FIELD METHODS AND WATER CHEMISTRY ANALYSES

Lake water and sediment samples were collected from 34 lakes between 11 and 22 July 2000. An additional 5 lakes were sampled in July or early August in 1996, 1997, or 1998 (Ash, Trout, Patrick, Hungry, and K819). Chironomid analyses were performed on sediment samples that were collected from near the center of the lake using a gravity corer (Glew, 1991), with the exception of the 1998 samples, which were collected with an Ekman Dredge (Hungry and K819) or clear plastic tube fitted with a piston. The upper 1 to 2 cm were extruded at 0.5 cm or 1 cm intervals using a vertical extruder (Glew, 1988).

Conductivity (Cond) and temperature (T) profiles were recorded using a YSI model 33 SCT meter, and oxygen profiles were taken with a YSI model 51B oxygen meter, measured at 0.5 m intervals. Surface-water pH was averaged from 2 or 3 different handheld pH meters (Oakton pHTestr™ 2 and 3) that were calibrated daily to standard buffer solutions. Water transparency (Secchi) was measured using a standard 22-cm-diameter Secchi disc. Water collection, filtering, and sample storage protocols followed Environment Canada (1983). Details of the filtering and preparation of water samples for chemistry analyses are given in Wilson and Gajewski (2002). Loss-on-ignition (organic matter content) and carbonate content of the sediment were determined by drying 2-cm³ samples overnight at 95°C and then igniting in a muffle furnace at 500°C for 3 h and 925°C for 4 h, respectively (Bengtsson and Enell, 1986).

CHIRONOMID ANALYSES

Samples were processed by deflocculating the sediment in hot 10% KOH for 30 min and sieving on a 90 μm mesh. Head capsules were hand sorted with forceps, from a Bogorov counting tray, under a 40× magnification using a stereo microscope, and placed on cover slips to dry. Cover slips were later mounted on slides using Entellan®. Head capsules were extracted from the uppermost 1 cm first, followed by the 1 to 2 cm sample if necessary to obtain a minimum of 50 head capsules (Heiri and Lotter, 2001; Quinlan and Smol, 2001).

Taxonomic identifications were performed at 100× to 400× magnification under bright-field illumination and were based on Wiederholm (1983), Oliver and Roussel (1983), and Walker (1988, 2000). Taxonomy within the subtribe Tanytarsina followed Walker (2000). A number of specimens were encountered that bore a likeness to both the Tanytarsus lugens group and Corynocera oliveri type (both mentum and mandible features) and could not be placed reliably in either category. These were put in a separate grouping designated T. lugens/C. oliveri type. Corynoneura was separated from Corynoneura/Thienemanniella when sculpturing was visible on the head capsule surface (Wiederholm 1983), while Cricotopus and Orthocladius were largely indistinguishable and therefore grouped in many cases. The Figure 3 caption also refers to specific figures and their sources for the identification of some taxa (e.g., Parakiefferiella cf. sp. B, Parakiefferiella Fig. 367, and various Psectrocladius taxa).

DATA ANALYSIS

A few of the sites sampled in 1996 to 1998 were without dissolved organic carbon, transparency (Secchi), oxygen, and chlorophyll a (Chla) data. Mean values were substituted in each case (Tabachnick and Fidell, 1989) (Table 1). Chemical variables that had values below the detection limit in more than half of the study lakes were eliminated from all statistical analyses. This was the case for many of the metals, with the exception of Al, Fe, Mn, and Mo (Table 1). In other cases where values were below the limit of detection, the detection limit value was substituted. Environmental variables were tested for non-normal distributions using CALIBRATE version 0.70 (Juggins and ter Braak, 1997). All variables were log-transformed with the exception of surface and bottom-water temperature, pH, conductivity, alkalinity, dissolved organic carbon, and silicon, which were square-root-transformed. Total phosphorus and total filtered phosphorus remained highly skewed after transformation.

Ordinations were performed using CANOCO version 4.0 (ter Braak and Šmilauer, 1998). Both canonical correspondence analysis (CCA) and redundancy analysis (RDA) were used to assess the relationship between the chironomid assemblages and the measured physical and environmental variables because the length of the gradient of the first axis in a detrended correspondence analysis (DCA) was relatively small (2.2 standard deviation units) (ter Braak and Šmilauer, 1998). Taxon abundances were square-root-transformed for all ordinations, and rare species downweighted in DCA and CCA. RDA was performed with species centering and without standardization. A principal components analysis (PCA) of the environmental data, a DCA of the taxon data, and leverage diagnostics in CANOCO were used to check for outliers.

