larger lakes within the study area. Instead, the study focused on sampling the distribution of the late Holocene MTDs rather than collecting a comprehensive inventory, as is commonly the goal with subaerial earthquake-triggered landslides for modern earthquakes (Harp et al. 2011). A sampling strategy was devised from surveys acquired in 2019 on Tee Lake and parts of Kipawa Lake, where Shilts (1984) and Doughty et al. (2010a) had reported MTDs, respectively (Fig. 3). The locational pattern was readily apparent; the MTDs occur preferentially at the base of relatively steep slopes and in depths greater than about 30 m or more. Locations of interest (small lake basins, sub-basins of larger lakes) were duly identified qualitatively on commercial and internet fishing bathymetric maps (many of unknown quality), and from discussions about water depths with local cottage owners and fishermen. The single exception to this is survey area 5 located at the southern end of Lake Timiskaming (Fig. 4), which was identified on a Canadian Hydrographic Service digital bathymetry. The prospective sites were prioritized to expand and constrain the MTD distribution previously reported by Shilts (1984) and Doughty et al. (2010a). All of the surveys were conducted as day trips, although several areas required multiple days (or part days) to complete (see Table S1).

Lake bottom samples were acquired from a frozen lake surface using two different coring techniques, depending on water depth and targeted sediments. The geographical coordinates of sites of interest were determined using the SAP data, and later located in the field using real-time corrected, streamed GPS data collected with a NovAltel Smart6L™ or Smart7™ receiver. The coring sites are shown in Fig. 6 and listed in Supplementary Material, Table S3.

In deep water (47–87 m deep), lake bottom samples were recovered in March 2021, 2022, and 2023, using a Tech-Ops open-barrel gravity corer with a transparent, polycarbonate tube, 1.2 m long and 10 cm diameter (see Alpay et al. 2016). A 0.3 m diameter circular screen was attached to the sampling tube about 0.8 m above the bottom of the tube to limit the depth of penetration in soft lake bottom sediments. The approximately 35 kg sampler was attached to a rope and suspended from a hoisting frame (either a heavy-duty step ladder or tripod). The sampler was then lowered down to the lake bottom through an approximately 0.5 × 0.5 m hole cut into the ice surface. Lowering was monitored using a Humminbird Solix 12™ fish finder sonar unit to ensure a controlled descent into the lake bottom. Lowering was especially slow over the lower 5 m to minimize disturbance of the lake bottom surface. The weight of the corer caused the sampling tube to sink into the lake bottom sediments. A gas-powered 1.3 hp capstan winch hoisted the sediment-filled core sampler from the lake bed. At the lake surface, a rubber bung was inserted into the bottom of the sampling tube, the corer disassembled, and the base of the sampling tube placed in a sealed coupler mounted onto an aluminum baseplate. Pressurized water was then pumped into the coupler to push the bung and overlying sediment up and out of the top of the sampling tube for subsampling at intervals of 0.5 cm between 0 and 10 cm depth, 1.0 cm between 10 and 20 cm depth, and 2 cm between 20 and 40 cm depth. The subsamples were placed in water-tight, plastic sample bottles, stored to prevent freezing in a hotel room and then a refrigerator at 601 Booth Street, Ottawa. They were later submitted to Flett Research Ltd. for Pb-210, Cs-137 and Ra-226 analyses.

In March 2022, the gravity corer was used to collect sediment on unfailed slopes, but there were problems penetrating into the deposits underlying the gyttja. A Livingston piston corer, therefore, was used in March 2023 (see Livingston 1955), but was limited to water depths of less than 20 m. Two sites were cored on Tee Lake, each consisting of a pair of sampling holes located about 1 m apart. Sampling used 50.8 mm (O.D.) aluminum tubes that recovered segments of core sediment about 1.0 m long. Sampling depth intervals were staggered between the pair of holes to overlap by about 0.5 m, resulting in the recovery of a continuous, composite sample of sediment between the two holes. Immediately upon recovery, the core tubes were sealed with plastic caps and taped securely, and then stored to prevent freezing.

The two sets of piston core samples and one of the gravity core samples were analyzed using a Siemens SOMATOM Definition AS + 128 CT Scanner at the Institut national de la recherche scientifique, Quebec City, Quebec, to produce tomodensitometry radiographs of the unsplit, recovered deposits. Later in the laboratory, the core sediments were split, photographed, logged, and subsampled for macrofossils, and sediment texture, carbon, and moisture analysis (following Girard et al. 2004). The piston core samples were measured for sediment compressive strength using a Pocket Penetrometer (see https://www.forestry-suppliers.com/Documents/4296_msds.pdf). At the time of writing, the piston cores are archived in cold storage at the Geological Survey of Canada in Ottawa. The split gravity core sediments were very soft (i.e., soupy) and not retained.

During core logging, three macrofossils found in core sediments were subsampled using clean, stainless-steel knives, wrapped in aluminum foil, and stored at 4 °C. They were subsequently cleaned using distilled H₂0 and oven dried at 80 °C for 24 h. The samples were then submitted to the Lalonde Accelerator Mass Spectrometry (AMS) Laboratory, University of Ottawa, for pre-treatment in HCl–NaOH–HCl and then for AMS radiocarbon analysis. The reported radiocarbon ages were calibrated to the 2σ confidence limits using OxCal 4.4 (Bronk Ramsey 2009) and IntCal20 (Reimer et al. 2020).

MTD facies are located at the base of steeper slopes within the SAP lines (Figs. 7 and 8; see Supplementary Material, Fig. S32, for larger versions of the profiles). Locally up to 13 m thick, the facies typically exhibits an irregular to hummocky topography on profile lines. Internally, the facies is transparent, but locally may be weakly chaotic. Locally within some survey areas, the base of the facies was indistinct because of the absence of a basal reflection. Where this occurred, the base of the facies was approximated during picking based on the general architecture of the sub-bottom stratigraphy. As exemplified in Fig. 7, the mapped MTDs all lack an overlying drape of post-failure lacustrine deposits that can be discerned from the SAP reflections. The apparent absence of such a drape implies that the mapped MTDs are very late Holocene in age, and possibly less than several hundred years.

