Introduction

It has long been noted that Chaoborus (Diptera: Chaoboridae) may use anoxic strata as a refuge from predators (Juday 1921, cited in Stahl 1966). Chaoborus taxa are able to ‘oxy-regulate’ and thereby tolerate low oxygen concentrations (Jäger and Walz 2002) by utilizing haemolymph as a source of anaerobic energy (Scholz and Zerbst-Boroffka 1998). Chaoborus larvae do not maintain a ventilation tube when burrowing into anoxic sediments, and are therefore likely more exposed to toxic H2S and contaminants in interstitial water compared to tube-dwelling chironomids (Gosselin and Hare 2003). Chaoborids are likely better able to tolerate these conditions, compared to chironomids, due to a relatively more impermeable exoskeleton (Munger et al. 1999).

The population dynamics of Chaoborus have been shown to be influenced by hypolimnetic oxygen dynamics in lakes. For example, hypolimnetic oxygenation of mesotrophic Lake Newman, USA, led to a large decline in chaoborid abundance as a result of increased fish predation due to expansion of useable fish habitat into the formerly anoxic hypolimnion (Doke et al. 1995). Intuitively, the inverse situation may occur over time: formation or expansion of the anoxic/hypoxic zone in the hypolimnion will reduce useable fish habitat and exclude fish predators, resulting in an increase in chaoborid abundance due to decreased fish predation via an expanded refuge from fish predation. Assuming that influences on Chaoboridae abundance such as competition between Chaoborus and other invertebrate predators, or introduction/extirpation of fish in a lake do not change substantially through time, it is reasonable to hypothesize that temporal changes in hypolimnetic oxygen will influence abundances of Chaoborus in lakes. Such changes will be reflected in abundances of subfossil Chaoborus in lake sediment cores.

Subfossil Chaoboridae have been used in paleolimnological studies to infer periods of eutrophication and contamination in lakes, primarily in Finnish case studies of Lake Saimaa (Simola et al. 1996) and Lake Päijänne (Meriläinen and Hamina 1993; Meriläinen et al. 2001, 2003; Hynynen et al. 2004). Continued presence and abundance of Chaoborus mandibles in stratigraphic intervals where chironomid head capsules nearly disappear from the stratigraphic fossil record (e.g. Meriläinen et al. 2001; Hynynen et al. 2004), including anoxia- and contaminant-tolerant chironomid taxa such as Chironomus plumosus, indicate a particularly high tolerance by Chaoborus for either stressor.

Paleolimnological studies in relatively unproductive lakes in Canada have shown that a typical chironomid assemblage response to increased nutrient loading and anoxia may be a collapse of the entire profundal community (e.g. Clerk et al. 2000, 2004; Little et al. 2000; Reavie et al. 2006). Extirpation of the profundal community includes loss of anoxia-tolerant taxa such as Chironomus, and this results in an increase in the relative proportion of littoral taxa in the subfossil chironomid record. As a consequence, the use of relative abundance data (%) in chironomid-based inference models results in inferences that likely underestimate the duration or extent of anoxia in study lakes. Chaoborus ecological preferences have been quantified in temperature inference models (Barley et al. 2006), and in a chlorophyll a inference model (Brodersen and Lindegaard 1999), but have not yet been determined for oxygen gradients in paleolimnological models (Little and Smol 2001; Quinlan and Smol 2001). However, we have noted that subfossil Chaoborus mandibles were more likely to be present in surface sediments of lakes with anoxic hypolimnia than in surface sediments of lakes with oxic hypolimnia (Quinlan and Smol unpublished data). We hypothesized that Chaoborus subfossil remains would show a stronger relationship with anoxic conditions in a surface sediment training set of volume-weighted hypolimnetic oxygen (VWHO), than would ‘traditional’ chironomid indicators of anoxia (e.g. Chironomus). Closely related to this, we also hypothesized that Chaoborus subfossils would represent a greater proportion of dipteran subfossils in lakes with more anoxic conditions (i.e. lower VWHO), such that a ratio of Chaoboridae to Chironomidae subfossils may be a useful index to infer past changes in hypolimnetic oxygen.

Materials and methods

A complete study site description and field methodologies are found in Quinlan and Smol (2001). The study lakes were previously the focus of chironomid-based techniques and the field/lab techniques used in this study followed standard protocols (Quinlan and Smol 2001). Chaoborus specimen identification follows Uutala (1990). Chaoborus taxa recovered as mandibles included C. (subgenus Sayomyia), C. flavicans and C. trivittatus.