To reduce the presence of correlated environmental variables, we performed a series of constrained CCAs and RDAs where each environmental variable was selected as the sole variable, and the significance of the first axis tested using a Monte Carlo Permutation test. Only significant (P < 0.05, 500 permutations) variables were retained. Following this, variables were checked for high variance inflation factors (a VIF > 20 indicates that a variable is almost perfectly correlated with another [ter Braak and Šmilauer, 1998]) and eliminated sequentially, beginning with the variable having the highest VIF. This proceeded until all VIFs were below five. Forward selection was then used to identify those variables that explained most of the variance in the data. Variance partitioning (Borcard et al., 1992) was used to assess the strength and independence of those variables that explained the majority of the variance in the data, as determined by the series of constrained RDAs, as some of these variables were highly correlated with each other.

WATER CHEMISTRY

A summary of lake chemistry and physical properties is given in Table 1. More detailed descriptions are provided in Wilson and Gajewski (2002). In general the lakes are mostly shallow (mean depth = 5.2 m; maximum depth = 15 m) with nearly half (43%) having a maximum depth of 2 m or less (Table 1). Many of the lakes are closed basin, while a few have significant stream inflow (Hungry, Otter Falls, and Little Louise). Surface-water temperatures measured at the time of sampling ranged from 11°C to 20.5°C, although temperatures were rarely above 17°C. The lowest surface-water temperatures were recorded in the northern British Columbia lakes that are located in subalpine tundra (e.g., West Twin, Kusawak, and Three Guardsmen). Temperature profiles were uniform in all of the shallow (<5 m deep) lakes, while lakes of intermediate depth (5–10 m) had either isothermal or weakly stratified water columns by mid-July. Lakes that were >10 m deep had well-developed thermal stratification with epilimnion thickness of 6 to 7 m (not shown).

Alkalinity of the lakes was high compared to many boreal and subarctic regions (mean Alk = 147 mg L⁻¹; median pH = 8.7), and only 2 lakes had a specific conductance less than 100 μS cm⁻¹. This may be due to the presence of abundant glacial till in this part of the Yukon and to the evaporative enrichment associated with closed basin lakes, particularly those near Kluane Lake. Overall, the variability in conductivities among the lakes appears to be related to regional differences in geology, as well as differences in vegetation and the amount of stream inflow to lakes.

Most of the lakes are in the oligotrophic (<10 μg L⁻¹) (52%) and meso-eutrophic (10–30 μg L⁻¹) (40%) ranges, based on epilimnetic total phosphorus (TP) values (Table 1). Dissolved organic carbon (DOC) is variable as well (0.8–43 mg L⁻¹) and appears to be related to the relative abundance of grassland or tundra versus forest in the immediate vicinity of the lake, and possibly catchment area as well (Wilson and Gajewski, 2002). Sediment organic content (as determined by LOI) ranged from 9% to 80% with a mean of 32% (Table 1). LOI was negatively correlated with depth and highly correlated with bottom oxygen (not shown).

CHIRONOMIDS

A total of 5356 head capsules were enumerated from all lakes (mean per sample = 137; minimum = 58; maximum = 344), including those lakes that were later eliminated from the statistical analysis as outliers. Sixty-eight chironomid taxa were identified from the lakes, but only those present at ≥2% abundance in at least 2 lakes were included in the statistical analyses (Quinlan and Smol, 2001), leaving 38 taxa in total. A few lakes had chironomid populations overwhelmingly dominated by a single genus, such as Donjek Kettle (Chironomus > 80%) and Teapot Lake (Glyptotendipes > 70%) (Fig. 2). Psectrocladius D type was abundant only in Kusawak Lake (40% relative abundance), a shallow, low alkalinity, ultra-oligotrophic lake. A number of rheophilic genera including Rheocricotopus, Brillia, Euorthocladius, and Thienemanniella? were present only in Little Louise and Otter Falls lakes, at low relative abundance (each < 1.5%; not shown). These 2 lakes also had the only occurrences of Simuliidae (black fly) larval head capsules in their surface sediments, indicating stream or river influence (Currie and Walker, 1992).

