Community composition of zooplankton exported from a shallow polymictic reservoir linked to wind conditions

Study site

Palatine Lake (39.541747°N, −75.169077°W; Fig. 1; Table 1) is a small polymictic reservoir located in Pittsgrove Township, Salem County, New Jersey (USA) that is owned and operated by the Palatine Lake Village Homeowners Association (PLVHA). The PLVHA gave us permission to use their facilities for this study and to sample the lake/outfall. Palatine Lake is fed primarily by Muddy Run and Palatine Branch, which together drain a watershed composed mostly of agricultural lands (53.0%) and wetlands (30.6%), but there are also relatively large areas of forest (7.2%) and urban (10.8%) habitat. Muddy Run continues downstream of Palatine Lake after passing through one of two outfalls on the dam, both of which are in continuous (concurrent) operation. All samples in the study presented here were taken from one site located in the confluence of the two outfalls (Fig. 1). The confluence pool has a natural bottom and edge that is pool-like, but the reservoir outfalls induce a strong mid-stream current that is run-like. Under base-flow conditions at the sampling location, the confluence has a width of 23.9 m, mean cross-sectional depth of 100.1 cm, and maximum cross-sectional depth of 158.5 cm. The maximum depth for the entire confluence is 214.0 cm.

Field methods

Thirteen bi-hourly samples were collected during each of three 24-h periods: 12–13 June, 10–11 July, and 8–9 August 2017. Each bi-hourly sample is an integrated sample of five discrete casts of a bucket into the middle of the outfall stream (into the current) such that the benthos was not disturbed. The discrete casts were all collected within 5 min of each other and filtered through a 63-micron Nitex mesh. The total volume of water filtered for each bi-hourly sample was 38 liters. Samples were preserved with a stepped preparation of ~50% isopropyl alcohol in the field followed by further concentration and preservation in the lab to ~90%.

For each bi-hourly zooplankton sample, corresponding environmental data were collected either on-site or from a nearby environmental monitoring station (more details on specific sampling devices are given in Table 2). On-site atmospheric data were collected using a handheld anemometer measuring maximum wind speed (Variable: Maximum Gust) and air temperature. A variety of handheld meters were used to collect in-situ physical–chemical water quality data including dissolved oxygen (DO), water temperature, and pH. Conductivity, total dissolved solids, and turbidity were measured ex-situ within 20 min of sample collection.

Off-site weather data were obtained from a weather station located 3.28 km to the southwest (39.524351°N, −75.200927°W) of our sampling location; the weather station is operated by the Rutgers University Agricultural Extension (Upper Deerfield, NJ Mesonet Station; https://www.njweather.org/station/284). The weather station variables we used were precipitation, total solar radiation, mean solar radiation, wind speed, and wind direction (Table 2). Numerical weather station data were collected at 5-min intervals, binned at a variety of time-periods relative to when on-site zooplankton sampling occurred (30 min concurrent with sampling and at 30, 60, and 120 min leading up to zooplankton sampling), and summarized (maximum, mean, or sum). Binning of data at a variety of time lags was done to account for possible discontinuity in timing between environmental conditions at the weather station and the response of zooplankton exports to changing conditions.

Wind direction data (categorical headings) obtained from the weather station were reported as a single direction (1 of 16 possible directions) for a 5-min period (a 5-min mean). These categorical wind directions were transformed to measures of variability in the direction of the wind (directionality) by expressing the data as (1) a count of how many different wind directions were observed prior to sampling (Maximum = 16; Variable: Count Wind Directions, Table 2) and (2) as a count of how many different wind directions were observed prior to sampling divided by the number of observations (Variable: Mean Wind Directionality, Table 2). The number of observations (5-min means) varied by the time period being considered (binning by 30, 60, or 120 min). For both wind directionality variables, higher values indicate winds blowing from an inconsistent direction whereas lower values indicate the wind is blowing from a consistent direction.

Hydrological data were collected from a USGS stream gauge in an adjacent watershed 9.4 km to the southeast (39.495750°N, −75.076496°W) of our sampling location; these data include mean and maximum daily discharge (given as ft³/second and fitted as m³/second) for the week (7 days) leading up to sampling.

Lab methods

Field-preserved samples were concentrated using a 63-micron Nitex mesh and preserved at a final concentration of ~90% isopropyl alcohol. The water volume filtered (38 L) and the final concentrated sample were standardized across all samples (n = 39; 13 bi-hourly samples per 24-h period) to allow for comparisons of abundance between samples. Zooplankton were identified according to Haney (2019) under 40× magnification using a one mL Sedgwick-Rafter counting chamber that was loaded from a well-mixed sample (McCauley, 1984). For samples with low abundance, the sub-sample was drawn entirely from the sample (one mL of mixed sample directly into the counting chamber). For samples with high abundance, the sub-sample was diluted in the counting chamber (e.g. 1:1 = 0.5 mL sub-sample + 0.5 mL isopropyl alcohol). For each sample, at least six sub-samples were counted. Dilutions of 1:1 and 1:3 were typical dilutions (maximum dilution 1:20 for one sample). In the final analysis density was standardized to individuals/liter.

