Abstract
As the largest body of water on the northern Great Plains of North America, Lake Winnipeg in central Manitoba, Canada, is crucial to the region’s hydrology, economy, and society. Previous research identified exposed subfossil stumps at several locations along the shore in both the lake’s north and south basins, and interpreted them as evidence of low lake levels c.
Introduction
As the largest body of water on the northern Great Plains of North America, Lake Winnipeg in central Manitoba, Canada (Figure 1a), is crucial to the region’s hydrology, economy, and society (Wassenaar and Rao, 2012). Draining more than 1 million km2 between Lake Superior and the eastern slopes of the Canadian Rockies, the lake receives inflows from the Winnipeg River (which mainly encompasses territory in northwestern Ontario), the Saskatchewan River (southern Alberta and central Saskatchewan), the Red River (southern Manitoba, southeastern Saskatchewan, and parts of Minnesota and the Dakotas), and several other smaller rivers. These tributaries provide water to more than 5.5 million people in Canada and the United States, and are the main source used by several major cities, including Calgary, Edmonton, Winnipeg, Fargo-Moorehead, and Grand Forks. Lake Winnipeg also functions as the largest reservoir in the province of Manitoba’s hydroelectric generating system, and its outlet, the Nelson River, powers several dams that collectively produce more than 4000 MW of electricity (Manitoba Hydro, 2013).

Maps showing (a) Lake Winnipeg and its environments, as well as (b and c) the location of the study site near the Spider Islands. In panel (c), the black circles mark the position of submerged tree stumps collected along the lake’s foreshore in 2012.
Because the lake is large, shallow, and has a relatively short residence time (4.3 years; Lévesque and Page, 2011), the amount of water in Lake Winnipeg can vary substantially from one year to the next. Since systematic measurements began in 1913, its surface elevation has fluctuated by more than 2.7 m and the lake has experienced several low stands lasting 8–26 months (Supplementary Figure S1, available online). The most extreme low-water conditions during the period of record were observed in June 1941, when the mean monthly elevation of the lake fell to 216.2 m a.s.l. Because depressed lake levels reduce the amount of hydroelectric power produced downstream, this episode of exceptionally low-water is used in scenario planning to test the resilience of the provincial energy system to severe and sustained drought (Rangarajan et al., 1999; Turanli, 2002). More broadly, because its watershed collects runoff from nearly the entire northern Great Plains, the level of Lake Winnipeg effectively functions as a surrogate measure of the region’s overall hydrological budget. In that respect, low lake stands can be used to gauge the impact of widespread drought on the total water resource shared among communities in Canada and the United States.
In other settings, the geochronological dating of drowned forests has proven to be an effective method to determine the timing and magnitude of past low stands prior to the onset of direct hydrological monitoring (e.g. Lewis et al., 2008; Shanahan et al., 2009; Stine, 1994). Nielsen (1998) reported subfossil tree stumps along the foreshore of Lake Winnipeg at six locations in the lake’s north and south basins. Radiocarbon dating of nine stumps produced ages ranging between 160 and 390 yr 14C BP, which led him to suggest the level of Lake Winnipeg was lower c.
Here, we report radiocarbon dates for a new set of submerged stumps recovered from the foreshore of Lake Winnipeg near the Spider Islands, which was the site closest to the lake’s outlet sampled by Nielsen (1998). We present evidence that subfossil trees currently exposed at this site are considerably older than the ages reported by the previous study and, as a result, argue they cannot be the evidence of depressed water levels caused by hydrological factors, such as severe drought, during the last millennium. Instead, we suggest these exposed stumps are the product of a gradual increase in the level of Lake Winnipeg and onshore transgression caused by isostasy.
Study site and methods
Located within the former basin of Glacial Lake Agassiz, Lake Winnipeg is shallow (it has an average depth of 12 m in the south and 16–17 m in the north; Todd et al., 1997), and oriented about a north-northwest/south-southeast long axis. Lake Winnipeg established roughly 7.5 cal. kyr BP as a series of independent lakes before post-glacial tilting due to isostatic rebound led to their expansion and coalescence at 2.9 cal. kyr BP (Lewis et al., 2001). At present, this part of North America is still responding to the removal of surface load associated with the most recent advance of the Laurentide ice sheet, and the lake’s outlet is rising more rapidly than its two main basins (Sella et al., 2007; Tackman et al., 1999). Lake levels in the south basin are rising relative to the north basin and the lake is slowly expanding, particularly along its southern shore (Department of Mines, Resources and Environmental Management, 1977). Since 1976, the level of Lake Winnipeg has been controlled by two dams and three diversion channels located near its outlet to the Nelson River. These facilities reduce the amplitude of the annual cycle in Lake Winnipeg’s water level and increase outflow from the lake during fall and winter so dams on the Nelson can satisfy the elevated seasonal demand for electricity.
