, Volume 162, Issue 3, pp 641–651

Population synchrony of a native fish across three Laurentian Great Lakes: evaluating the effects of dispersal and climate


    • Great Lakes Science CenterUS Geological Survey
  • Jean V. Adams
    • Great Lakes Science CenterUS Geological Survey
  • Owen T. Gorman
    • Great Lakes Science CenterUS Geological Survey
  • Charles P. Madenjian
    • Great Lakes Science CenterUS Geological Survey
  • Stephen C. Riley
    • Great Lakes Science CenterUS Geological Survey
  • Edward F. Roseman
    • Great Lakes Science CenterUS Geological Survey
  • Jeffrey S. Schaeffer
    • Great Lakes Science CenterUS Geological Survey
Population Ecology - Original Paper

DOI: 10.1007/s00442-009-1487-6

Cite this article as:
Bunnell, D.B., Adams, J.V., Gorman, O.T. et al. Oecologia (2010) 162: 641. doi:10.1007/s00442-009-1487-6


Climate and dispersal are the two most commonly cited mechanisms to explain spatial synchrony among time series of animal populations, and climate is typically most important for fishes. Using data from 1978–2006, we quantified the spatial synchrony in recruitment and population catch-per-unit-effort (CPUE) for bloater (Coregonus hoyi) populations across lakes Superior, Michigan, and Huron. In this natural field experiment, climate was highly synchronous across lakes but the likelihood of dispersal between lakes differed. When data from all lakes were pooled, modified correlograms revealed spatial synchrony to occur up to 800 km for long-term (data not detrended) trends and up to 600 km for short-term (data detrended by the annual rate of change) trends. This large spatial synchrony more than doubles the scale previously observed in freshwater fish populations, and exceeds the scale found in most marine or estuarine populations. When analyzing the data separately for within- and between-lake pairs, spatial synchrony was always observed within lakes, up to 400 or 600 km. Conversely, between-lake synchrony did not occur among short-term trends, and for long-term trends, the scale of synchrony was highly variable. For recruit CPUE, synchrony occurred up to 600 km between both lakes Michigan and Huron (where dispersal was most likely) and lakes Michigan and Superior (where dispersal was least likely), but failed to occur between lakes Huron and Superior (where dispersal likelihood was intermediate). When considering the scale of putative bloater dispersal and genetic information from previous studies, we concluded that dispersal was likely underlying within-lake synchrony but climate was more likely underlying between-lake synchrony. The broad scale of synchrony in Great Lakes bloater populations increases their probability of extirpation, a timely message for fishery managers given current low levels of bloater abundance.


Moran effectPopulation fluctuationSpatial synchronyFishDispersal


Understanding the mechanisms underlying fluctuations in population density has long been a primary goal in ecology. During the 1990s, however, ecologists became increasingly aware of how spatially distinct populations frequently exhibited synchrony in their abundance or growth rates (see reviews by Ranta et al. 1995; Koenig 1999; Liebhold et al. 2004). One commonly cited mechanism to explain this synchrony is the Moran effect (Moran 1953), whereby spatially autocorrelated climate synchronizes disparate populations that have a similar density-dependent structure (see Hanski and Woiwood 1993; Grenfell et al. 1998; Koenig 2002). Other potential explanations include dispersal among populations that act to synchronize local dynamics (Ranta et al. 1995; Haydon and Greenwood 2000) or mobile predators that can synchronize prey populations (Ydenberg 1987; Ims and Steen 1990).

Among animal populations, climate and dispersal are the two most commonly cited explanations for spatial synchrony (Koenig 1999). Because these two mechanisms can also interact to influence synchrony (Haydon and Steen 1997; Kendall et al. 2000), ecologists commonly attempt to determine the relative importance of each by building models (e.g., Ranta et al. 1995), designing experiments (e.g., Huitu et al. 2005), or even finding natural settings (e.g., Grenfell et al. 1998) where dispersal can be eliminated from consideration. Among freshwater and marine fish populations, most studies have concluded that climate is the primary mechanism underlying synchrony (Myers et al. 1997; Grenouillet et al. 2001; Tedesco et al. 2004; Marjomäki et al. 2004; Cheal et al. 2007; Phelps et al. 2008). In some cases, evidence supporting dispersal has been lacking (Grenouillet et al. 2001; Cattanéo et al. 2003; Marjomäki et al. 2004) and in other cases it was ruled out because different watersheds were sampled (Tedesco et al. 2004). A meta-analysis of fish populations revealed a second generality: the scale of spatial synchrony in marine populations (~500 km) is an order of magnitude greater than in freshwater ones (~50 km; Myers et al. 1997). The authors hypothesized that the relatively low synchrony observed in freshwater populations was due to localized biotic factors that overwhelmed the role of climate (Myers et al. 1997). Although subsequent studies on freshwater populations have documented spatial synchrony to be as high as 150–300 km, climate has remained the primary explanatory factor (Grenouillet et al. 2001; Tedesco et al. 2004, Marjomäki et al. 2004; Phelps et al. 2008).