Zooplankton were identified to the following taxonomic groups: copepod nauplii (Naup), adult copepods and copepodites (Cop), Bosmina (Bos; a cladoceran genus), Chydorids (Chy; a cladoceran family), other cladocerans (Clad), rotifers (Rot), midges (Midge; Chaoborus, a genus of planktivorous midge larvae that are members of the zooplankton), and ostracod (Ostra). These taxonomic groups were chosen a priori based on a combination of factors including confidence of taxonomic identification at 40× magnification, minimizing the undue influence of rare species on ordinations, and on known behavioral/ecological responses to environmental conditions.

Statistical methods

The role of environmental variables in structuring CCZE was explored via ordination of the export community (indirect gradient analysis approach; non-metric multi-dimensional scaling (NMDS); metaMDS function in the Vegan package of R; Bray–Curtis dissimilarity). For each of the three 24-h periods, temporal shifts in CCZE were visualized by plotting bi-hourly samples with an NMDS ordination and then vector-fitting environmental variables to the underlying ordination to independently assess their predictive power for CCZE. In these analyses (for each 24-h period), space is held as constant (the same location is repeatedly sampled) and time is assumed to vary (points in the ordination diverge because of differences in time). Environmental variables (Table 2) were fit to the NMDS ordination using the envfit function in the Vegan package of R. Some turbidity values during 10–11 July as well as conductivity and TDS values during 8–9 August were not useable, so these variables were not fitted to the corresponding 24-h ordinations or to the combined seasonal (comparing 24-h periods) ordination detailed below. The hydrological data were not fit to NMDS ordinations because of the potential for discontinuity between measurements from a different watershed at nearly 10 km distance from the location where CCZE was being assessed. Likewise, species vectors are not displayed in the NMDS ordinations because it is difficult to interpret species vectors relative to environmental vectors in an unconstrained ordination.

Seasonal (between sampling period) differences in the CCZE were explored for each taxonomic group via Kruskal–Wallis test by ranks with Steel-Dwass pairwise tests (JMP 13, SAS Institute) and a NMDS ordination using all bi-hourly samples (all three 24-h sampling periods combined; R). This ordination contrasts samples taken at two different time scales (bi-hourly and monthly across 24-h sampling periods) and has utility for assessing differences in CCZE between months (factor-fit). The fit of environmental variables to this ordination was assessed conservatively using α = 0.001 instead of 0.05 to avoid a type 1 error associated with using data collected at multiple scales in the same ordination. Comparison of hydrological data between 24-h sampling events (months) was conducted using Kruskal–Wallis test by ranks and Steel-Dwass pairwise tests, α = 0.05.

Fit of environmental variables to CCZE

Non-metric multi-dimensional scaling ordinations were plotted in two dimensions (stress: June = 0.112; July = 0.106; August = 0.119) and environmental explanatory variables were fit to the underlying ordination (see Appendix Table A1, for fitting statistics and p-values for all environmental variables). No light variables were significant predictors of CCZE during any 24-h period (June, July, or August). Mean wind directionality measured concurrently with on-site zooplankton sampling was correlated with the zooplankton community ordinations in both June (Fig. 2A, r² = 0.535, p = 0.023) and July (Fig. 2B, r² = 0.510, p = 0.031). Additional wind directionality variables were significantly correlated with the June ordination (Fig. 2A; 60-min mean wind directionality, r² = 0.496, p = 0.031; 120-min mean wind directionality, r² = 0.574, p = 0.012) and July ordination (Fig. 2B; 30-min count wind directions, r² = 0.476, p = 0.039; 120-min count wind directions, r² = 0.515, p = 0.029). No measured environmental variables were significant predictors of CCZE in August.

Differences in zooplankton exports between 24-h periods (months)

Zooplankton density (all taxa of zooplankton combined) was highest in July (significantly higher in July than in June, p = 0.008), but was highly variable between samples and between 24-h sampling periods (months; Tables 3 and 4). Naupliar density increased significantly across months while midge density decreased. Peak density for most groups occurred in July; some groups (Bosmina and cladoceran) were significantly more common in July samples than in either June or August samples (p < 0.001). Copepods significantly increased in density from June to July but, while density was much lower in August than in July, this decrease was only marginally significant (p = 0.079). Midge and ostracod density decreased significantly between the July and August sampling periods, while rotifer and chydorid groups did not differ between months.

For the seasonal (all months combined) NMDS ordination (stress = 0.165), CCZE varied significantly between months (Appendix 1 and Fig. 3; months, r² = 0.652, p = 0.001). A few environmental variables were significant predictors of the NMDS ordination at the alpha = 0.05 level, but only physical–chemical parameters were significant at an adjusted alpha of 0.001 (DO, r² = 0.320, p < 0.001; pH, r² = 0.397, p < 0.001; water temperature, r² = 0.356, p < 0.001). Hydrological conditions were significantly different between monthly 24-h sampling events (Fig. 4; Kruskal–Wallis comparing 24-h sampling period between months, p < 0.001; absolute value of Z in Steel-Dwass test; June–July: Z = 3.066, p = 0.006; June–August: Z = 2.814, p = 0.014; July–August: Z = 3.06998, p = 0.006).