Our results potentially extend evidence for the beginning of North American occupation by B. cederströmii from the 1980s to the early part of the twentieth century in four lakes, and much earlier in one of those lakes (Three Mile Lake, Ontario). They raise significant questions about the origin and subsequent range expansion of B. cederströmii on the continent and demonstrate the presence of protracted periods of time without detection by traditional plankton net tows or any other means (e.g., ensnared on angling gear). If our results are valid, they demand rethinking about the invasion dynamics of B. cederströmii and our methods of study. If our results are artifactual, they imply that the paleolimnological approaches that we used, otherwise widely adopted for studies of many species of crustacean zooplankton where large numbers of specimens are identified, are potentially unsuitable for the study of species invasion (such as by B. cederströmii), for which only presence and absence data are relied upon. We acknowledge that in typical paleolimnological investigations, the finding of one or two specimens would never be considered significant; however, when investigating a potential invader, such data (if reliable) are clearly important.
The prevailing view regarding continental invasion of B. cederströmii is that it became established in North America during the late 1970s to early 1980s when propagules arrived in the ballast water of ocean-going cargo ships (Sprules et al. 1990). This scenario, however, has never been confirmed with direct evidence. The purported timeframe is based on analyses of plankton-net assemblages and fish-stomach contents collected during the 1980s from the Great Lakes, which provided the first evidence of the species on the continent (Johannsson et al. 1991; Bur et al. 1986). The inferred mechanism of transport in ship ballast water is based on the timing of the opening of the St. Lawrence Seaway in 1959, which accelerated transport of ballast water from foreign cargo ships into the Great Lakes (Mills et al. 1993; Ricciardi 2006; Sturtevant et al. 2019). But shipping between the Great Lakes and the Atlantic Ocean was already underway long before 1959 and therefore it is possible that ballast water containing B. cederströmii could have been released into the Great Lakes (Mills et al. 1993) up to a about century earlier, following completion of the first Welland Canal around Niagara Falls (1829) and the first St. Lawrence River canal system (1847) (LesStrang 1981). By one estimate, ballast water is credited with the arrival of 6 non-native species in the Great Lakes before 1956 (Ricciardi 2006). This nuanced history of Great Lakes shipping, together with our sediment evidence, suggests that the origins of B. cederströmii in North America could have occurred through ballast water discharge, but at a much earlier initial date than currently described in the literature.
If our results are valid, they raise questions regarding the timing and sequence of lake invasions during continental range expansion. If B. cederströmii first arrived on the continent in ship ballast water, it likely first established population centers in the Great Lakes and then subsequently spread inland, possibly associated with recreational boating activities (MacIsaac et al. 2004). This scenario predicts that Great Lakes sediments should house chronological records that establish their position as a continental beachhead. However, uncovering this type of evidence seems unlikely for two reasons. First, a published study on the paleolimnology of B. cederströmii from Lake Erie reports that the oldest exoskeletal remains there date to near 1980 (Keilty 1988), which is inconsistent with an invasion that predates our inland lake records. Second, the lack of evidence of B. cederströmii among Great Lakes plankton-net surveys done throughout the 1900s argues against its presence in the Great Lakes prior to inland lakes (Wells 1970; Watson and Wilson 1978). An alternative scenario of continental range expansion is that B. cederströmii initially arrived by a vector other than ship ballast water and initially established population centers in inland lakes, only later spreading to the Great Lakes.
Although still tentative, our single record that places B. cederströmii in North America pre-1650s (Three Mile Lake, Ontario) hints at the possibility that the genus could be native. However, this is a speculative hypothesis. Distinguishing between the two aforementioned scenarios of continental range expansion will require more evidence, some of which might be gathered through paleolimnological work on millennial time scales from many lakes and locations. Notably, these scenarios of range expansion are not mutually exclusive, making it possible that B. cederströmii in North America is an ensemble of multiple origins and gene pools, some of which may have arrived recently from Eurasia (Colautti et al. 2005). Although extant populations of Bythotrephes in North America represent the morphological species B. cederströmii (Korovchinsky and Arnott 2019), not all populations have been studied, including historical ones, leaving open the possibility that multiple morphological species have existed through time, or may currently coexist on the continent.