Using the largest freshwater ecosystem in the world, the Laurentian Great Lakes, our study affords a unique opportunity to explore spatial synchrony in a freshwater fish population over a broad spatial scale (up to 1,000 km), where dispersal is possible within each lake and is even possible (to varying degrees) between lakes. Populations of bloater (Coregonus hoyi) remain extant and abundant in three of the Great Lakes (Superior, Michigan, and Huron). In Lake Ontario, bloater collapsed in the 1960s and were nearly extirpated by the early 1980s for reasons not well understood (see Christie 1973). Bloater never occurred in Lake Erie, likely owing to its relatively shallow depth. Bloater have been best studied in Lake Michigan, where current knowledge suggests that the population is regulated by intrinsic factors, such as skewed sex ratios, rather than extrinsic factors, such as predation, winter and spring weather, or the relatively small commercial fishery (Bunnell et al. 2006, 2009). Bloaters undergo ontogenetic habitat shifts from the profundal (i.e., benthic, offshore waters) zone as embryos, to the offshore epilimnetic zone as larvae and early juveniles, and then return to the profundal zone as adults (Rice et al. 1987). Bloaters are rarely consumed by piscivores in lakes Michigan and Huron (Madenjian et al. 1998, 2006), but bloaters and kiyis (Coregonus kiyi) are commonly consumed by offshore siscowet lake trout (Salvelinus namaycush) in Lake Superior (Harvey and Kitchell 2000). Because offshore siscowet lake trout do not occur in lakes Michigan and Huron, we can effectively rule out the “common mobile predator” as a synchronizing mechanism in our study.

Hence, climate and dispersal are the most plausible explanations for any spatial synchrony observed across these bloater populations. The likelihood of dispersal synchronizing populations between any two of the Great Lakes can be qualitatively characterized. Dispersal is most likely between lakes Michigan and Huron given that water distance separating sampling sites is relatively small (mean = 463 km, range 138–785 km) and that the lakes are hydrologically connected by the Straits of Mackinac, where a deep (>70 m) and narrow (0.5–1.0 km) channel exists, through which all life stages of bloaters could move. Dispersal is of intermediate likelihood between lakes Superior and Huron because the water distance separating sampling sites is relatively intermediate (mean = 568 km, range 144–961 km) and the hydrological connection is the St. Mary’s River, which has a bottleneck at the Sault Locks where fish must traverse the rapids (1–3 m depth) or shipping locks (7–9 m in depth, 24–33 m width). Dispersal is least likely between lakes Superior and Michigan because water distance separating the sampling sites is relatively great (mean = 819 km, range 309–1,248 km) and both the Straits of Mackinac and Sault Locks must be traversed. Given that we expect climate to be largely synchronized across the three upper Great Lakes (sensu Myers et al. 1997; Koenig 2002), we can use the differing likelihoods of dispersal as a natural experiment to determine the relative contribution of dispersal and climate to any bloater synchrony that we observe between different lake populations.

For each of the three lakes, we collected annual catch-per-unit-effort (CPUE) estimates of bloater recruits (i.e., age-0 or age-1) and the overall population from 1978 to 2006. We estimated whether spatial synchrony occurred with these two metrics, which provided insights into the long-term, regional trends of the bloater population. In some cases, however, long-term population trends can obscure synchrony that occurs in short-term fluctuations (Koenig 1999; Liebhold et al. 2004). As a result, we also removed the long-term trends from the recruit and population data by calculating the annual rate of change for each time series (see Liebhold et al. 2004). In total, we sought to determine whether spatial synchrony existed in both long- or short-term bloater recruitment and population trends, and whether any observed synchrony could be explained by dispersal or climate.