The dominance of Glyptotendipes in Teapot Lake might be explained by a number of factors. First, Teapot Lake is relatively deep (12 m) with a very small surface area (6500 m²), where metalimnetic oxygen maxima of 150% to 200% and a very shallow (2–3 m), mostly anoxic (O₂ < 2 mg L⁻¹) hypolimnion have been observed over the past 2 summers (unpublished data). Total phosphorus is much higher in the metalimnion and hypolimnion (summer averages were approximately 23 and 100 μg L⁻¹, respectively) compared to the epilimnion (9 μg L⁻¹). The small volume of oxygenated hypolimnion that the above conditions allow would explain the near absence of profundal chironomids in the surface sediments compared to lakes of similar depth, and the dominance of littoral taxa. Chironomids living in the littoral zone of the lake could easily be transported down slope (e.g., Schmäh, 1993; Walker et al., 1993), especially given the small diameter and steep sides. Second, Glyptotendipes is known to thrive in mesotrophic-eutrophic and highly conductive environments (Saether, 1979; Walker, 1987; Walker et al., 1995; Lotter et al., 1998), which describes well the littoral environment of Teapot Lake (specific conductance = 616 μS cm⁻¹). Furthermore, the abundance of macrophyte remains in the anoxic, deep-water sediments of Teapot Lake suggests that these are plentiful in the littoral zone, providing a substrate for the many species of Glyptotendipes that mine the leaves and stems of aquatic plants (Weiderholm, 1983). Based on Weiderholm (1983), the head capsules found in the surface sediments probably belong to species group A, which includes G. barbipes, G. pallens, and G. paripes. Lods-Crozet and Lachavanne (1994) found the G. paripes/G. pallens group exclusively in the submerged macrophyte zone of Lake Geneva, and G. barbipes is characteristic of high conductivity lakes (Wiederholm, 1983). Anoxic conditions in much of the water column of Donjek Kettle might explain the dominance of Chironomus in this lake, as members of the genus are known to thrive in low-oxygen environments (e.g., Hofmann, 1988; Quinlan et al., 1998).

RELATIONSHIP BETWEEN CHIRONOMIDS AND LIMNOLOGICAL VARIABLES

The Otter Falls sample was eliminated from analysis because of significant river/stream inflow from a much larger, deeper lake; Otter Falls is actually a wide area of a river. Three Guardsmen Lake was eliminated because its sample scores were outside the 95% confidence limits in both the PCA and DCA. The variables Fe and Chla were eliminated from the final ordination analyses because of extreme influence (6.9× and 5.9×, respectively) in some lakes, as revealed with leverage diagnostics (ter Braak and Šmilauer, 1998). Because we had relatively few sites, we thought it preferable to eliminate these variables rather than the lakes from the ordination. Secchi depth was also eliminated from the final ordination because these measurements were often constrained by shallow lake depths, yielding inaccurate measures of transparency (Table 1).

The series of constrained CCAs and RDAs and forward selection with both ordination methods produced similar results (not shown). However, because the percentage of the total variance in the chironomid data explained by the first 2 axes and the explained variance of individual variables when constrained to axis 1 in RDA was slightly higher than with CCA (21% versus 20% using the 14 variables in Table 3), we present only the results for the RDA.

Of the 14 variables identified as statistically significant in the constrained RDAs, eight remained after elimination of those with the highest variance inflation factors—bottom water temperature (T_b), bottom oxygen (DO_b), depth, alkalinity (Alk), dissolved organic carbon (DOC), total phosphorus (TP), Mn, and loss-on-ignition (LOI) (Fig 3). The forward selection procedure identified LOI, TP, bottom water temperature, and alkalinity as the variables that could explain most of the variation in the chironomid data.

The first 2 axes of the RDA, consisting of the above 8 variables, were identified as statistically significant using Monte Carlo Permutation testing (P < 0.05, 500 permutations). These 2 axes explain 19.5% of the variation in the chironomid data (eigenvalues of 0.12 and 0.08, respectively). Loss-on-ignition, depth, and bottom water temperature were significantly correlated with axis 1, based on the t-values of the canonical coefficients (significant at P < 0.05), while TP and Mn were highly correlated with axis 2 (Table 2). LOI explained the majority of the variance in the data (9.8%), followed by bottom water temperature (7.7%), Secchi depth (7.5%), which was later eliminated, and depth (7.4%) (Table 3).

The results of variance partitioning between the three most significant environmental variables (as determined by constrained RDAs, but excluding Secchi depth) indicate that LOI and bottom water temperature can be considered statistically independent, as each variable captures a significant amount of variation regardless of the covariables used in the analysis (Table 4). By contrast, lake depth covaried with LOI and bottom water temperature when entered together as covariables. However, lake depth did contribute independently to the overall variance when LOI and bottom temperature were entered as single covariables (Table 4). LOI proved to be the strongest variable of the four, explaining greater than 7% of the total variance regardless of the covariables used.