Our results also indicate that B. cederströmii populations persisted in lag phases (decades of prolonged low abundances) before escalating in the 1990s and 2000s (Fig. 4). Lag phases are characteristic of invasive species, though their causes (e.g., poor environmental conditions, and genetic and reproductive barriers) can be elusive (Simberloff 2009). In the case of B. cederströmii, climate anomalies, reductions in predation-based mortality, genetic subsidy from Eurasia, or reduction in mating barriers could all have possibly triggered their population expansions (Walsh et al. 2016b; Manca and DeMott 2009; Wittmann et al. 2011). Commonality in the time period (1990s and 2000s) when B. cederströmii transitioned from low to high abundance in the four inland lakes points to the likelihood of a single, regional triggering mechanism. A cooler than average July mean temperature is one possibility that is supported by circumstantial evidence. Data show that the entire region experienced low temperatures in the summer of 1992, around the time that accumulation rates of exoskeletal remains began to escalate across the study lakes. This is around the year that B. cederströmii also began appearing in the water column for the first time in inland lakes in Canada and the USA (Yan et al. 1992; Branstrator et al. 2006). Whether the close timing of cool temperatures and first appearances is causal or coincidental remains unclear. Walsh et al. (2016b) pointed to a cool summer in 2009 as the triggering mechanism that led to a surge in B. cederströmii density and its first detection in the water column of Lake Mendota, Wisconsin (USA). The mechanism by which cool temperatures drive a release of B. cederströmii from a lag phase may involve a thermal preference (Grigorovich et al. 1998), a dissolved oxygen preference (Sorensen and Branstrator 2017), or changes in predatory regime (Walsh et al. 2016b). Some fish species have declined in abundance in some of our study lakes over the last several decades (Anthony and Jorgensen 1977; Venturelli et al. 2014). However, the correlation between predatory regime changes through a decline in fish abundance and an increase in B. cederströmii abundance needs further study. In contrast to our results, two previous studies that used methods similar to ours report < 16 years between first presence of B. cederströmii in the sediments and first detections in the water column (Walsh et al. 2016b; Branstrator et al. 2017), which underscores the fact that long lag phases are not required for establishment of B. cederströmii populations.
Shortly after B. cederströmii was first recorded in North America, paleolimnology was heralded as a method that could be used to document the timing of lake invasions (Keilty 1988). The species’ barbed tail spine, which is diagnostic and preserves well in lake sediments, gave promise to the prospect that its arrival and presence in a lake could be reconstructed through analysis of exoskeletal remains. To date, however, only four published studies in North America have used paleolimnology to assess B. cederströmii invasions (Keilty 1988; Hall and Yan 1997; Walsh et al. 2016b; Branstrator et al. 2017). Therefore, we are still learning whether the approach of identifying very low numbers of specimens has vulnerabilities and incongruities with other methods when used to reconstruct the historical presence. In particular, we remain challenged to explain how B. cederströmii could have persisted for so long in these lakes without detection by plankton nets or anecdotal observation. Populations of B. cederströmii in invaded lakes vary in abundance during the growing season, but were continuously found in the water column in two studies examining seasonal abundance of the species (Brown et al. 2012; Yan et al. 2001). In Lake Kabetogama, extensive zooplankton monitoring with plankton nets from 1978 to 2003 failed to record a single B. cederströmii specimen (Kallemeyn et al. 2003; Kerfoot et al. 2016). However, given our sediment-based reconstructed estimates of the densities of B. cederströmii living in Lake Kabetogama during 1978–2003 (DeWeese 2020), such a failure of detection by plankton nets during this time period is statistically unlikely (Walsh et al. 2018). Likewise, in Lake Mille Lacs, aggressive zooplankton monitoring by net tows that began in 2006 by the Minnesota Department of Natural Resources staff failed to detect B. cederströmii until 2009 (DeWeese 2020). Moreover, both Lake Kabetogama and Lake Mille Lacs are common destinations for sport anglers, and the lack of any anecdotal evidence of an earlier presence of B. cederströmii, despite statewide awareness campaigns since the early 2000s, is puzzling and points to a gap in our understanding of the relationship between the neo- and paleo-records.