MTD facies occurs in close association with a lacustrine facies, which blankets most of the lake bottoms throughout all of the surveyed lakes (Fig.9 and 10; see Supplementary Material, Fig. S33 for larger version of the profiles). Internally, the lacustrine facies is transparent or contains weak, parallel, internal reflections. Its surface generally has a smooth curvilinear to flat topography (Figs. 9A–9C and 10A–10C). Where present together, MTD facies can be differentiated from lacustrine facies based on surface topography and deposit architecture, as exemplified in Figs. 7–10. Turbidite facies forming a transparent bed extending from MTD facies across lacustrine facies were observed in SAPs only locally within the study area, most obviously within survey area 22 (off Deshêtres Bay; Fig.4). The facies probably is more common, but is too thin to be discerned on the SAPs. Additional facies are present in the survey area that are grouped collectively as a glacial sediments-bedrock facies, as shown in Figs. 7 and 9.

Examples of mapped MTDs within selected survey areas are shown in Fig. 11 (see also Supplementary Material, Figs. S7–S10, S12 and S13, S19–S23, and S28 and S29). The extent of the mapped deposits is variable between survey areas, ranging from about 900 to 100 000 m². The largest deposit is about 800 m long and up to 165 m wide (survey area 22; off Deshêtres Bay; Figs. 4 and 11C) and the smallest is an isolated occurrence mapped in survey area 28 that easily could have been missed (Fox Pass; Figs. 4, 9,D, and S29). The larger MTDs commonly are composed of coalesced material originating from separate source areas. The latter are evident as distinct areas of slopes immediately overlying MTDs devoid of lacustrine facies (see Figs. 7B, 7C, and 7E–7H) separated by areas of intact lacustrine facies, as discerned from crossing SAP lines. The mapped MTDs are locally up to 13 m thick, being thickest where concentrated within narrow, elongated trenches oriented transversely to the source areas(s). The greatest number of separate mapped MTDs are within Tee Lake (survey area 7; Fig. 11A), which is located proximal to the 1935 CE earthquake epicentre (Fig. 2).

The mapped MTDs were generated by shallow failures, as is apparent by well-defined head scarps, up to several metres high, which mark the abrupt downslope edge of intact lacustrine deposits that are missing from the lower slope (see Figs. 7B, 7G, and 7H). The missing sediments clearly slid down a planar failure surface and now form an MTD located at the base of the slope (compare Figs. 7B and 9B). This failure mode is also evidenced on slopes lacking a head scarp, where lacustrine deposits are absent or thin and there is an MTD located at the base of the slope (see Figs. 7C, 7E, and 7F).

The composition of lateral equivalents to failed deposits was sampled at two piston coring sites, TeeLk-2023-05 and -09, both located upslope of well-defined head scarps perched above MTDs located at the base of the slope (e.g., see Figs. 6, 7B, and 11A; Supplementary Material, Table S3). Composite sediment logs, physical properties, and mosaics of CT-scan and photo images of the core sediments are shown in Figs. 12A and 12B. Both cores contained analogous upward sequences of varved glaciolacustrine, mixed silt-organic, and organic-rich (gyttja) deposits, albeit of differing thicknesses of the upper two units. Textural analysis of bulk samples of varved glaciolacustrine sediments indicates that the sediments fall within the fine, medium, or coarse silt size classes. The upward increase in organic matter content is readily apparent from the changes in dry mass percentage of 21%–569% and loss on ignition (LOI) of 0.2%–20.1% between the three units of both cores (Figs. 12A and 12B). Correspondingly, as better exemplified from the longer core TeeLk-2023-09 (Fig. 12A), sediment compressive strength decreases significantly upward from stiff glaciolacustrine sediments, ranging from 1.5 to >4.5 kPa, into a soft glaciolacustrine sediment bed, ∼5 cm thick (between about 1.5–1.45 m depth), and soft mixed silt-organic and gyttja units. Strength within the soft glaciolacustrine bed and the two overlying units are lower than could be measured with a Pocket Penetrometer (i.e., <∼ 0.5 kPa; Figs. 12A and 12B). The marked change in strength was obvious during coring at both sites, as the core sampler was easily pushed manually into the lake bottom until a stiff layer was encountered at the equivalent depth of glaciolacustrine deposit; additional penetration required pounding the sampler using a fence post driver.

Consistent with the composition of the two Tee Lake cores is core 2022-Kipawa-HB06 recovered using a gravity corer from the middle of an unfailed slope (see Fig. 6; Supplementary Material, Table S3). Shown on Fig. 12C, this core sample is composed of an analogous upward sequence of glaciolacustrine, mixed silt-organic sediment and gyttja units, and upward increase in LOI. All of the recovered sediments were soft, similar to the Tee Lakes core samples. Three additional gravity cores were collected to sample unfailed slope sediments on Kipawa Lake, but failed to penetrate into glaciolacustrine sediments (cores 2022-Kipawa-DB03, 2022-Kipawa-DB03A-1, and 2022-Kipawa-DB03A-2; Table S3;Fig. 6).

Based on the position of the abrupt decrease in sediment strength at the top of the glaciolacustrine deposits in the two Tee Lake cores (Figs. 12A and 12B), the failure plane is inferred to be located at or near the base of the soft sediments, within either the soft upper-most glaciolacustrine sediments or the mixed silt-organic unit, or at the boundary between them. The reported white, milky, or café au lait colours of water at Tee and parts of Kipawa lakes immediately after the 1935 CE earthquake (references cited above) indicate that a number of failures occurred within the top of glaciolacustrine sediments, which is the obvious source of the silt and clay clastic sediment. This inferred stratigraphic position of the failure plane and the composition of the overlying core sediments (see Fig. 12) indicate that gyttja, the thickest of the failed deposits, forms the majority of the MTDs.