Materials and methods

Field collection

Bloaters were sampled during annual US Geological Survey Great Lakes Science Center bottom trawl surveys on each lake. We used data from 1978–2006, as 1978 is the earliest year that data were collected in all lakes. Within this time period, however, some sampling was missed owing to vessel malfunctions or improper deployment of the sampling gear (e.g., Huron in 1992, 1993, 1998, 2000; Michigan, 2000; Superior, 1980). Some sampling procedures were common across lakes, such as the use of diurnal tows and the sorting of catch by species for abundance and biomass estimates. Depending on the size of the bloater catch, either a subset or the entire set of individuals was measured to the nearest mm total length (TL). Other sampling procedures varied across lakes. In Lake Superior, trawls were fished for approximately 60 min across depth contours, beginning near the 18-m (interquartile range 15–20 m) depth contour and ending near the 56-m (interquartile range 44–74 m) depth contour, at 53–56 sites. Conversely, in lakes Michigan and Huron, 10-min trawls were fished at up to 11 different discrete depths (ranging 9–110 m) at one of seven sites on Lake Michigan or five sites on Lake Huron. Sampling in Lake Superior occurred in spring whereas sampling in lakes Michigan and Huron occurred in autumn. Lakes Michigan and Superior used the same trawls (headrope of 12-m), and this trawl was also used on Lake Huron from 1978–1991. From 1992–2006, however, a larger trawl with a 21-m headrope (same cod-end mesh) was used. To join these two time series, separate fishing power corrections were estimated for age-0 and age-1 and older bloaters in Lake Huron (Adams et al. 2009).

Data processing and analyses

In each lake, we estimated bloater CPUE (fish/ha) for both the entire population sampled as well as the youngest age-class captured (i.e., recruits). In lakes Michigan and Huron where the survey occurred in autumn, recruits were age-0 fish, whereas in Lake Superior, where the survey occurred in spring, recruits were age-1 fish. During the Lake Superior survey, age-0 bloaters were only in their egg and larval stages, and thus unavailable to the bottom trawl. Recruits were not fully recruited to the bottom trawl in any lake, which meant that the absolute estimate of year-class strength was biased low. In a relative sense, however, the CPUE of recruits was reflective of relative year-class strength. In Lake Michigan, for example, bloaters fully recruit to the bottom trawl at age 3 (Bunnell et al. 2006), but there is a strong positive correlation (r = 0.68; P < 0.0001) between age-0 CPUE and age-3 CPUE for the year-classes from 1962 up to and including 2000 (D .B. Bunnell, unpublished data). The area swept by the bottom trawl (ha) in each lake was calculated according to results of net mensuration that identified net width and time on bottom as a function of net depth. Length frequency plots were used to determine the maximum TL of recruits in each lake annually: Lake Superior (130 mm TL), Lake Michigan (range = 95–125 mm TL, mean = 102 mm TL), and Lake Huron (range = 105–125 mm TL, mean = 123 mm TL).

Because of differences in survey design, mean CPUE estimates were calculated differently between Lake Superior and lakes Michigan and Huron. In Lake Superior, between six and ten sites were grouped into one of six ecoregions, which share similar geography and geomorphology (i.e., shore substrate and steepness; see Gorman et al. 2007). For each eco-region, we calculated the mean arithmetic recruit and population CPUE. For lakes Michigan and Huron, we calculated the mean arithmetic recruit and population CPUE for each site across depth contours, but restricted our sample to include only six depths (27, 37, 46, 55, 64, 73 m) to standardize sampling depths with Lake Superior. This did not bias the results, however, as we found that trends estimated with the subset of depths in Lake Michigan were highly correlated (r ≥ 0.94 for both recruits and population CPUE) with the trends that used all available depth contours. To homogenize the variance and normalize the distributions, long-term trends of recruits (Rt) and population (Pt) CPUE data at a given site in year t were transformed with a natural log (x + 1) transformation (sensu Koenig 1999). To characterize the short-term fluctuations or trends, we calculated ΔRt = Rt − Rt−1 and ΔPt = Pt − Pt−1 for recruits and the population, respectively (Liebhold et al. 2004).