In Eurasia, where Bythotrephes is native, researchers have documented its exoskeletal remains and resting eggs in numerous lakes and to great (> 50 cm) sediment depths (Herzig 1985; Nilssen and Sandoy 1990; Milan et al. 2017). However, even there, exceptions occur. For example, in Lago Maggiore, Italy, Bythotrephes are easily detected with plankton nets, but two independent paleolimnological studies failed to find a single piece of exoskeletal evidence in sediment cores from the lake (Manca et al. 2007; Nevalainen et al. 2018). Possible explanations for the absence of sedimentary evidence of Bythotrephes in Lago Maggiore include poor exoskeletal preservation and spatial patchiness. Bythotrephes has a low calcium content compared to other cladocerans (Kim et al. 2011), which may explain differences in its sedimentary preservation. Additionally, previous studies have found that other cladoceran exoskeleton preservation may vary throughout the sediment record due to physical and biological factors, including fish consumption (Leppänen and Weckström 2016). This gap between the neo- and paleo-records in Lago Maggiore, like our results, is cautionary and points to the presence of still unknown relationships between living Bythotrephes populations and the fate of their sedimentary remains.
If the early timelines of B. cederströmii presence suggested by our results are artifactual, it is incumbent on us to identify the sources of error and consider how these could be resolved in future studies. In the interpretation of paleo-records of other crustacean zooplankton, it has been noted that sediment focusing and resuspension, as well as exoskeletal preservation, can vary by species and inter-annually (Korhola and Rautio 2001; Nykänen et al. 2009). For our purposes, we grouped the possible sources of error as: (1) those associated with the procedures and equipment that we used for core collection, processing, and analysis; and (2) those associated with natural redistribution of exoskeletal remains in lake sediments prior to our core collections.
In order to minimize error associated with core collection, processing, and analysis, our two teams used procedures and equipment that have been widely vetted in the discipline of paleolimnology. One procedural distinction of note between our two approaches was that during the extrusion step of core processing we trimmed and discarded the outer 0.5-cm edge of the intervals of the sediment cores collected from Lake Kabetogama and Lake Mille Lacs (USA), but not those collected from Lake Nipissing and Three Mile Lake (Canada). Trimming should have eliminated any material translocated along the core barrel walls during collection and extrusion.
If, however, smearing did extend inward farther than 0.5 cm it could have affected our timelines of species presence. Only further study can resolve whether this is of concern. If smearing leads to a predictable rate of decline in numbers of fragments downcore, this could possibly be used as a null hypothesis against which core data could be compared. The great length (> 0.5 cm) and barbed anatomy of the tail spine of Bythotrephes might increase its likelihood of ensnaring on surfaces and smearing, and this could present special problems for paleolimnological reconstruction. However, our extensive taxonomic evidence for B. cederströmii presence also includes small fragments of the tail spine, as well as mandibles and resting eggs (Fig. 2), none of which is unique in their size or anatomy compared to those of other crustaceans commonly studied using sediment cores. Regardless of the remains considered, the evidence suggests early presence in each of the four study lakes. It could be hypothesized that we incorrectly identified the exoskeletal remains of B. cederströmii, but that is highly unlikely given the unique morphological characteristics of the tail spine and mandible. Nonetheless, to be conservative we reported only the most diagnostic forms of the remains (mandibles, tail spine kinks, and resting eggs), and not the large numbers of additional tail spine fragments in the samples that lacked the kinked portion.