In this study we introduce an index of hypolimnetic oxygen status, the ‘chaob:chir’ ratio, which represents subfossil Chaoborus mandible abundance relative to subfossil Chironomidae head capsule abundance:

$$ {{\text Chaob:chir}}= (\# {\text{of chaob mandibles/2}})/((\# {\text{of chaob mandibles/2}})+\# {\text{of chir head capsules}})$$
(1)

This hypolimnetic oxygen index is analogous to a chrysophyte cyst:diatom frustule ratio, commonly used in paleolimnology as an index of trophic status (Smol 1985). It is based on a conceptual model of increased anoxia/hypoxia resulting in expanded predation refugia for Chaoborus populations, with the result that there are increased abundances of subfossil Chaoborus mandibles in lake sediments relative to abundances of subfossil chironomids. While Horppila et al. (2004) showed that low light intensity in deep, stratified clay-turbid lakes enhanced the extent of refugia for Chaoborus up into the metalimnion, no such lakes exist in our dataset. There were several clay-turbid lakes in our original 86-lake dataset, however these were all shallow and polymictic, and thus removed from analyses. As such, patterns observed in our study reflect the influence of oxygen dynamics, rather than turbidity dynamics.

There was concern that the absence or low abundance of subfossil Chaoborus mandibles in a lake’s sediments was a consequence of relatively low deposition rates of mandibles at mid-basin sites. Therefore, we conducted additional analyses to determine if the absence or low abundance of Chaoborus mandibles in lakes with few dipteran subfossils had a substantive influence on statistical analyses. For patterns of chaob:chir, additional analyses were conducted on lakes where >40 chironomid head capsules were sifted from the same surface sediments from which Chaoborus mandibles were retained (Quinlan and Smol 2001). It was assumed that for these lakes there was sufficient recovery of dipteran subfossil remains such that the absence or low relative abundance of Chaoborus subfossils was likely to reflect “true” Chaoborus absence or low subfossil abundances relative to chironomid remains. The relationships between chaob:chir in surface sediments (0–1 cm) and various limnological variables (trophic status, major ion chemistry, morphometry) were explored using linear regression in Statistical Packages for the Social Sciences (SPSS) v. 16. Surficial sediments were from an 80-lake training set of thermally-stratified lakes, and a reduced training set of 59 lakes with >40 chironomid head capsules (Quinlan and Smol 2001).

Chironomid- and chaoborid-based inference models of VWHO were developed using C2 version 1.5 (Juggins 2007). Regression techniques used in model development included partial least squares (PLS), weighted-averaging partial least squares (WA-PLS), weighted-averaging, and the modern analog technique (MAT) (Birks 1998). Only “best” model results are presented further (highest jack-knifed r 2, lowest root mean squared error of prediction (RMSEP), lowest predictive bias). Model results for the different regression techniques were similar to that of Table 4 in Quinlan and Smol (2001), with the “best” model consisting of weighted-averaging regression with tolerance downweighting and inverse deshrinking (RMSEP = 1.98 mg O2 l−1, maximum bias = 2.76 mg O2 l−1), with slightly higher maximum bias in WA-PLS models (3.01 mg O2 l−1), and slightly higher RMSEP in PLS models (2.27 mg O2 l−1).

Results

Patterns of chaob:chir

In analyses of either the full or reduced dataset (80 sites and 59 sites, respectively), a number of environmental variables were significantly correlated (P < 0.05) with chaob:chir, with hypolimnetic oxygen (VWHO), nutrients (TP) and morphometry (depth) having the strongest correlations (Table 1). There was no significant correlation between chaob:chir and volume-weighted hypolimnetic temperature (VWHT). Partial correlations with VWHO, TP or depth as a covariate resulted in the partial correlation coefficients of nearly all variables being reduced to non-significant values, with the exception of VWHO, which retained a significant coefficient regardless of covariate (Table 1). Stepwise multiple linear regression of chaob:chir and environmental variables, for both the full and reduced datasets and for both forward selection and backwards elimination procedures, yielded VWHO as the only significant environmental variable retained in the regression.

Table 1 Pearson correlation coefficients of the relationship between a Chaoborus mandible:chironomid head capsule ratio (chaob:chir) and environmental variables in 80 thermally-stratified lakes, with and without covariates included in the correlation analysis
Full size table

Paleolimnological inference models including Chaoborus

The ‘best’ (lowest RMSEP, highest r 2) VWHO inference model using chironomid-only or chironomid + chaoborid approaches was a WA regression model with inverse deshrinking and tolerance downweighting. Including Chaoborus in chironomid-based inference models of VWHO resulted in a modest increase in predictive power (3.9% decline in RMSEP) for the VWHO inference model, compared to a chironomid-only model (RMSEP 1.98 vs. 2.06 mg O2 l−1, respectively; Fig. 1). The weighted-averaging VWHO optima of Chaoborus taxa were substantially lower than that of Chironomus (e.g. 2.1 mg O2 l−1 for both C. flavicans and C. trivittatus, versus 3.4 mg O2 l−1 for Chironomus), typically considered one of the major chironomid indicators for deepwater anoxia. Huisman-Olff-Fresco (HOF) models (HOF software version 2.3, Oksanen and Minchin 2002) indicated that both Chironomus and Chaoborus taxa had statistically significant monotonic relationships (Type II model) with VWHO, however the relationship had a much stronger skew to the lower end of the VWHO gradient for Chaoborus (Fig. 2).