Our study site is located in the northeastern quadrant of Lake Winnipeg, directly east of the Spider Islands archipelago (Figure 1a). At this position, the shoreline is made up of two narrow (20 m wide) barrier beaches that separate the lake from a 2.5 km2 lagoon (Figure 1b). The closest lake gauge is sited approximately 16 km north-northwest of the Spider Islands at Montreal Point. In 1996, Dr Nielsen found more than two dozen stumps exposed along the foreshore of the northwest facing beach (Figure 1b and Supplementary Figure S2, available online). At this location, the lake’s bottom is composed of sand out to a depth of 5 m; farther offshore, it grades into eroded Lake Agassiz deposits with scattered outcrops of Precambrian bedrock (Forbes, 2000). The sand used to construct the barrier beaches is thought to derive mainly from ice-marginal glacial deposits and Lake Agassiz sands eroded from the bottom of Lake Winnipeg (Forbes, 2000). Aerial photographs of this site archived at the Canadian National Air Photo Library show no obvious changes in the position or structure of the barrier beaches occurred between c. 1970 and present (Dr. Stephen Wolfe, personal communication).
We conducted an initial reconnaissance visit on 21 August 2011, when the lake was 0.6 m higher compared to 1996 (Water Survey of Canada, 2013). No stumps were subaerially exposed, and the beach was blanketed by a ~30-cm layer of peat that washed ashore some time after 1996. Tree fragments were found submerged offshore under 0.5–1 m of water, and samples were collected from four stumps that appeared to be rooted in the sandy lake bottom (Figure 2a and b). When the site was visited again on 21 August the following year, the lake had fallen by 0.6 m (returning to approximately the same elevation as 1996), the peat deposit had been removed by wave action, and several new stumps were subaerially exposed (Figure 2c and d). We sampled seven stumps rooted underwater, apparently in growth position, along a 110-m transect parallel to the shoreline (Figure 1c). All specimens were solid wood with no evident rot, had smooth surfaces without bark, and were dark brown in color. The largest samples included the root ball and portions of the lower stem(s).

(a and b) Subfossil stumps (respectively, SPIDER11-1 and SPIDER11-5) collected from the Spider Islands site on 21 August 2011. (c and d) Submerged tree stumps along the foreshore of Lake Winnipeg near the Spider Islands on 21 August 2012.
Wood samples from the outer portion of each stump were radiocarbon dated using the accelerator mass spectrometry method by Beta Analytic Incorporated (Miami, Florida). The outermost rings of each specimen were removed so the material submitted for dating did not include the stump’s exposed surface. Prior to analysis, samples were cut into small (3–5 mm) fragments and washed with an acid–alkali–acid pretreatment to remove carbonates and organic acids. Calibration was performed using OxCal 4.2.2 (Bronk Ramsey and Lee, 2013) and the IntCal13 atmospheric calibration curve (Reimer et al., 2013). Because none of the specimens had more than 30 growth rings, we did not attempt to date them using dendrochronology.
Results and discussion
Wood samples from SPI-12-01 and SPI-12-06 exhibited piceoid cross-field pitting, fusiform rays on the tangential surface, and an abrupt earlywood-latewood transition, indicating that these trees were Larix (Hoadley, 1990), the same genera reported at the site by Nielsen (1998). The new set of 11 stumps recovered near the Spider Islands had calibrated ages ranging between 2880 and 4150 cal. yr BP (Table 1 and Figure 3). We did not find any clear association between the age of the stumps and their distance from the shoreline. The stumps farthest offshore (SPI-12-01, SPI-12-02, and SPI-12- 03) may be slightly older than those closer to the beach (SPI-12-04, SPI-12-06, and SPI-12-07), but there was still considerable overlap between the calibrated ranges of those specimens. The most recent date in the entire set was obtained for SPI-12-05, which was located between stumps that are probably 400–500 years older. The oldest date was obtained for specimen SPIDER11-2, but this result should be interpreted with caution; this specimen did not include any portion of the root ball and may not have been preserved in growth position. As a whole, these new dates indicate the subfossil stumps currently exposed at this location are 3–4 kyr old, which places them at the approximate boundary between the mid- and late Holocene (Walker et al., 2012). The calibrated ages of most stumps (except for SPIDER11-2 and SPI-12-05) have overlapping 2 σ ranges, meaning that they could have died at the same time (and potentially, from the same cause), but it is more likely the establishment and death dates of these trees occurred at different times over a 500-year period.
Radiocarbon ages of submerged stumps rooted in the foreshore of Lake Winnipeg.
Errors are ±1σ.
Determined from the calibration dataset IntCal13 (Reimer et al., 2013) using the program OxCal 4.2.3 (Bronk Ramsey and Lee, 2013). Age ranges are ±2σ.

Calibrated radiocarbon dates for submerged tree stumps recovered from the foreshore of Lake Winnipeg. The shading represents the probability distribution for each date, and the gray lines beneath illustrate the 95% probability ranges. Calibration performed using OxCal 4.2.3 (Bronk Ramsey and Lee, 2013) and the IntCal13 atmospheric calibration curve (Reimer et al., 2013).