In addition to possible translocation of exoskeletal remains by core-wall smearing, the physicality of core collection and transport to the laboratory for the Minnesota cores (the Ontario cores were extruded on shore) could have redistributed exoskeletal remains down core. To minimize this, we applied Zorbitrol (Tomkins et al. 2008) to stabilize the sediment–water interfaces of the Minnesota cores immediately post-collection. Those cores were transported to the laboratory and stored horizontally in a walk-in refrigerator before analysis. Occasional visual inspection of the cores did not reveal any movement in the sediment or pore water along the core walls. Possible movement of exoskeletal remains could have happened in the interior regions of the sediment matrix that we could not ascertain by visual inspection. Whereas some minor movement of exoskeletal remains by this mechanism cannot be ruled out, the core profiles of 210Pb and 7Be are inconsistent with substantial sediment mixing. In addition, major changes in the microfossil assemblage of diatoms and midges that were evident mid-core in Lake Nipissing site 1 (Favot 2021) suggest that any mixing by coring, transporting, or extruding was not pronounced.
Our 210Pb dating results indicated that sedimentation rates in the top 25 cm of the cores differed by a factor of about 2 to 3 between Lake Kabetogama and Lake Mille Lacs (Fig. 3). Consequently, the depths of the time horizons dating to equivalent ages (e.g., the year 2000, 1980, or 1930; see Table 1) ranged accordingly. If B. cederströmii did first invade Lake Kabetogama in 2007 and Lake Mille Lacs in 2009 (which are dates of first water column records, respectively), it follows that the entirety of the inventories of exoskeletal remains that we recovered from earlier sediments in both lakes are artifactual. If so, it is curious that in both lakes we see a surge in densities of exoskeletal remains around 1990–2000. For this to be the case, significant numbers of exoskeletal remains would have had to translocate about 2–3 times deeper in the sediments of Lake Kabetogama than Lake Mille Lacs. This demands extraordinary and consistent errors that seem untenable. Moreover, the water and organic contents in the top 25 cm of Lake Kabetogama sediments were less than those in Lake Mille Lacs sediments (Table 1) and this should have reduced their vulnerability to mixing. Dating error, another possible artifact, can largely be ruled out as an explanation for the early appearance of exoskeletal remains in our cores. The exponential 210Pb activity profiles and low uncertainty associated with twentieth century dates (< 5 years) precludes timing errors of that magnitude.
As a second potential source of error, we considered whether physical resuspension at the sediment–water interface and bioturbation could have redistributed B. cederströmii exoskeletal remains prior to our core collections. A wide taxonomic range of macroinvertebrates, and some fish, disrupt sediments through occupancy (e.g., burrowing) and feeding, which can redistribute constituent particles, including exoskeletal remains of crustacean zooplankton (Adámek and Maršálek 2013). Chaoborus commonly burrow up to 3 cm depth (LaRow 1969; Gosselin and Hare 2003), and Hexagenia, which is among the deepest burrowers, may excavate tunnels up to 10 cm depth in lake sediments (Charbonneau et al. 1997). In three of our four lakes (Mille Lacs, Nipissing, and Three Mile), these burrowing depths could be sufficient to account for redistribution of exoskeletal remains to at least the 1980 horizons, but in Lake Kabetogama the 1980 horizon is too deep for disruption by such burrowing activity (Table 1).