Two radiocarbon dates provide chronology for the cored lacustrine deposits. These are 7257 ± 23 B.P. (8170–8010 year cal B.P.) and 8892 ± 22 B.P. (10 165–9905 year cal B.P.), yielded from detrital wood sampled from the mixed silt-organic units of cores 2022-Kipawa-HB06 and TeeLk-2023-05, respectively (Table 1; Figs. 12B and 12C). Both dates thus represent maximum ages for the respective sampling depths. Assuming the older date, 8892 ± 22 B.P. (10 165–9905 year cal B.P.), reflects the age of the encapsulating sediments, it represents the time by which Tee Lake and the higher Kipawa Lake basins had become isolated from glacial Lake Barlow–Ojibway (spanning 11.0–8.2 kyr cal B.P.; references cited previously) and an early postglacial depositional environment had been established. Both ages, plus the apparent lack of an unconformity within the overlying core deposits and within respective SAP lines, are consistent with the full sequence of overlying mixed silt-organic and gyttja deposits being preserved at both coring sites. On the basis of these two dates, and the lack of unconformities at these locations and within the SAPs elsewhere, preserved depositional sequences are also inferred for the intact lacustrine deposits blanketing the slopes throughout the surveyed areas, as exemplified in Figs.,7E and 9A–9F.

A third radiocarbon date is also listed in Table 1. This is 1430 ± 20 B.P. (1351–1298 year cal B.P.) yielded from pine needles subsampled at 44 cm depth, within the gyttja unit of core 2022-Kipawa-HB06 (Fig. 12C). Situated in the middle of combined mixed silt-organic and gyttja units (Fig. 12C), the age seems young for its depth location. Presumably, this reflects differential compaction during gyttja accumulation at the coring site, as indicated by the up-core increase in gyttja LOI in Fig. 12C.

To confirm the connection of the mapped MTDs to the 1935 CE earthquake, six gravity cores were collected at sites that either directly overlie or are located proximal to thick MTDs, as chosen from SAP lines. The coring sites were located within the Tee Lake (two), off Nine Mile Point (two), and off Deshêtres Bay (one) survey areas (Figs. 4 and 6; Supplementary Material, Figs. S7C, S22C, and S23C; Table S3); the sixth was located within the Hunter Lake area of Kipawa Lake over an MTD mapped by Doughty et al. (2010a; Fig. 6).

The recovered core samples contained organic-rich, gyttja deposits, 54–71 cm thick (Supplementary Material, Table S3), that exhibit faint bedding imparted by slight colour changes to the gyttja, as observed through the transparent core collection tubes (Fig. 13). Five of the six cores also contained a light-grey, fine silt layer, 0.25–1.5 cm thick, situated between 9 and 16 cm depth, as measured through the transparent sampling tube, that contrasts markedly against the under- and overlying gyttja deposits. The thickest silt layer, contained within core TeeLk-2023-01 #1, is shown in Fig. 13.

Radioisotopic analyses revealed “approximate” to “irregular, approximate” exponential decreases of ²¹⁰Pb activity as a function of depth in the upper 3.75–9.5 cm of each core, which is consistent with a regular rate of lacustrine gyttja sedimentation (see summaries of the laboratory reports, Supplementary Material, Tables S4–S9). Over each of these depth ranges, a linear regression age–depth model was applied that was used to calibrate a continuous rate of supply (CRS) model. The resulting CRS age–depth determinations were checked against the depth of the ¹³⁷Cs peak representing 1963 CE in the respective cores. If the age differences exceeded ± 5 years at the depth interval of the ¹³⁷Cs peak, then the CRS ages were adjusted by the difference. Adjustments of 9.8–24.1 years were made to the CRS ages of four of the six cores. The need for these adjustments is attributed to the loss of sediment at the gyttja–water interface during coring (see Supplementary Material, Tables S4–S6 and S9).

The resulting ²¹⁰Pb age–depth plots are shown in Fig. 14. The deepest and oldest age of each plot falls between 1938 and 1951 CE and represents an approximate maximum age for the start of the overlying gyttja accumulation. These ages also indicate that this gyttja sedimentation began approximately in the middle of the 20th century and, where present, after the deposition of the silt layers.

Immediately below the ²¹⁰Pb age–depth plots in each sediment core, the laboratory reports describe the deposits over intervals of 1.5–15 cm as being composed of different sourced sediments that are the product of irregular sedimentation processes or short-term depositional events, based on abrupt changes in bulk density and the shape of the ²¹⁰Pb curve (see Supplementary Material, Tables S4–S9). Deeper deposits in the sediment cores approach or are at ²¹⁰Pb background level, except for core Kipawa-2022-BB, where deposits between 19 and 34 cm depth again exhibit an exponential decrease in ²¹⁰Pb activity as a function of depth (see Supplementary Material, Table S8). The irregular sedimentation processes or short-term depositional events are inferred to reflect the presence of MTDs, as is consistent with the coring site locations. Since the MTDs are overlain by lacustrine sediments that started accumulating in the middle of the 20th century (Fig. 14), the MTDs are thus interpreted to have originated from failures triggered by the 1935 CE earthquake. Based on a stratigraphic position between the MTDs and the overlying lacustrine gyttja, the silt layers are interpreted to have originated from the accumulation of suspended fine silt clastic sediments associated with turbidites generated by the failures. While in suspension, these sediments caused the discolouration of waters of Tee and Kipawa lakes observed immediately after the 1935 CE earthquake.