For the climate data, we calculated the mean monthly winter (December–February) and summer (July–September) air temperature at nine sites along lakes Superior, Michigan, and Huron that were interspersed among our trawling sites (see map in Online Resource), as multiple time series of historical water temperature data were not available. We downloaded monthly mean air temperature data from the Global Historical Climatology Network temperature data base version 2 (Peterson and Vose 1997): http://www.ncdc.noaa.gov/oa/climate/ghcn-monthly/index.php (accessed 24 November 2008). We chose winter because winter has been hypothesized as a critical period for survival of bloater eggs and embryos (e.g., Rice et al. 1987; Bunnell et al. 2006). We chose summer because bloater larvae occupy the epilimnion during these months (Rice et al. 1987), and thus may be influenced by air temperature to a greater extent than the adult life stage that occupies the hypolimnion. To detrend the climate data, we calculated residuals from a linear regression for each time series (sensu Koenig 2002).

Modified correlogram

Following the analytical approach of most spatial synchrony papers, we used the modified correlogram method (Koenig and Knops 1998) to evaluate whether bloater recruitment and population data, as well as climate data, were spatially synchronized across and within the upper Great Lakes basin. We calculated distance between each pair of sites for bloater (by water) and climate (by air). In calculating water distance, we assumed that bloaters would swim the most direct distance between sampling sites. We used ArcView GIS 3.3 (ESRI) to estimate the latitude/longitude of intermediate, connecting points along each direct route, and then used the great circle estimation method to calculate the total water distance. Between each pair of sites, the Pearson correlation coefficient was calculated for both bloater CPUE and climate. The pairs were grouped into distance categories and the mean Pearson correlation coefficient was calculated for each category. To gain statistical inference (as the correlation pairs are not independent), we calculated bootstrap 95% confidence intervals (95% CI) for each distance category by resampling time points (rather than sites) with replacement 1,000 times per distance category (Lillegård et al. 2005). If the 95% CI overlapped with 0 in a given distance category, then we concluded that synchrony did not occur at that spatial scale.

For the bloater data, we first constructed the modified correlogram for all possible data pairs within and between lakes Superior, Michigan, and Huron. This correlogram provided a measure of basin-wide synchrony, and we selected five distance categories (≤200, >200–400, >400–600, >600–800, >800–1,300 km) to maximize the number of categories yet maintain similar sample sizes among categories. To further understand the basin-wide pattern we constructed six correlograms, each constructed with one of six unique subsets of time-series pairs to enable comparisons of within-lake and between-lake patterns. Three correlograms were constructed with within-lake pairs of data (within Huron, within Michigan, within Superior) and three correlograms were constructed with between-lake pairs of data (between lakes Huron and Superior, between lakes Michigan and Huron, between lakes Michigan and Superior). The between-lake correlograms had four distance categories (≤300, >300–600, >600–900, >900 km), whereas the within-lake correlograms had only three (≤200, >200–400, >400–600 km). For the climate data, there were nine sites: three on Lake Superior, four on Lake Michigan, two on Lake Huron. Only three distance categories were selected (≤200, >200–400, >400–700 km), and owing to smaller sample sizes, only two unique correlograms were constructed (within and between lakes).


Modified correlograms yielded very similar qualitative results whether population or recruit data were used. For brevity, we present correlogram results only from recruitment CPUE data, and explicitly note the instances where results differed for population CPUE. Recruitment data can be considered a more robust evaluation of spatial synchrony because they limit the effects of temporal autocorrelation that are present in the population data given that bloater longevity is at least 13 years. Conversely, recruitment in any 2 consecutive years has a much lower probability of being correlated given the high interannual recruitment variability typically observed for fish species (though not necessarily true for bloater; Bunnell et al. 2006), likely owing to the role of density independent factors that can swamp the effect of the number of spawning adults (Houde 1987; Maceina and Pereira 2007).