To further explore whether natural sediment mixing (e.g., by wave action, porewater advection, and bioturbation) could help explain the patterns in exoskeletal remains, we analyzed 7Be decay profiles to estimate depths of modern, near-surface mixing in Lake Kabetogama and Lake Mille Lacs. Our results indicate that modern sediment mixing is confined to the top 2 cm in Lake Kabetogama and top 3 cm in Lake Mille Lacs. Given our sediment age-at-depth calculations for Lake Mille Lacs, the results imply that exoskeletal remains there could have been mixed to depths at which the sediment is nearly 20 years older than the age of the remains themselves. In Lake Kabetogama, however, where sedimentation rates are faster, exoskeletal remains could have been mixed by natural processes to depths at which sediments are only 4 years older, at most, than the age of the remains themselves. Even if we postulate that B. cederströmii were first present, died, and settled to the lake sediments in the 1980s (the decade of first detection of B. cederströmii on the North American continent; Johannsson et al. 1991), our 7Be-based estimates of modern sediment mixing in each lake, extrapolated across a respective core’s history, cannot explain the stratigraphic patterns. The only exception to this is with core site 1 from Lake Mille Lacs where low sedimentation rates yielded a sediment age of 1930 at 5 cm depth, which is only 3 cm below the 1980 horizon (Table 1). Although Hall and Yan (1997) demonstrated that B. cederströmii tail spines do not move vertically in sediments in Harp Lake, Ontario, their study was conducted within 1–2 years of first records of B. cederströmii in the lake, and it is possible that longer residency could cause deeper redistribution of remains if the phenomenon is time-dependent.
We also examined historical cores collected prior to B. cederströmii detection in open water to supplement our findings from recently collected cores. We expected to find exoskeletal remains in historical cores at similar ages and densities as the recently collected cores; however, we failed to recover any B. cederströmii remains. The complete absence of B. cederströmii exoskeletal remains suggests that sediment mixing may have occurred, although we took thorough steps to prevent mixing during core collection and extrusion and our isotope analysis did not indicate substantial mixing had occurred. These results remain a confounding element in our analysis.
In conclusion, ecological studies have long been plagued by inadequate methods to detect species that are small-bodied, behaviorally cryptic, low in abundance, or limited in distribution (Hoffman et al. 2011; Walsh et al. 2017). This gap has impeded our ability to reconstruct pathways and timelines of range expansion, and adequately document the distribution of biodiversity, for native and non-native species alike. In view of this, our results, if valid, highlight the potential that paleolimnology holds for the discovery and assessment of rare species in aquatic environments. The four lakes that we studied are popular tourist and angling destinations, and some are well-studied scientifically. Despite considerable management attention, B. cederströmii was not recorded in them until the 1990–2000s. If our results are valid, they suggest that sediment records may be far more sensitive to small founder populations than traditional detection tools (Walsh et al. 2017). Such prolonged lags in the discovery of invasive species have broad implications for the capacity of humans to manage this dimension of global environmental change (Vitousek et al. 1997).
Nonetheless, our results raise a number of issues that will eventually demand reconciliation. These include: (1) the lack of a clear narrative of continental invasion that is congruent with ballast water shipping and the timing of first records in the Great Lakes and inland lakes, (2) the lack of a clear mechanism for a transition from a lag phase to a growth phase in our four lakes, and (3) the lack of evidence of B. cederströmii using traditional methods prior to the 1980s in North America. For these reasons, we suggest that the early timelines for presence suggested by our results may be artifactual, and therefore should be regarded as ‘preliminary findings’ at this time.
We should emphasize that, in any typical paleolimnological assessment, the simple presence of one or two specimens would never be considered definitive, except for the special case of tracking an initial invasion (as we have attempted here). If indeed our seemingly early detection of B. cederströmii in all four of our lakes is an artifact, it does not call into question the large number of paleolimnological studies that have successfully used these approaches to reconstruct changes in zooplankton community structure, since most studies exclude very small numbers of remains. Future studies could help validate our results. Experiments that determine whether exoskeletal remains of B. cederströmii preferentially translocate downwards in sediments in response to biological and physical disruption, including during core collection and/or extrusion, could reveal errors in our methods and the nature of how lake sediments archive B. cederströmii remains. Additionally, examination of B. cederströmii in historical plankton collections, fish stomachs (e.g., museum-archived specimens), or sediment cores that were collected prior to dates of first records in the water column could help support or refute our results, which point to early presence. Until further study, however, we caution that paleolimnology may not be a secure method to pinpoint very early detection and assess initial timelines of colonization of B. cederströmii in ecosystems, if numbers are extremely low as they were in our cores.