Overall, the results of the six isotopic analyses confirm the interpretation that the MTDs at the coring sites are the result of failures triggered by the 1935 CE earthquake. This interpretation is inferred to apply to the very late Holocene MTDs mapped within the Témiscaming study area all of which similarly lack an overlying drape of lacustrine sediment in the SAPs. Where present overtop of an MTD, a distinctive light-grey silt layer in the upper roughly 10 cm of the lake bottom deposits demarcates 1935 CE MTD from overlying lacustrine sediment. The 1935 CE age of the silt layers also confirms the inference of Doig (1991) for the age of the stratigraphically youngest silt layers within his Tee Lake cores.

MTDs attributed to the 1935 CE earthquake were mapped at 17 of the 31 survey areas on 11 lakes, as shown in Fig. 15. The new data considerably augment the sites of Doughty et al. (2010a), also depicted in Fig. 15. A dashed polygon bounds the survey areas with mapped MTDs.

Relative to the centre of the mapped MTD locations (see Fig. 15), the new data significantly expand the MTD distribution to the east and south, and moderately to the west and northwest. In the northwest area, the most distant mapped MTD is at survey area 28 (near Fox Bay Pass; Fig. 15), which contains a single MTD occurrence that is the smallest mapped MTD within the study area (Figs. 9D, 10D, and S29). The mapped MTDs at the nearby survey areas 22 and 27 are sizeable and unambiguous (see Figs. 7D, 11C, and 11D). The northwestern position of the bounding polygon is constrained by six nearby survey areas lacking MTDs (23, 24, 25, 26, 29 and 30;Fig. 15).

The most westerly mapped MTD is within survey area 5, located at the southern end of Lake Timiskaming, adjacent to Town of Témiscaming (Fig. 15). Here, a single MTD is present that is the product of a failure from the eastern slope of the lake channel (Fig. 7G). The westerly MTD extent is constrained by the absence of MTDs at survey areas 1–4 located 12–21 km west of survey area 5. There were no suitable lakes to survey between survey areas 5 and 3–4. The western position of the bounding polygon is positioned toward the survey area 5 datapoint.

The most easterly MTD in Fig. 15 was along a single profile line (Figs. 9E and 10E) at the southern end of the north basin of survey area 16 (McKillop Lake). No MTDs were identified in survey areas 15 and 17 (Sheffield and Pant lakes) located to the south and north, respectively. There are no survey areas located further to the east, but the isolated MTD at McKillop Lake and the MTD absence at Sheffield and Pants lakes are considered to indicate that the MTD distribution is falling off, similar to what is observed at survey area 28 (Fox Bay Pass). The eastern position of the bounding polygon in Fig. 15 reflects this inference.

The survey area 8 (West Beauchêne Lake) contains the most southerly mapped MTDs (Fig. 15). Situated about 9.5 km southwest from the nearest mapped MTD (survey area 9; Petit Beauchêne Lake), the site is relatively isolated. Conducting additional surveys appreciably to the south and southeast, however, was hampered by a lack of lakes considered promising for surveying MTDs in combination with accessibility challenges (see below). The southern position of the bounding polygon is positioned just south of survey area 8, based on available data. The southeast position of the bounding polygon is an extrapolation between survey areas 9 and 13 (West Beauchêne Lake and Bois Franc Lake), but not the null result at survey area 11 (Otter Lake). The absence of MTDs at Otter Lake reflects a low susceptibility of the lake bottom slopes to failure (see below).

Overall, the bounding polygon in Fig. 15 is considered to be semi-constrained by surveyed areas that lacked MTDs. It encloses about 1270 km², which represents the known area affected by subaqueous failures triggered by the 1935 CE earthquake within the study area.

Assuming there was passable road access to and an available boat launch at a given lake of interest, a factor inhibiting accessibility to some parts of the study area was vehicle travel time, which is proportional to distance and affected by road quality (e.g., gravel vs. paved surface, smooth vs. washboard or potholed, public vs. private road, etc.). Extended travel times can reach a duration where the time available for surveying is too limited to collect data efficiently. Survey areas 16 and 17 (Fig. 15) on the eastern side of the study area represent the practical limit to surveying lakes by a day trip working out the Town of Témiscaming.

Excessive travel time hampered SAP surveying in the northeast, southeast, and east portions of the MTD distribution in Fig. 15. The problem is particularly apparent to the northeast, where there is an obvious gap in the surveyed areas. Access limitations probably contribute to the underestimation of the distribution area of MTDs, assuming reasonably similar susceptibility of lake bottom slopes to failure, which is not necessarily the case.

Conducting a quantitative analysis of slope stability is beyond the scope of this study, but basic factors controlling where MTDs are present or absent can be assessed qualitatively. As mentioned above, the survey areas were selected primarily based on water depths greater than about 30 m and slopes being “relatively steep” on available charts and maps from various sources. This approach successfully located MTDs within the initial couple of hours of data collection at survey areas 10, 12a, 12b, 13, 14, 18, 22, and 27 (Fig. 15). MTDs were eventually located at survey areas 8, 9, 10, 16, 19, 20, and 21 by progressively expanding the profiled areas (see Supplementary Material, Figs. S1–S31). The white circles at these latter survey areas represent sub-areas where MTDs were absent locally, demonstrating that MTDs do not occur ubiquitously within some basins/sub-basins.

Survey areas 15, 17, 23–26, and 29–30 lack mapped MTDs and are present at or outside the margin of the bounding polygon on Fig. 15. These sites all contain slope morphologies reasonably similar to the nearby survey areas containing mapped MTDs. It is thus inferred that the lack of MTDs within survey areas 15, 17, 23–26, and 29–30 reflects local shaking levels that were below the threshold of failure, as is reflected by the position of the bounding polygon in the northwestern and eastern part of the MTD distribution.