Long-term trends in bloater CPUE

Across all three Great Lakes, CPUE of bloater recruits was generally highest in the 1980s, lowest in the 1990s, and intermediate in the 2000s (Fig. 1a–c). The pattern for population CPUE was similar to the recruitment pattern in lakes Michigan and Huron, but the population exhibited a less clear temporal trend in Lake Superior (Fig. 1d–f). When a correlogram was constructed with all recruit CPUE data, mean correlation between sampling sites declined with water distance (≤200 km = 0.69, >200–400 km = 0.49, >400–600 km = 0.27, >600–800 km = 0.25, >800–1,300 km = 0.15), yet significant correlation occurred up to >600–800 km. When all population CPUE data were pooled, significant correlation occurred only up to >400–600 km. Correlograms constructed from within-lake time-series always had a higher scale of spatial synchrony than correlograms constructed from between-lake time-series (Fig. 2). For each within-lake correlogram of recruit data, all distance categories exhibited spatial synchrony (Fig. 2b, d, f); when population CPUE data were used, the only difference was no significant correlation in the >400–600 km distance category within Lake Superior.
Fig. 1

Time series of recruits (ac) and the population (d–f) of bloaters (Coregonus hoyi) sampled with a bottom trawl in lakes Superior (a, d), Michigan (b, e), and Huron (c, f) from 1978 to 2006. For lakes Michigan and Huron, each line represents the annual average for each site calculated across depth contours. For Lake Superior, each line represents the annual average for an ecoregion calculated across sites. Note log scale used on y-axis. CPUE Catch-per-unit-effort

Fig. 2

Modified correlograms for long-term synchrony (CPUE) of bloater recruits (Rt) when the time-series pairs were separated into six unique categories: a between lakes Huron and Superior, b within Lake Huron, c between lakes Michigan and Huron, d within Lake Michigan, e between lakes Michigan and Superior, f within Lake Superior. a–f Each open circle represents the correlation in two time series as a function of their separating water distance. Within each distance category (vertical gray lines), the filled triangle value [and its bootstrapped 95% confidence interval (CI)] represents the mean correlation. Spatial synchrony for a given distance category in a given figure part was determined to be significant when the lower 95% CI was greater than 0

Comparing spatial synchrony results across the three between-lake correlograms provided insight as to whether climate or dispersal was a more important mechanism. For the between-lake Michigan and Huron correlogram (where dispersal was most likely), spatial synchrony occurred out to the >300–600 km distance category for recruit CPUE (Fig. 2c) and out to the >600–900 km one for population CPUE. For the between-lake Huron and Superior correlogram (where probability of dispersal was intermediate), spatial synchrony did not occur for recruit CPUE (Fig. 2a) and occurred in only the >600–900 km distance category for population CPUE data. For the between-lake Michigan and Superior correlogram (where dispersal was least likely), spatial synchrony occurred in the >300–600 km distance category (Fig. 2e) for both recruit and population CPUE data; no data were available in the ≤300 km distance category. Finding a different scale of spatial synchrony across these correlograms was not consistent with the hypothesis that broad-scale regional climate synchronizes these populations. At the same time, evidence for dispersal was inconclusive, given the relatively high scale of synchrony that occurred between lakes Michigan and Superior.

Short-term trends in bloater CPUE, and annual rate of change for recruit and population CPUE

The annual rate of change for both recruit (ΔRt; Fig. 3a, b, c) and population (ΔPt; Fig. 3d, e, f) CPUE showed no temporal trend for lakes Superior, Michigan, and Huron. When a correlogram was constructed with all pairs of ΔRt data, correlation between sampling sites declined slightly with water distance (≤200 km = 0.45, >200–400 km = 0.27, >400–600 km = 0.11, >600–800 km = 0.10, >800–1,300 km = −0.05) but the correlations were significant only up to >200–400 km. When all ΔPt data were pooled, spatial synchrony was significant up to >400–600 km, slightly greater than ΔRt. Within each lake, spatial synchrony was significant in the ≤200 and >200–400 km distance categories for both ΔRt and ΔPt data for all lakes (Fig. 4 b, d, f). For the three correlograms constructed with data from between-lake sites, spatial synchrony never occurred for ΔRt data (Fig. 4a, c, e), and occurred only for the >600–900 km distance category between lakes Huron and Superior for the ΔPt data. Finding synchrony only in within-lake data indicates that synchrony in short-term trends can be explained by either dispersal (at the within-lake scale) or localized climate regimes.
Fig. 3

Time series of short-term trends (i.e., annual rate of change) for recruits (ΔRt, ac) and the population (ΔPt, df) of bloaters sampled with a bottom trawl in lakes Superior (a, d), Michigan (b, e), and Huron (c, f) from 1978 to 2006