These results are comparable to Olander et al. (1999) and Larocque et al. (2001), who also found sediment organic content (as measured by LOI) to be the chief explanatory variable of chironomid assemblages in subarctic lakes of northern Fennoscandia and northern Sweden, respectively. The LOI gradient represented by our Yukon lakes was relatively long (9% to 80%; Table 1) as was that of Olander et al. (17% to 88%) and Larocque et al. (3% to 87%). This may explain why Walker et al. (1991) did not find sediment organic content to be a factor determining the distribution of chironomids across treeline in Labrador, where the LOI gradient was much shorter (6.5% to 40.5%). A number of studies on chironomid ecology have pointed to the importance of substrate in determining species composition and abundance (e.g., McGarrigle, 1980; Pinder, 1986). However, the relative importance of sediment type versus other limnological variables in determining chironomid community composition has been the focus of debate (e.g., Warwick, 1989; Walker and Mathewes, 1989a; Walker et al., 1992). Often, sediment organic content is highly correlated with lake depth and surface water temperature as well, when lakes are sampled along strong latitudinal and altitudinal gradients (Walker et al., 1991; Lotter et al., 1997; Olander et al., 1999), making it difficult to separate the effects of these variables on chironomid communities. Surface water temperature was not a factor in determining chironomid assemblages in this study largely because lakes were not sampled along a strong climatic gradient. LOI was negatively correlated with depth (Fig. 3), although the results of the variance partitioning indicate that LOI contributed a large and statistically significant independent contribution to the overall variance when the effects of lake depth and bottom water temperature were partialled out. Olander et al. (1999) and Larocque et al. (2001) also found LOI to vary independently from both surface-water and/or air temperature and lake depth.

Other variables that were important in influencing chironomid composition in these lakes, as determined by forward selection, included total phosphorus, bottom water temperature, and alkalinity. Ion strength and composition are known to affect chironomid distributions (Pinder, 1986; Walker et al., 1995). Bottom water temperature is usually a measure of water depth and the degree of thermocline development, with the deepest lakes having the lowest bottom temperatures. Lake depth and bottom temperature were negatively correlated in this study (Fig. 3); however, T_b did have an independent influence on chironomid distributions that was unrelated to lake depth (Table 4). Water temperature has been shown to have a strong influence on chironomid communities (Walker et al., 1991; Olander et al., 1999), mostly through its effects on the growth and development of larvae (Oliver, 1971; Walker and Mathewes, 1989b). Numerous studies have shown that chironomids respond to changes in the availability of hypolimnetic oxygen caused by changes in the trophic conditions of lakes (e.g., Merilainen et al., 2000; Little et al., 2000; Francis, 2001). Few studies have identified TP as a strong variable for identifying these changes (e.g., Clerk et al., 2000), although Brodersen and Lindegaard (1997) show TP to be important in determining chironomid communities in shallow lakes where hypolimnetic oxygen is not a factor. Similarly, many of our sites were shallow with uniform oxygen profiles, although the TP values for our study lakes were much smaller than those of Broderson and Lindegaard (1997).

Although LOI seems to be a major determinant of the chironomid assemblages in these lakes, a large number of environmental variables control the distribution of larval chironomids, and some of these factors determine the nature of the sediment organic matter itself. The study of sediment type appears to be an efficient means to study the joint impact of many of the above variables, as the sediments reflect the interactions of nutrient levels, light penetration, aquatic productivity, wind and wave action, distance from shore, and vegetation type (McGarrigle, 1980).

ECOLOGY OF CHIRONOMID TAXA AND COMPARISON WITH OTHER REGIONS

The RDA biplot placed chironomids indicative of high sediment organic content (LOI) on the far left side of the diagram (e.g., Tanytarsus sp. C, Procladius, Tanytarsus pallidicornis type, Allo/Mesopsectrocladius, Psectrocladius spp., Microtendipes, Dicrotendipes, and Polypedilum), and taxa indicative of more minerogenic sediments on the far right (e.g., Chironomus, Paratanytarsus, Sergentia, Corynocera oliveri type, Cricotopus/Orthocladius [C. tremulus type], and Micropsectra atrofasciata type) (Fig. 3). The first axis also represents a gradient of lake depth, with shallow lakes occurring on the left and deeper lakes on the right. Chironomids characteristic of the deepest lakes with cold hypolimnetic or bottom waters are positioned in the upper right quadrant (e.g., Sergentia, Corynocera oliveri type, Micropsectra atrofasciata type, Heterotrissocladius, Eukiefferiella, Tanytarsus lugens/C. oliveri type) while chironomids typical of the shallow lakes with higher bottom water temperatures are positioned in the lower left quadrant (e.g., Microtendipes, Polypedilum, Cladopelma, Dicrotendipes, Tanytarsus lugens group, Cladotanytarsus mancus group, and Psectrocladius E/F type). This latter group of taxa is also characteristic of high nutrient (TP and DOC) lakes (Fig. 3). Shallow lakes with warm bottom temperatures are typically more productive, having an abundance of algae and aquatic plants that yield higher organic sedimentation—hence, the position of these lakes on the left side of the ordination in the direction of increasing LOI.