There are survey areas, however, where the absence of mapped MTDs probably reflects lower susceptibility of the slopes to failure rather than attenuated shaking per se. This is apparent at survey areas 6 and 11, both of which lack mapped MTDs, but are situated well within the bounding polygon (Fig. 15). Most obvious is survey area 6 (Gordon Lake) located near survey area 7 (Tee Lake), which contains multiple MTDs (Fig. 11A). The lack of MTDs at survey area 6 probably reflects relatively gentler slopes (less than 11°) than area 7 and, in particular, an apparent lack of noticeable gyttja deposits locally on the steeper slopes (see Fig 9A). The lack of MTDs at survey area 11 (Otter Lake) reflects generally short slopes (about 30 m long) and shallower slope angles (less than 6°) within what was a much shallower than expected basin (up to about 13 m deep), which was surveyed based on local advice. The lack of MTDs at both survey areas 6 and 11 is attributed primarily to a relatively low susceptibility of the lake bottom slopes to failure. The position of the bounding polygon in Fig. 15 is not influenced by the lack of MTDs at either survey area.

Survey areas 1–4 (Serene, McConnell, Spring, and Dymond lakes) are the most westerly surveyed basins/sub-basins within the study area (Fig. 15). All lack MTDs, which fits the general distribution of mapped MTDs and thus seems reflective of diminished shaking. Aside from survey area 1 (Serene Lake, up to about 25 m deep), which had relatively gentle slopes (less than 9°), the other three lakes are deeper (up to 35–60 m deep) and have topographies with highly variable slopes, but there is a general lack of gyttja sediments on the steeper slopes, as exemplified in Figs. 9F and 10F. The absence of MTDs at survey areas 2–4 probably more reflects a lower susceptibility of the slopes to failure than diminished levels of shaking per se. Thus, the western position of the bounding polygon (Fig. 15), which is defined by survey area 5, probably is influenced by the lower susceptibility of slopes to failure west of the Lake Timiskaming-Ottawa River.

Mid-way into this study, 4 × 4 m resolution digital bathymetry of the Lake Timiskaming became available from which the MTD at survey area 5 was recognized and subsequently profiled. The debris fields of seven additional MTDs were identified to the north of survey area 5, as shown in Fig. 15, but there was insufficient time to survey them. All seven sites are located toward the northwest and outside of the mapped distribution. The ages of the sites are not known. The authors are unaware of contemporary reports of turbidity changes to Lake Timiskaming immediately following the 1935 CE earthquake that might be attributed to these failures, but this does not preclude any having a 1935 CE age. Additional research is required to investigate the age of these MTDs as well as other MTDs identified further north within the Lake Timiskaming basin. Depending on these results, the west-northwestern portion of the mapped MTD distribution could be revised. The possible occurrence of additional MTDs in unsurveyed locations is an uncertainty inherent to a study of subaqueous landslides within an area of abundant lakes.

Researchers, such as Keefer (1984, 2002), Rodríguez et al. (1999), and Delgado et al. (2011), have used global inventories of primarily subaerial landslides to generate a series of empirical plots relating landslide distribution (area affected by landslides or maximum distance of landslides from the epicentre/rupture zone) to earthquake magnitude. The landslide data were also subdivided by geological material (rock-soil) and movement characteristics (disrupted slides and falls, coherent slides, lateral spreads and flows) to produce sub-plots. Upper bounds were fitted to the plots that represented the greatest area or distance likely to be affected by landslides for a given magnitude. Conversely, the upper bound curves also represent the minimum magnitude expected to generate a known area or distance affected by landslides in a paleoseismic investigation where paleoearthquake magnitude is not known. The plots most relevant to paleoseismic studies in central Canada are those that relate area affected by landslides versus magnitude because the location of a probable surface rupture typically is not known.

The area affected by landslides versus magnitude plot of Delgado et al. (2011) is shown in Fig. 16. It displays a dataset from the instrumental period of mostly subaerial landslides. Solid and dashed curves define different placements of the upper bound curves from Keefer (1984) and Rodríguez et al. (1999), respectively. The two diamond-shaped points labelled A and B on Fig. 16 represent Témiscaming study area data. Point A is defined by the 1270 km² MTD distribution area and M_w 6.1 magnitude of the 1935 CE earthquake. This point falls below, but close to, both upper bound curves, indicating that the scale of mapped MTD distribution is generally consistent with the global data, despite being derived from shallow, subaqueous failures involving mostly organic-rich gyttja deposits.

On Fig. 16, point B is where a horizontal line defined by an MTD distribution area of 1270 km² crosses the dashed upper bound curve. As extrapolated from the x-axis, the minimum magnitude of earthquake that can generate this area is between M_w/M_s 5.7–5.8, which is less than the instrumental M_w 6.1 magnitude of the 1935 CE earthquake. As a check and to account for the semi-constraint of the MTD distribution, as mentioned above, point B′ on Fig. 16 represents a high estimate of minimum magnitude of between M_w/M_s 5.9–6.0, as defined from the intersection of the upper bound curve and an area of 1270 km² plus 50% (i.e., 1905 km²). This expanded area is sufficient to encompass the seven uninvestigated MTDs located further north within the Lake Timiskaming basin (see Fig. 15) and the resulting estimated minimum magnitude is also less than the instrumental M_w 6.1 magnitude. Therefore, the estimate of minimum magnitude of M_w/M_s 5.7–5.8 based on MTD distribution area of 1270 km² is considered to be a reasonable estimation of minimum magnitude, if this application had been for a paleoseismic study. More broadly, and representing only a single example, this extrapolated magnitude suggests that an empirical plot of area of landsliding versus magnitude, compiled with mostly subaerial, global data, is satisfactory for estimating minimum magnitude of earthquakes in a study where the MTD distribution can be considered reasonably well mapped. However, an important factor influencing the underestimation of magnitude is that the mapped distribution of subaqueous MTDs is fundamentally controlled by the location, density, and morphology of lakes and, in areas where lakes are abundant, practical limits for data collection (e.g., lake accessibility and time availability). Conversely, subaerial landslides triggered by a modern earthquake can be mapped readily and comprehensively using remote sensing imagery (e.g., Sotiris et al. 2016).