Fig. 4

Modified correlograms for short-term synchrony (annual rate of change; ΔRt) of bloater recruits when the time-series pairs were separated into six unique categories: a between lakes Huron and Superior, b within Lake Huron, c between lakes Michigan and Huron, d within Lake Michigan, e between lakes Michigan and Superior, f within Lake Superior. a–f Each open circle represents the correlation in two time series as a function of their separating water distance. Within each distance category (vertical gray lines), the filled triangle value (and its bootstrapped 95% CI) represents the mean correlation. Spatial synchrony for a given distance category in a given figure part was determined to be significant when the lower 95% CI was greater than 0

Long- and short-term trends in climate synchrony

Both winter and summer air temperatures were strongly synchronized at all distance categories, independent of whether raw or detrended data were used. During winter, mean correlation ranged from 0.97 (<200 km) to 0.94 (>400–700 km) for raw data, and from 0.96 to 0.92 for detrended data. Given that no correlation was lower than 0.86 during the winter, separating the pairs into between and within lake was unnecessary. During summer, mean correlation ranged from 0.88 (<200 km) to 0.78 (>400–700 km) for raw data and from 0.90 to 0.80 for detrended data. When we separated the summer pairs into within- and between lake, we found significant spatial synchrony in all distance categories of each correlogram.


Pooling data across lakes Superior, Michigan, and Huron, bloater recruits exhibited spatial synchrony up to a scale of 800 km for long-term trends, and 400 km for short-term trends, both of which surpass the scale that has previously been observed in freshwater fish populations. Spatial synchrony consistently occurred up to 400 km within each of these lakes, for both long- and short-term trends. Synchrony between lakes was less common. For long-term trends of recruits (a more conservative measure than using population long-term trend data), no spatial synchrony occurred between lakes Huron and Superior, but synchrony occurred out to 600 km for the other two lake pairs. Conversely, for short-term trends of recruits, synchrony between lakes was never observed. These results failed to reveal a simple explanation for either dispersal or climate synchronizing these bloater populations. Support for climate synchronizing these populations was weakened by different scales of synchrony observed among the between-lake correlograms. Similarly, evidence that dispersal was synchronizing the variable between-lake synchrony was equivocal. In support of the dispersal mechanism, we found that spatial synchrony was relatively broad (i.e., out to 600 km) between lakes Michigan and Huron, where dispersal is most likely to occur. Conversely, similarly broad synchrony occurred between lakes Michigan and Superior, where dispersal is unlikely given the low probability that any life stage of bloater could move between these two systems. Below, we further consider the evidence for climate and dispersal as synchronizing mechanisms for bloater populations, both within and between these three lakes.

Previous studies that have evaluated synchrony among fish populations have largely concluded that regional or broad-scale climate patterns (e.g., temperature, precipitation-driven hydrology) is the primary underlying mechanism (Myers et al. 1997; Cattanéo et al. 2003; Marjomäki et al. 2004; Tedesco et al. 2004; Phelps et al. 2008). In some of these cases, dispersal was ruled out (Tedesco et al. 2004; Marjomäki et al. 2004), and in other cases connected sites exhibited no greater synchrony than unconnected sites (Grenouillet et al. 2001; Cattanéo et al. 2003; Marjomäki et al. 2004). In our study, climate (i.e., winter and summer air temperatures) exhibited high spatial synchrony at the greatest spatial scale (up to 700 km), which was consistent with previous documentation of broad-scale (up to 2,500 km) synchrony in air temperatures and precipitation in the continental United States (Koenig 2002). Given the apparent regional, broad-scale synchrony in climate that we observed, the most convincing evidence for broad-scale, climate-driven synchrony would have been similar scales of synchrony for the three correlograms constructed with between-lake pairs of data. This consistent synchrony pattern did not emerge for either short- or long-term bloater trends.

Despite the lack of evidence for regional, climate-driven synchrony, we cannot reject the hypothesis that climate synchronizes bloater populations. First, climate is the most parsimonious explanation for the synchrony in long-term bloater trends between lakes Michigan and Superior; dispersal between these two systems is highly unlikely and they do not share common predators. If climate synchronized these lakes, then it seems logical to argue that the long-term synchrony between lakes Michigan and Huron also was climate related. Under this line of reasoning, one need only explain why synchrony did not occur between lakes Huron and Superior. Myers et al. (1997) argued that regional climate synchronization could be overwhelmed by some other factor, such as biotic interactions that can differ between two systems. In this case, Lake Superior is unique in that it has a high biomass of offshore, siscowet lake trout predators that consume bloaters (Harvey and Kitchell 2000; Ray et al. 2007); siscowet lake trout do not occur in lakes Huron or Michigan and predation on bloaters by other Lake Huron predators is relatively uncommon (Madenjian et al. 2006). This difference could account for the asynchrony between lakes Huron and Superior, but this argument is weakened by the fact that predation on bloaters in Lake Michigan is also relatively uncommon (Madenjian et al. 1998). We considered other factors (e.g., bloater prey resources, evidence for disease, etc.) that might be uniquely different between lakes Huron and Superior but none were apparent. Therefore, it remains unclear why a food web difference between lakes Huron and Superior might have overridden climate-driven synchrony in these lakes, but not between lakes Michigan and Superior.