The above trends are comparable to other studies of modern chironomid assemblages. Heterotrissocladius, Sergentia, and Micropsectra are common in the profundal zone of deep, thermally stratified lakes in temperate regions (Walker, 1991). Heterotrissocladius and Micropsectra are more typical of oligotrophic lakes (Saether, 1979), although Micropsectra has been known to occur across a trophic gradient (Lods-Crozet and Lachavanne, 1994) and was common in mesotrophic lakes in the Swiss Alps (Lotter et al., 1998). Sergentia was most abundant in oligotrophic (<10 μg L⁻¹ TP) lakes in our data (Table 1, Fig. 3), although it commonly thrives under mesotrophic conditions (moderate oxygen depletion) (Walker, 1991; Quinlan et al., 1998). However, as very few deep, mesotrophic lakes were sampled in this study, we may simply have missed the preferred niche of this genus. The shallow-water/littoral taxa observed in our lakes are also typical of other regions (e.g., Schmäh, 1993; Walker and MacDonald, 1995; Korhola et al., 2000). One exception may be the Tanytarsus lugens group, which in our data was common in shallow, productive lakes (Figs 2 and 3), but is considered an oligotrophic taxon (e.g., Saether, 1979; Brundin, 1956, in Hofmann, 1986), and a cold stenotherm (Hofmann, 1988). T. lugens was also present in many oligotrophic lakes of northern Fennoscandia (Korhola et al., 2000). However, Lotter et al. (1998) show that T. lugens group is common in mesotrophic lakes from the Alps, and it was a dominant taxon in the mesotrophic profundal zone of Lake Geneva (Lods-Crozet and Lachavanne, 1994). Alm and Willassen (1993) and Hofmann (1988) also note that it occurs as a profundal species in oligotrophic to weakly eutrophic lakes of middle Europe but can occur in the littoral zone of subarctic lakes where temperatures are lower. These apparent differences in ecology may be related in part to taxonomic uncertainties. Difficulties can arise in distinguishing C. oliveri and T. lugens as noted above (T. lugens group and C. oliveri type grouped in some cases) and by Alm and Willassen (1993). C. oliveri is considered a cold stenotherm (Pinder and Reiss in Olander et al., 1999) and did occur in the deepest lakes with the lowest bottom water temperatures in our sites (Figs. 2 and 3). However, given the lack of ecological information on littoral chironomid communities (Hofmann, 1986), it is possible that T. lugens has a more widespread trophic and bathymetric distribution than previously documented.

The chironomid distribution with respect to LOI in our data compares well with Olander et al. (1999). Dicrotendipes, Psectrocladius, Allo/Mesopsectrocladius, Microtendipes, and Polypedilum occur in highly organic sediments, and Heterotrissocladius, Micropsectra, and Corynocera oliveri type occur in more minerogenic sediments in both regions. Similarly, Larocque et al. (2001) found Dicrotendipes, Polypedilum, and Microtendipes in more organic sediments, as well as Tanytarsus sp. C and Tanytarsus pallidicornis type, which were prominent in the most organic-rich sediments in our study lakes (Figs. 2 and 3). Heterotrissocladius and Micropsectra were also more abundant in Swedish lakes with lower LOI values. One major difference was the dominance of Chironomus in Yukon lakes with the least organic sediment, compared to the Fennoscandian and Swedish lakes, where it was most abundant in highly organic lake bottoms. McGarrigle (1980) also found species of Chironomus to be more abundant in sediment with high organic content. The three Yukon lakes where Chironomus is most abundant (Donjek, Rat, and Small) have low to moderate oxygen concentrations near the lake bottom (Table 1), and in the case of Small Lake, stable oxygen readings were difficult to obtain. It is possible that low oxygen is responsible for their success in these lakes and that the low organic matter content is simply a coincidence. Chironomus was dominant in Donjek Kettle, and this site undoubtedly has a large influence on the relationship of LOI with Chironomus abundance. Donjek Kettle receives large inputs of loess from the Donjek River valley. Perhaps the loess input dilutes the organic matter in the sediment, while the lake otherwise has conditions suitable for Chironomus survival.

This study confirms the importance of sediment organic matter in determining the distribution of chironomid populations in lakes from the southwest Yukon. Sediment organic matter in turn depends on the productivity of the lake in addition to watershed conditions. Future work will investigate the coupling of these ecosystems and the factors causing them to change in time and space.