There is one additional reported failure attributed to the 1935 CE earthquake that represents a far-field landslide relative to the mapped MTDs in Fig. 15. This is a subaerial failure of a railway embankment near Parent, Quebec (Fig. 1; 47.9°N, 74.6°W), located about 350 km north-northeast of the centre of the MTD distribution. Based on a site visit, Hodgson (1935, 1936, 1945) described the failure as more than 200 ft (about 61 m) long that occurred where a stream was undermining the base of the embankment slope with the erosion causing fill to slide into a small lake. He also mentioned that “there were many rock cuts [in the area] and in some of them the rocks were on the point of moving and could be moved by hand yet did not move with the earthquake (Hodgson 1935, p. 8).” On the basis of the lack of rock movement in local rock cuts and the about 350 km distance to the epicentre, Hodgson considered that the embankment slope to be at the “point of slipping” at the time of the earthquake.

A goal of the present study was to utilize the mapped distribution of MTDs from a paleoseismic perspective and estimate a “paleo”-magnitude using the area affected by landslides, as is done above, based exclusively on the mapped MTDs. From a prehistorical context, however, it is difficult to correlate a Parent-like failure to the MTD records in the Témiscaming study area with any confidence, even having high-precision dating control, because of the lack of failures within the about 350 km distance between the locations. Hence, the Parent embankment failure was not incorporated into the area affected by landslides, as presented above.

Nevertheless, it is useful to consider the effect of including the Parent failure into the estimation of minimum magnitude, as can be demonstrated using a plot from Delgado et al. (2011, his fig. 2) that relates maximum distance to epicentre/surface rupture versus magnitude for disrupted landslides during the instrumental period. Using the 350 km distance between Parent and the centre of the MTD distribution, the extrapolated minimum magnitude falls between M_w/M_s 7.8–7.9. This far exceeds the instrumental M_w 6.1 magnitude of the 1935 CE earthquake and is even greater than the M_w 7–7.5 1663 CE Charlevoix earthquake, the largest historical and currently-known postglacial earthquake in central Canada (see Brooks and Perret 2023). The high estimate may reflect the Parent embankment literally being at the point of slipping at the time of the earthquake, as Hodgson (1935, 1936, 1945) reasoned. Alternatively, it is also possible that there is no cause and effect between the earthquake and the failure and that the timings were coincidental.

Although an extreme example, the Parent embankment failure highlights the possible scale of overestimation arising from including a far-field landslide (or landslides) into an estimation of minimum magnitude using a distribution of landslides. Far-field, earthquake-triggered landslides occur at large distances from an earthquake epicentre/surface rupture and generate outlier points on empirical area/distance versus magnitude plots (see Fig. 16). These landslides occur for a single or combination of reasons, including (i) soil/rock having been weakened by a recent, previous earthquakes; (ii) a combination of soil/rock prone to instability from seismic shaking, in steep, rugged terrain, where the slopes are undercut by erosion or road cuts; (iii) significant rainfall before or during the earthquake; and (v) amplification of shaking from site effects (Delgado et al. 2011). Of these, (ii) best applies to the Parent embankment failure, based on the description by Hodgson (1935, 1936, 1945).

It is apparent that paleoseismic investigations focused on mapping common-aged landslides on the Canadian Shield need to consider the possible presence of a far-field landslide(s) in the dataset when estimating a minimum magnitude. Factors to be considered for recognizing a far-field landslide(s) should include the number of mapped landslides, the relative spatial density of the landslides, chronological uncertainty in correlating landslides, the range of landslide types and settings, and importantly, comparison of the estimated minimum magnitude relative to the largest historical earthquakes regionally. The last point is clearly exemplified by the estimated minimum magnitude determined using the Parent embankment failure that greatly exceeds the magnitude of the actual triggering 1935 CE earthquake. From a paleoseismic perspective, the great physical separation between an isolated failure and the mapped MTD distribution is a red flag that an estimate of a minimum magnitude for a triggering earthquake may be overestimated. In such cases, the estimate of minimum magnitude should be made both using and not using the suspected far-field datapoint(s) with the respective estimates being appropriately assessed.

The centre of the mapped MTD distribution is located about 15 km northeast of the epicentre of the 1935 CE earthquake (Fig. 15). In a paleoseismic study where there is no rupture surface or obvious nearby active fault, this centre point provides an obvious basis for estimating the epicentre location. However, the location is a function of various factors affecting the MTD distribution, e.g., the spatial pattern of shaking, variable site effects, differing susceptibility of lake bottom slopes to failure, and the completeness of the mass movement dataset, rather than being strictly controlled by the location of the epicentre per se. The centre of a mapped MTD distribution from a paleoseismic perspective, therefore, should be considered as an imprecise approximation of the epicentre location.

Older, buried MTDs are present at survey areas 7 and 29 (Tee Lake and Kipawa Channel, respectively; Fig. 15). At Tee Lake, the buried deposit is located within the western basin opposite the mouth of the east arm (Fig. 11A). It covers about 20 000 m², is up to 4 m thick, and is composed of a transparent to weakly chaotic facies that is inferred to be an MTD. This deposit is overlain by MTD facies attributed to the 1935 CE earthquake, but locally also by lacustrine facies up to about 4.5 m thick, as shown on Figs. 9G and 10G. As assessed from several crossing SAP lines, the overlying lacustrine facies thickens gradually down the lowest portion of adjacent slopes, which did not fail in 1935 CE. The facies then extends directly across the top of the buried MTD, eventually becoming disturbed by the 1935 CE MTD (Figs. 9G and 10G). This architecture is consistent with the vertical sequence of lacustrine facies being intact (i.e., complete) and thus it probably began to accumulate at the start of the lacustrine phase of sedimentation in Tee Lake basin. The lacustrine facies therefore postdates the buried MTDs.