Climate also could be synchronizing bloater populations within each lake, if water temperatures within each lake were not as synchronous as the air temperatures suggested. The three lakes differ markedly with respect to bathymetry (Lake Superior mean depth = 147 m, Lake Michigan mean depth = 85 m, Lake Huron mean depth = 59 m), which can cause water temperatures to respond differentially to similar air temperatures. Austin and Colman (2007) reported that since 1979 Lake Superior surface water temperature has warmed at a faster rate than expected based on air temperatures alone, perhaps because of more limited ice cover in recent years. Although similar analyses were not conducted in lakes Michigan and Huron, inter-lake differences in bathymetry could result in asynchronous water temperatures among the lakes which, in turn, could synchronize populations within each lake but limit the extent of synchrony between the lakes. We are less willing to accept this localized climate hypothesis because the synchrony in long-term trends (up to 600 km) between lakes Michigan and Superior is most parsimoniously explained by broad-scale climate synchrony and invoking both local- and broad-scale climate patterns to simultaneously explain bloater synchrony is difficult to defend.

The hypothesis that dispersal can synchronize spatially distinct populations has been supported by both theoretical models and empirical data. Models have revealed that populations that exchange individuals each generation can be brought into synchrony (Ranta et al. 1995, 1998; Kendall et al. 2000), even if inter-population variation in density-dependent structure exists (Blasius et al. 1999; Liebhold et al. 2006). Conclusive empirical support for the synchronizing effect of dispersal is fairly rare, as it is difficult to rule out climate effects. For fish populations, we know of only one study that considered the dispersal mechanism and they concluded that both dispersal and climate could be synchronizing synchrony in coral reef populations at a relatively small spatial scale (i.e., <100 km; Cheal et al. 2007). Beyond fish populations, Schwartz et al. (2002) documented high gene flow among Canada lynx (Lynx canadensi) populations over the same range that synchrony had been documented, and argued that dispersal was the most parsimonious explanation even though climate effects could not be discounted. Two other studies that supported the synchronizing role of dispersal involved bird populations where species with greater dispersal capabilities were found to exhibit higher synchrony than those populations with lower dispersal capabilities (Paradis et al. 1999; Bellamy et al. 2003).

Similar to previous studies, our work provided some evidence that dispersal could be synchronizing bloater populations, but we also acknowledge that dispersal, alone, is insufficient to explain our field patterns. It is possible that dispersal could be synchronizing time series within each lake, as well as the between lakes Michigan and Huron long-term trends. Evidence of bloater movement at the scale of hundreds of kilometers, either through tagging studies or genetic analyses that reveal gene flow, would provide supporting evidence for the dispersal hypothesis. We know of only one study that determined bloater movement through tagging. In 1931, Smith and Van Oosten (1940) recovered six out of 56 adult deepwater ciscoes (Coregonus spp.) tagged in Lake Michigan. Bloater was one of six deepwater cisco species in the lake at this time, and it comprised ~31% of the deepwater cisco community (Smith 1964). All six recaptured fish were collected within 16 km of their tagging site, indicating that some, albeit limited, dispersal occurs at this life stage. Additional movement data are required to determine the true extent of bloater dispersal, but the current knowledge suggests a low likelihood that adult dispersal is synchronizing the Great Lakes bloater populations.

Unlike the adult tagging data, a recent genetic study of bloater populations in lakes Superior, Michigan, and Huron provided evidence that dispersal occurs between lakes Michigan and Huron and within each of the three lakes (Favé and Turgeon 2008). Bloater from lakes Huron and Michigan were not genetically differentiated (Favé and Turgeon 2008), indicating that some level of gene flow occurs. Additionally, bloater from lakes Michigan and Huron were differentiated from bloater in Lake Superior, and no differentiation was observed among bloater from within Lake Superior (Favé and Turgeon 2008). These results indicated that gene flow (and dispersal) occurs within each lake, as well as between lakes Michigan and Huron, where we identified dispersal to be most likely.

