Among the 21 species of zooplankton Daphnia pulex, D. similis, D. rosea, D. galeata mendotae, Diaphanosoma birgei, and Ceriodaphnia lacustris were the most abundant cladocerans. Leptodiaptomus sicilis and Diaptomus nevadensis were abundant calanoid copepods, and Diacyclops thomasi was the most common cyclopoid species. The brachiopod Artemia franciscana dominated in hypersaline lakes. Littoral sampling resulted in 18 classes of macroinvertebrates, with amphipods (Hyalella azteca and Gammarus lacustris) and zygopterans being most abundant. Yellow perch (Perca flavescens), nine-spine stickleback (Pungitius pungitius), Brook stickleback (Culaea inconstans), fathead minnow (Pimephales promelas), northern pike (Esox lucius), and walleye (Sander vitreus) were the most common fish species.
Stable isotope patterns
The 20 study lakes showed a broad range of lake-specific variation in stable isotope values of individual taxa over the two-year study period. Linear regressions demonstrated that the variability in δ13C was mostly dependent on nutrient levels and productivity measures (Table ). For cladocerans, the most important parameters were TP, SRP, and Chl a, and variability in δ13C for copepods was related to Chl a, SRP, Secchi depth, and TP. Amphipod variability in δ13C had no significant correlation to any parameters. For δ15N, cladoceran values had a significant positive relationship to surface area, while Secchi depth had a significant negative effect. The δ15N values of copepods showed a positive correlation to Chl a. Amphipod variability in δ15N values had significant relationships with SRP, Chl a, and TP. Stepwise multiple linear regressions determined that TP and fish complexity had significant positive influences on Cladoceran δ13C, while Chl a had a positive correlation to Copepod δ13C (Table ). For Cladoceran δ15N, Secchi depth showed a significant negative correlation and lake surface area had a significant positive relationship, while SRP had a significant positive correlation to amphipod δ15N (Table ). Copepod δ13C and Amphipod δ15N were not significant correlation to any environmental parameters.
Pearson correlation coefficients (r) between environmental parameters and temporal variation (range) of δ13C (C) and δ15N (N) for cladocerans (clad), copepods (cope) and amphipods (amph) over the two year period
Stepwise multiple linear regression results for temporal variation (range) in δ13C and δ15N of cladocerans, copepods and amphipods as a function of environmental parameters
Circular statistics revealed minor, yet systematic temporal changes in δ13C and δ15N. Intra-annual variation (June to August) showed a small significant shift to higher δ13C values for zooplankton and macroinvertebrates, indicated by mean vectors of change (μ) of 61.0° in 2007 (Rayleigh's test, p < 0.01, n = 36; Figure ) and 101.2° in 2008 (Rayleigh's test p < 0.01, n = 33; Figure ). As the angle of change was largely associated with the X-axis (δ13C), no systematic intra-annual changes were obvious for δ15N (Y-axis). Yet in both years, several lakes had lower δ15N during August, shown by individual angles of change (θ) associated with negative values along the Y-axis (Figure and ). Inter-annual variation was identified only for the June sampling, which showed a small significant increase in δ15N from 2007 to 2008, with a mean vector of change (μ) of 345.0° (Rayleigh's test p < 0.01, n = 33; Figure ).
Figure 2 Results of circular statistics to quantify direction and magnitude of temporal variation (Schmidt et al., 2007) of δ13C and δ15N values in the 20 study lakes for selected bioata. Periods of temporal change were either seasonal: (A) June (more ...)
Figure 3 A and B: Stable isotope bi-plots for the 20 study lakes, with δ15N on the Y-axis and δ13C on the X-axis. Lakes are sorted by salinity (see Table 1), increasing within rows from row 1 to row 10. Data are averaged over the four sampling (more ...)
Food-web dynamics - Based on stable isotope data, freshwater lakes had the most complex food webs, containing the greatest number of trophic levels as well as the most taxa (Figures and ). Cladocerans (Daphnia sp., Ceriodaphnia sp., Bosmina sp., and Diaphanosoma sp.) occupied the trophic position of the primary consumer within the pelagic habitat (low δ13C), and invertebrate predators such as Chaoborus and Leptodora were secondary pelagic consumers. Fathead minnows, as well as small (< 10 cm) walleye and yellow perch were predominantly zooplanktivorous as evidenced by intermediate δ15N and low δ13C values. In the littoral habitat (high δ13C), amphipods were the most common primary consumer, with predatory insect larvae (e.g. Zygoptera) as secondary consumers. Intermediate sized perch (10-15 cm), shiners, and Brook and nine-spine stickleback foraged more in the littoral zone and, at times, acted as tertiary consumers in the littoral. Larger piscivorous fish (walleye, pike, and perch > 15 cm) were top-predators in freshwater lakes and often incorporated both pelagic and littoral prey into their diet, as shown by intermediate δ13C and high δ15N values.