Within the Kipawa Channel survey area, a transparent to weakly chaotic MTD facies, up to 11 m thick and covering about 50 000 m², is overlain by lacustrine facies, up to about 3 m thick, within a sub-basin that does not contain late Holocene MTDs (Figs. 8H and S30). Here, the lacustrine facies is similar in thickness to the deposits that extend southward along the axis of the sub-basin and beyond both the location of the buried MTD and the apparent source of the failed sediment. Similar to Tee Lake, this overlying lacustrine facies seems intact and also probably began accumulating at the start of the lacustrine phase of sedimentation. It also postdates the buried MTD.

The authors are aware of a probable third location of buried MTDs within the Hunter Lake basin of Kipawa Lake, just west of the coring sites I and J in Fig. 6 (M. Doughty, unpublished data). Here, two occurrences of buried, inferred MTDs are situated about 95 m apart within a single SAP line and may be separate occurrences of the same deposit. Both pinch out at approximately the interface of the glaciolacustrine and lacustrine facies and were deposited at about the time of the associated transition in sedimentation.

At all three locations, an estimated minimum age for the start of lacustrine facies accumulation, and thus the minimum age of the buried MTDs, is roughly 8.0 ka cal B.P., based on the calibrated radiocarbon age (8170-8010 year cal B.P.) of wood situated near the base of the lacustrine deposits in Core 2022-Kipawa-HB06 (Table 1; Fig. 12C). This age postdates the demise of glacial Lake Barlow–Ojibway and fits the regional chronology (see Dyke 2004; Brouard et al. 2021). The three locations of buried MTDs are about 15–27 km apart, but there are insufficient data to connect them temporally or stratigraphically, except in a general sense. The notion that these MTD occurrences individually or collectively are evidence of a prehistoric earthquake cannot be substantiated. The only MTD signature within the study area unequivocally linked to seismic shaking is that generated by the 1935 CE earthquake.

In contrast to the buried MTD evidence presented above, Doig (1991) reported coring results from the west basin of Tee Lake that he interpreted to be evidence of two paleoearthquakes occurring about 1500 and 400 years ago. Key to his interpretations are deposits associated with the 1935 CE earthquake that he uses as an analogue for the older events. He describes the evidence of the 1935 CE earthquake as a silt layer, 1–2 cm thick, in combination with an overlying bed of chaotic gyttja and silt fragments, 15–22 cm thick, contained near the top of three core samples collected from sites overlying areas of “redeposition” (that presumably are MTD sites based on the locations; Doig 1991). The silt layer is equivalent to that shown in Fig. 13 and probably is also part of a turbidite. The connection of the overlying chaotic gyttja and silt fragments to the 1935 CE earthquake, however, is not supported by the ²¹⁰Pb profile from Tee Lake that overlies the silt layer in Fig. 14A.

Nevertheless, Doig’s (1991) evidence of the 1500-year-old event is a deeper silt bed, about 6 cm thick, located just above the base of one of the aforementioned cores, as well as silt layers and overlying beds of chaotic gyttja and silt fragments in core samples collected from two undisturbed gyttja sites on basin side slopes. He considered that these deposits may be evidence of a paleoearthquake similar to or larger than the 1935 CE earthquake, based on the greater thicknesses of the older silt beds and the two side slope locations being unaffected by sedimentation from the 1935 CE earthquake. Evidence of the 400-year-old event is inferred from silt layers, 1–2 mm thick, situated at intermediate depths in core samples from the three resedimented coring sites. He considered these thin layers to correlate and be evidence of a paleoearthquake of either lesser intensity than the 1935 CE earthquake or one located more distantly.

A key factor for interpreting a seismic trigger for a subaqueous mass movement or turbidite is establishing synchronicity between it and at least several MTDs that occur at common stratigraphic level (or event horizon) within a lake sub-bottom (e.g., Schnellmann et al. 2002). Greater confidence of the seismic interpretation can be obtained from higher numbers of MTDs within the event horizon and especially when the correlation can be made between well-spaced sub-basins of a lake or separate lakes (Praet et al. 2017). The interpretations by Doig (1991) that the two deeper silt layers each represent an older paleoearthquake are not supported by synchronous MTD deposits from the Tee Lake, and neither this study or Doughty et al. (2010a) recognized older MTDs interbedded within nearby areas of Kipawa Lake that could be the geomorphic signature of either event. Available evidence, therefore, indicates that the paleoearthquake interpretations from Tee Lake should be considered as uncertain, especially the inference of the older layer representing a similar or larger earthquake than the 1935 CE event.

The 1935 CE MTD signature thus seems to be the only unequivocal evidence of a significant earthquake within the Témiscaming study area since roughly 8 ka cal B.P. The presence of the three occurrences of older MTDs demonstrates that lake bottom deposits experienced failures in the distant past, suggesting that the lack of an analogous, older MTD signature(s) is indicative of the absence of an equivalent shaking event(s) to the 1935 CE earthquake over this period, as also concluded by Doughty et al. (2010a). However, it is possible that MTDs from the 1935 CE earthquake buried and thus overprinted MTDs of an older signature(s) making them unrecognizable in the SAP lines. Overprinting of MTDs was identified as an issue in a study by Daxer et al. (2022), but as hampering mapping individual landslides in certain parts of an Austrian alpine lake, resulting in the under- or overestimation of the number of individual failures and event horizon volume. Overprinting was not recognized as completely masking an event horizon signature. For complete overprinting to have happened within the Témiscaming study area, every older MTD must have been located within the distribution of the 1935 CE MTDs (see Fig. 15) and situated at sites that failed in 1935 CE at sufficient scale to obscure the older deposits, all of which seems unlikely, except for a localized, minor signature.