If intra- and inter-lake dispersal occurs, it most likely occurs in July and August in the larval life stage, when bloater ontogenetically shift from profundal (bottom waters in 40–100 m depth) to epilimnetic (surface waters above the same bottom depth) habitat at a size of 15 to 20 mm TL (Rice et al. 1987). At this size, bloater larvae are predicted to maintain an average swimming speed of 2–3 cm/s and attain burst speeds of 13–15 cm/s (Miller et al. 1988). Although these larvae are not passive plankton, they likely are still influenced by prevailing currents within each lake and may be redistributed far from their spawning sites, as has been recently documented for larval yellow perch (Perca flavescens; Dettmers et al. 2005) and cisco (Coregonus artedi; Oyadomari and Auer 2008) in lakes Michigan and Superior, respectively. In fact, predicted burst speed of bloater larvae is similar to the mean current velocity (15–23 cm/s) and far less than the maximum velocity (45–82 cm/s) measured by drifters in Lake Michigan in July 2001 (Höök et al. 2006). Therefore, it seems possible that in years of high larval production, currents could distribute larvae elsewhere, leading to high recruitment throughout the lake.

For dispersal to synchronize the lakes Michigan and Huron populations, epilimnetic larvae must move through the Straits of Mackinac, where the net direction of flow is from Lake Michigan to Lake Huron (Saylor and Sloss 1976). To arrive near the Straits with the aid of prevailing currents (see Beletsky and Schwab 2001), embryos could hatch from one of two general spawning areas in the northern basin. First, an eastern current (toward the straits) prevails (Beletsky and Schwab 2001) above the Beaver Island complex in northern Lake Michigan. Second, below the Beaver Island complex currents in the northern basin generally move in a counter-clockwise gyre. Offshore from Manistee and Charlevoix, Michigan, however, some of the northward currents in the gyre continue northward to the Straits. Hence, in years of strong larval production in northern Lake Michigan, some fraction of larvae could disperse into Lake Huron. If dispersal were driving the synchrony between lakes Michigan and Huron, then we would expect higher correlations at closer distances in this between-lake correlogram. We failed to see this pattern, however. Therefore, this low dispersal might be sufficient to maintain gene flow between the two lakes (and the absence of genetic differentiation), but we doubt that it is sufficient to synchronize population densities.

In conclusion, our analyses revealed synchrony in time series of recruit and population CPUE of bloater at a greater spatial scale than has previously been documented in freshwater populations. The mechanisms underlying this synchrony were not easily elucidated, as no consistent evidence emerged for regional, broad-scale climate or for dispersal. Consideration of all the information leads us to favor broad-scale climate as the mechanism underlying the long-term synchrony between the lakes, and dispersal underlying the high within-lake synchrony for both short- and long-term bloater trends. Because the oscillations of the bloater populations are suggestive of cycling (see Madenjian et al. 2002; Bunnell et al. 2006), bloater synchrony across the Great Lakes has added implications for fishery managers. Populations in synchrony have a greater probability of extirpation (Heino et al. 1997), especially if dispersal is a contributing factor (Ranta et al. 1998). If populations were asynchronous, then one population nearing extirpation could be reinvigorated by others that were at higher levels of abundance. Consequently, fishery managers in the Great Lakes should consider this risk to bloater populations given that all three populations are currently at relatively low levels of abundance.


We thank T. Desorcie for assistance in estimating distance between sampling sites and developing the figures. We thank S. Delean and M. Drever for assistance with bootstrap coding. We also thank the many biologists, technicians, and vessel crew that have worked to maintain the long-term bottom trawl survey on lakes Superior, Michigan, and Huron. M. Ebener, J. Savino, D. Yule, and several anonymous reviewers provided helpful comments on early manuscript drafts. This article is contribution 1556 of the US Geological Survey Great Lakes Science Center.

Supplementary material

442_2009_1487_MOESM1_ESM.doc (323 kb)
Supplementary material 1 (DOC 323 kb)

Copyright information

© Springer-Verlag 2009