Dates obtained for the first subfossil stumps recovered from the Spider Islands site indicated those trees died between 390 ± 70 and 255 ± 75 14C yr (305–529 cal. yr BP and −4–497 cal. yr BP, respectively; Nielsen, 1998). Noting that modern Larix laricina grow at least 0.5 m above the water table, Nielsen (1998) argued the location of these stumps meant the lake had risen by 20 cm/century at that location (and others) since approximately
This scenario cannot be valid if the subfossil stumps exposed at the Spider Island site are instead several millennia old. According to the uplift tilting model developed by Lewis et al. (2001), the shoreline of Lake Winnipeg was several hundreds of meters away from this location at that time. Pollen records indicate the region surrounding the lake’s north basin has become progressively wetter from 4 cal. kyr to present (Lewis et al., 2001; Ritchie, 1976), which means an additional decline in lake level caused by prolonged aridity in the lake’s watershed could have coincided with the lifespans of these trees, but that type of climatic shift is not required to account for either the estimated change in shoreline position or the observed tree mortality at our site. Our set of new dates implies that the establishment and death of these trees were not connected to Lake Winnipeg, and instead, these trees grew, died, and were preserved within a forest hundreds of meters inland. Under this alternative scenario, the lake has gradually transgressed shoreward toward the Spider Islands site over the last 3–4 cal. kyr as inferred by Forbes (2000) and is now exposing and eroding subfossil trees. If wave action and lake ice are acting to remove these stumps from their foreshore position after they become exposed, those processes could explain why so many stumps were present in 1996 while so few were visible in 2011 and 2012.
Lake Winnipeg is located near the hinge point separating glacial isostatic uplift from subsidence in central North America (Sella et al., 2007) and, as a result, its outlet is rising more rapidly than its northern or southern basins and the lake is transgressing south-westward. The influence of post-glacial tilting is detectable in lake-level gauge records (Tackman et al., 1999), but, over the last century, the water level of Lake Winnipeg has been primarily controlled by hydrological variability (and, since 1972, by the twin dams on its outlet). However, over the last several millennia, isostasy has been the dominant influence on both the configuration of the Lake Winnipeg basin and the elevation of the lake’s surface (Lewis et al., 2001). Studies of other large lakes in high-rebound regions have also shown the relative importance of these two factors to depend on the timescale considered. For example, over the last 1200 years, strandlines along Lake Michigan correlate with surface moisture proxies such as tree rings and peat records, suggesting that lake level fluctuations during that interval were driven primarily by hydrological variability (Argyilan et al., 2010). In contrast, over the last 5 kyr, the level of Lake Michigan has fallen by more than 10 m, and more than 60% of that change is due to vertical movement of its basin (Baedke and Thompson, 2000). In situations like these, where isostasy has had a significant cumulative effect on lake elevation and the position of the lake shore, mid-Holocene age material (or older) near the modern shoreline may not be related to the lake at the time of its deposition (or in the case of trees, their establishment) and would not constitute evidence of paleo-lake levels.
Conclusion
By virtue of its status as the largest natural reservoir within the northern Great Plains of North America, Lake Winnipeg provides a unique perspective on the hydrological resource shared between three American states, four Canadian provinces, and 5.5 million people. If it were possible to estimate the magnitude and timing of past low stands, that information could be used to clarify how severe, system-wide drought can affect total runoff across the northern Great Plains and develop regional planning scenarios based on real droughts that occurred in the past. Nielsen’s (1998) report of in situ stumps rooted below the modern water line provided the first indication that geochronological dating of drowned forests might be a viable approach to understand how Lake Winnipeg behaved prior to the onset of direct hydrological monitoring. Moreover, because stumps at the Spider Islands site and other shoreline locations appeared to be less than 500 years old, they were young enough to be plausibly connected to a climatically driven low stand that occurred during the last few hundred years. Having conducted an expanded survey of subfossil stumps near the Spider Islands, we found the relic forest exposed at this site to be substantially older than indicated by that earlier study. We are not certain why our age estimates are so different from those reported by Nielsen (1998), but if these new dates are correct, they argue against a simple hydrological explanation for these submerged trees. Instead, our results suggest these trees died 3–4 kyr ago and are now exposed because of gradual, isostatically driven changes in the basin configuration and shoreline position of Lake Winnipeg. Because they date to the mid- or late Holocene, we conclude these subfossil stumps do not constitute clear evidence of hydrologically caused low lake stands in Lake Winnipeg or widespread drought on the northern Great Plains.
Footnotes
Acknowledgements
We thank Dr Erik Nielsen for sharing photographs of his 1996 field collections, accompanying us to the field site in 2011, and for providing the spark for many productive discussions of Lake Winnipeg. We are also grateful to Blue Water Aviation for providing transport to the study site. Wood identification was conducted by Dr Robert Seavey at the University of Minnesota. The final version of this manuscript was improved by comments from Dr Stephen Wolfe, Dr Harvey Thorleifson, Dr Joe Mason, and two anonymous referees.
Funding
This work was supported by the University of Minnesota’s Office of the Vice President for Research through its Grant-In-Aid program.