On the other hand, the relevant time period could be shorter than 8 ka cal B.P. The failures triggered by the 1935 CE earthquake involved primarily gyttja deposits that are the product of lacustrine sedimentation. The gyttja, thus, thickened and would have become more prone to failure over time, other factors being equal (i.e., slope angle, sediment strength, length of slope). The time period relevant to the absence of an analogous MTD signature, therefore, may instead begin when gyttja deposits became thick enough to be prone to failure from a seismic event equivalent to the 1935 CE earthquake. This is similar in concept to the substantially diminished or possibly a missing signature associated with the younger of two earthquakes of comparable magnitude, when there has been insufficient accumulation of sediment on slopes to be unstable during the second event (e.g., Moernaut et al. 2007, 2017). In the present case, this relates to a single phase build up of gyttja sediment on slopes that did not catastrophically fail until 1935 CE. How much this time period might be shorter than 8 ka cal B.P. is not known, but warrants further investigation. Nevertheless, the lack of an older MTD signature is consistent with the 1935 CE earthquake being the largest to occur within the study area likely for a significant portion of the post 8 ka cal B.P. time period.

Late Holocene MTDs are common in the basins and sub-basins of Kipawa Lake and a number of smaller, surrounding lakes, near Témiscaming, Quebec. The MTDs are the product of subaqueous, shallow slope failures, up to several metres thick, that occurred within the postglacial lacustrine sediments. Core samples from Tee Lake indicate that unfailed deposits on lake bottom slopes consist of an upward sequence of glaciolacustrine, mixed silt-organic, and organic-rich (gyttja) deposits. The failure planes were probably within soft sediments present within the upper-most part of glaciolacustrine deposits, at the base of overlying lacustrine deposits, or at the boundary between them.

Radioisotope analyses of sediment samples from two cores recovered from Tee Lake and four from selected locations within Kipawa Lake confirm that the MTDs at these sites are the result of failures triggered by the 1935 CE earthquake. This association is inferred to apply to all of the late Holocene MTDs mapped within the Témiscaming study area.

MTDs triggered by the 1935 CE earthquake were mapped at 17 of the 31 survey areas on 11 lakes, considerably expanding the distribution of MTDs reported by previous studies. The MTD distribution encompasses about 1270 km², but is considered to be semi-constrained by survey areas that lacked MTDs.

A magnitude of M_w/M_s 5.7–5.8 was extrapolated from a published empirical plot relating area affected by primarily subaerial landslides versus magnitude and represents the minimum magnitude for an earthquake that can generate a landslide distribution of 1270 km². This magnitude is less than the instrumental M_w 6.1 magnitude of the 1935 CE earthquake and is deemed a reasonable estimate of minimum earthquake magnitude, if taken in a paleoseismic context. On the same empirical plot, the position of a datapoint defined by the 1270 km² area and M_w 6.1 magnitude indicates that the extent of mapped, subaqueous MTDs is generally consistent with a global dataset composed of primarily subaerial landslides.

An embankment failure, near Parent, Quebec, occurred about 350 km from the Témiscaming study area and is reported to have been triggered by the 1935 CE earthquake. An extrapolated estimate of minimum magnitude based on this distance far exceeds the instrumental M_w 6.1 magnitude of the 1935 CE earthquake. Although an extreme example, the Parent embankment failure demonstrates that paleoseismic studies using earthquake-triggered landslides and empirical plots to estimate minimum earthquake magnitudes need to consider the possible inclusion of far-field landslide(s) in the compiled landslide dataset. In such cases, the estimate of minimum magnitude should be made both using and not using the suspected far-field datapoint(s) and with the respective estimates appropriately assessed.

The distribution of MTDs triggered by the 1935 CE earthquake represents the only such signature within the Témiscaming study area since roughly 8 ka cal B.P. The lack of an analogous, older MTD signature(s) is consistent with the absence of an equivalent shaking event(s) to the 1935 CE earthquake over this timespan. The relevant period may be shorter, however, and begin when gyttja deposits within the study area became thick enough to be prone to failure from such an event. The significance of this difference is not clear, but warrants further investigation.

J. Adams (Canadian Hazards Information Service), D. Perret and A. Pugin (GSC), two anonymous journal reviewers, and the associate editor provided helpful comments that improved the paper. We thank GSC co-workers J. Zheng (assistance coring and sub-bottom profiling), T. Cartwright (assistance coring), and A. Pugin and B. Dietiker (advice and assistance with geophysical software). We also thank students K. Loker-Fulcher (assistance collecting sub-bottom acoustic profiles, July–August 2019), and D. Levesque and R. Kennedy (assistance processing SAP data in 2023). The cooperation of Kebaowek First Nations during the undertaking of this study is greatly appreciated. We also thank La Réserve Beauchêne Wilderness Lodge for access to West Beauchêne and Petit Beauchêne lakes, numerous cottage owners for use of private boat launches, and numerous outfitters, cottage owners, and fishermen for advice on local lakes. M. Doughty and N. Eyles kindly provided data from the Hunter Lake portion of their 2007 Kipawa Lake SAP survey. Canadian Hydrographic Service provided digital bathymetry of Lake Timiskaming. Work within the Blue Lake End Moraine (Dymond Lake) and Spring/Cut Lake Esker (Spring Lake) conservation reserves was authorized by Ontario Parks, Ministry of the Environment, Conservation and Parks. This research was supported by Public Safety Geoscience Program, Lands and Minerals Sector, Natural Resources Canada, and the Nuclear Waste Management Organization.

Data can be made available on request.

Conceptualization: GRB, AG, KB

Formal analysis: GRB

Funding acquisition: GRB

Investigation: GRB, AG, KB

Methodology: GRB, AG, KB

Project administration: GRB

Resources: GRB, KB

Software: GRB

Supervision: GRB

Validation: GRB

Visualization: GRB

Writing – original draft: GRB

Writing – review & editing: AG, KB

Supplementary data are available with the article at doi:10.1139/cjes-2024-0080.