Figure 4 A and B: Stable isotope bi-plots for the 20 study lakes, with δ15N on the Y-axis and δ13C on the X-axis. Lakes are sorted by salinity (see Table 1), increasing within rows from row 1 to row 10. Data are averaged over the four sampling (more ...)
In contrast, saline lakes had simplified food webs and strongly reduced taxonomic richness (Figure ). Fishes were absent in mesosaline and saline lakes, except Redberry Lake. Instead, the highest trophic levels in the pelagic and littoral areas were occupied by the predatory copepod D. nevadensis and zygopterans, respectively, even in Redberry Lake, where sticklebacks had slightly lower δ15N values than D. nevadensis. In hypersaline Snakehole Lake, primary consumers represented the highest trophic level in both pelagic and littoral habitats, as only halophilic A. franciscana and larvae of the brine fly Ephydra sp. were encountered. In the second hypersaline lake, Little Manitou, taxa that are usually associated with the littoral zone (e.g. harpacticoids) were isotopically indistinguishable from pelagic species such as A. franciscana and L. sicilis.
Littoral invertebrate predators (e.g. Zygoptera, Dysticidae, and corixids) did not actively invade the pelagic zone after the exclusion of fishes to take over the position of the top predator. Even though these taxa were frequently encountered in pelagic samples, based on their stable isotope values they continuously relied on littoral prey, indicating passive dispersal rather that an active migration into the open water.
Finally, in three lakes (Kipabiskau, Humboldt and Redberry), we consistently encountered copepods that had low δ13C values relative to all other pelagic crustaceans. These δ13C values were also more negative than δ13C of primary producers that were analyzed as particulate organic carbon (POC, data not shown).
Stable isotope metrics
The differences in food-web structures among prairie lakes derived from bi-plots were confirmed by community-wide metrics. Freshwater lakes had a large δ13C range (CR; 4.9 - 9.4, mean: 6.9), compared to mesosaline (CR; 3.6 - 8.8, mean: 5.5), and saline lakes (CR: 2.0 - 5.6, mean: 3.9) (Table ). Freshwater lakes also had the highest δ15N ranges (NR: 5.2 - 10.8, mean: 7.7; Table ). Redberry Lake had the highest NR for mesosaline lakes (5.8), while Snakehole Lake had the lowest NR among all lakes (0.7). Total area (TA) showed the greatest differences between fresh and saline systems. Freshwater lakes had a TA range of 13.7 to 49.2 (mean: 34.6; Table ). In contrast, mesosaline lakes ranged from 6.8 to 25.5 (mean: 13.6) and saline lakes ranged from 2.8 to 14.1 (mean 7.6). The mean centroid distances (CD, indicating trophic diversity) in freshwater lakes (1.9 - 3.6, mean: 2.8) were slightly larger than in mesosaline lakes (1.7 - 3.4, mean: 2.3) and saline lakes (1.1; 2.0, mean: 1.7), while saline lakes had greater mean nearest neighbor distances (NND, trophic redundancy). The standard deviations of NND (SDNND; Table ) were not significantly different among the lakes.
Table 4 Summary of community-wide metrics based on δ13C and δ15N analysis  for the 20 study lakes
Freshwater lakes are listed first, followed by mesosaline and saline lakes. A large CR (max δ13C - min δ13C) implies a significant difference in basal resources (littoral-pelagic). A large NR (max δ15N - min δ15N) demonstrates the trophic level variation. Total area (TA) encloses all species within the bi-plot space, representing total trophic diversity within the lake. Centroid distance (CD) is the average Euclidean distance of each species to the mean δ13C-δ15N value for all species in the bi-plot, measuring the average trophic diversity within the food web. Mean nearest neighbor distance (NND) is the average Euclidean distance of each species to its nearest neighbor and measures overall species packing. Standard deviation of NND (SDNND) measures evenness of species packing
Analysis of Variance (SPSS version 16.0) determined that the differences in metrics were significant for CR, NR, TA, CD, and NND (p < 0.05 for all) between freshwater and mesosaline lakes and freshwater and saline lakes.
Community wide metrics changed predictably with environmental variables (Table ). CR had a positive correlation to fish complexity, salinity and TKN. NR had a significant negative association with salinity, fish complexity, TKN, DOC, and TP. Total area had a significant positive correlation to fish complexity and significant negative correlations to salinity, TKN, and DOC. Centroid distance had a significant positive correlation with fish complexity, and negative correlations with TKN, salinity, and TP. NND had positive correlations to DOC and salinity and a negative correlation to chlorophyll a concentration. Standard deviation of NND was not significantly correlated to any environmental parameters. Stepwise multiple linear regressions determined that salinity had a negative influence on NR, while fish complexity had significant positive correlations to CR, TA, and CD (Table ).
Pearson correlation coefficients (r) of the environmental influence on community-wide metrics
Results of stepwise multiple linear regression coefficients of the environmental influence on the community-wide metrics