Fig. 1
figure 1

“Best” VWHO inference model results for a a chironomid-only model and b a chironomid and chaoborid model (54-lake training set, weighted-averaging with tolerance downweighting, inverse deshrinking). Diagonal solid line is a 1:1 line between observed and inferred values

Full size image
Fig. 2
figure 2

a VWHO versus% Chironomus, b VWHO versus% Chaoborus (as a % of Chironomidae). A second-order polynomial was fit to the data, approximating Type II HOF models (Oksanen and Minchin 2002)

Full size image

Applying the expanded dipteran-based VWHO model to a previously published stratigraphy where eutrophication has occurred and anoxia has increased in response to human disturbance (Little et al. 2000), the resulting VWHO inferences produced lower O2 conditions compared to a chironomid-only inference model (Fig. 3a), while chaob:chir values corresponded to trends in chironomid-only inferred VWHO (Fig. 3b).

Fig. 3
figure 3

VWHO inferences and chaob:chir derived from subfossil assemblages of a Gravenhurst Bay (Lake Muskoka, Ontario, Canada) stratigraphy (Little et al. 2000). a Inferred VWHO reconstruction using chironomid-only inference models (“Chir. inferred VWHO”), and all Chironomidae and Chaoboridae subfossils as a % of Diptera (“Chir. + Chaob. inferred VWHO), versus observed VWHO values; b chaob:chir versus chironomid-only inferred VWHO. Chaob:chir was multiplied by ×100, and the chaob:chir axis was reversed

Full size image

Discussion

Chaob:chir and deepwater oxygen

Deepwater oxygen is the strongest determinant of the presence/absence of Chaoborus in small inland shield lakes with fish populations (Quinlan and Smol unpublished data), with Chaoborus mandibles much more likely to be found in surface sediments of lakes with low VWHO, presumably due to the existence of an anoxic or hypoxic hypolimnetic refuge from fish predation. There were numerous environmental variables that were significantly correlated with chaob:chir, including nutrient levels (TN, TP), water chemistry (pH, specific conductivity) and morphometric characteristics (surface area, Z max) that would influence food levels and extent of hypolimnetic habitat. Statistical approaches that accommodated the inclusion of covariates to partial out the variation associated with VWHO (correlation matrices of chaob:chir and environmental variables) indicated, however, that their relationship with chaob:chir was largely associated with correlations between VWHO and other limnological gradients (e.g. correlation between VWHO and TP = −0.68; correlation between VWHO and Z max = 0.82). The inclusion of nutrients or morphometric variables as covariates did not reduce the influence of VWHO on chaob:chir patterns. These statistical results reinforce the inference that deepwater oxygen condition, presumably as a major determinant of the extent of a hypolimnetic refuge from fish predation, is a primary influence on Chaoborus populations in lakes, with resultant effects on chaob:chir values of dipteran subfossils recovered from surficial lake sediments. It is assumed that this greater relative importance of Chaoboridae in lakes with lower VWHO reflects greater extent of deepwater habitat that functions as a refuge from fish predation, resulting in increased Chaoborus abundance and consequently greater deposition of subfossil Chaoborus mandibles relative to subfossil Chironomidae head capsules. The significant relationship between chaob:chir and VWHO in linear regression approaches may reflect a continuous gradient between the extent of Chaoborus refugia and VWHO, as opposed to a simple dichotomous distribution of refugia in lakes that have either wholly anoxic hypolimnia or fully oxygenated hypolimnia (or wholly-hypoxic versus no-hypoxia).

Chaoborus in paleolimnological studies to infer changes in deepwater oxygen

The results of our study are in sharp contrast to those of Luoto and Nevalainen (2009), who showed no statistically significant influence of hypolimnetic dissolved oxygen on subfossil Chaoborus distribution, and a significant influence of July air temperature. These different results are likely due to the different characteristics of the lake set for each study. Our lakes were relatively deep (minimum Z max = 10.7 m) thermally-stratified lakes across a short climatic (latitudinal) gradient (~0.6°), while lakes in Luoto and Nevalainen (2009) were mostly polymictic lakes (maximum Z max = 7.0 m, hypolimnion present in only 31 of 80 lakes) across a long latitudinal gradient (10°).

While the RMSEP of our inference model was only modestly lower when including Chaoborus (3.9% decline; RMSEP 1.98 vs. 2.06 O2 mg l−1), more detailed analyses comparing model residuals suggested that model improvement is understated, as there is little or no improvement in model prediction for sites where either Chaoborus is absent or low in % abundance. In anoxic or hypoxic sites where Chaoborus was present and abundant, the inclusion of Chaoborus in the VWHO inference model improved the model prediction by up to ~2 mg O2 l−1 versus the chironomid-only inferred value. The maximum reduction in model residual in low-VWHO lakes (0–4 mg O2 l−1) was 2.3 mg O2 l−1, while the maximum reduction in model residuals in high-VWHO lakes (>8 mg O2 l−1) was only 0.4 mg O2 l−1.

Application of a VWHO inference model, with and without Chaoborus included in the analyses, to the Gravenhurst Bay sediment core stratigraphy from Little et al. 2000 (Fig. 3) suggests that the quantitative VWHO inference that includes Chaoborus is likely more realistic. The relatively short time span and high variability of the measured O2 record does not allow us to definitively determine, based on VWHO inferences alone, if the inclusion of Chaoborus in this particular case study makes the inference results more ‘realistic.’ Examining chaob:chir values relative to chironomid-only inferences, however, suggests that the dipteran-based VWHO inferences of ~1 mg O2 l−1 are more realistic than the chironomid-based VWHO inferences of ~2 mg O2 l−1, as values of the chaob:chir ratio (>20) in these sediment intervals were exceeded by only two anoxic lakes (VWHO < ~ 1 mg O2 l−1) within the reduced 59-lake modern dataset.

Despite the strong relationship between % Chaoborus and VWHO and the fact that a modified VWHO inference model that includes Chaoborus may be used to infer, more accurately, low VWHO values, caution should be used in inferring past VWHO changes in lakes that have experienced a substantial change in depth. Paleolimnological studies designed to determine long-term environmental change sometimes include ‘top–bottom’ approaches (Smol 2008). In this approach, long-term change is assessed by comparing inferences from surface deposits in a sediment core (‘top’ 0–1 cm of sediment, representing modern-day conditions) to inferences from a ‘bottom,’ i.e. deeper interval in the sediment core, which represents pre-disturbance conditions. This provides an evaluation of ‘net change’ in the lake environment since human disturbance began on a large scale. In southern Ontario lakes, the ‘bottom’ sample is collected at approximately >15 cm sediment depth, and represents material deposited >150 years ago, i.e. before 1850 AD (Quinlan et al. 2008; Smol 2008).

An examination of top–bottom chironomid-inferred VWHO changes versus top–bottom chaob:chir changes reveals a relationship that is conditional on lake depth history. Lakes with naturally determined water levels, i.e. no changes regulated by artificial structures, displayed, for the most part, a negative relationship between top–bottom changes in chaob:chir and chironomid-inferred VWHO, which was the expected relationship. Dammed lakes, on the other hand, displayed a positive relationship, i.e. an increase in the chaob:chir ratio with an increase in chironomid-inferred VWHO (Spearman correlation, r = 0.52; Fig. 4). These dam-at-outlet lakes, however, are relatively shallow for stratified lakes (Z max ~ 20 m) with relatively shallow hypolimnia, and the increase in hypolimnetic volume associated with damming may have provided larger refugia for Chaoborus such that their populations increased despite an improvement in hypolimnetic oxygen conditions. These results indicate that assessments of long-term hypolimnetic oxygen change using a ‘top–bottom’ paleolimnological approach that rely on inferences from shifts in chaob:chir, might prove problematic. Historical changes in lake depth that influence the presence/absence or extent of a hypolimnetic refuge, unrelated to changes in hypolimnetic oxygen, would confound the relationship between ‘top–bottom’ change in chaob:chir and inferred change in VWHO. In such cases, a more detailed stratigraphic study or use of a VWHO inference model that included Chaoborus is recommended. Despite this caveat with respect to using subfossil Chaoborus mandibles in paleolimnological inference approaches, given that subfossil chironomid-based VWHO inference models for southern Ontario are not strong indicators of low VWHO conditions (VWHO < 2 mg l−1), the use of Chaoborus mandibles improves the ability to track past changes in hypolimnetic oxygen, particularly if oxygen declines to hypoxic or anoxic levels.

Fig. 4
figure 4

Top–bottom changes in chaob:chir compared to top–bottom chironomid-inferred changes in VWHO (Quinlan and Smol 2002), in lakes with natural hydrological regimes versus lakes with constructed dams at their outlet. Chaob:chir was multiplied by ×100

Full size image