Seasonal fish community variation in headwater mangrove creeks in the southwestern Everglades: an examination of their role as dry-down refuges

Home >Methods Results Discussion Acknowledgments Literature Cited Tables and Figures PDF Version

Site Description. --We sampled the large and small fish community in the oligohaline to mesohaline headwater reaches of six creeks in the RB region and six creeks in the NW region (Fig. 1). RB sites included four creeks in the main stem of Rookery Branch, as well as Squawk and Otter creeks (RB7 and RB12, respectively). NW sites were located along three Watson River creeks and three North River creeks. Creeks in NW drain shorter hydroperiod marshes than RB creeks. By hydroperiod, we refer to the number of days the marsh is flooded in a yearly cycle. Marshes are typically considered dry if water levels drop below 5 cm, at which depth little standing water remains (Loftus and Eklund, 1994). According to data from hydrologic stations P35 and P38 (Fig. 1), over the past 20-yr period (1986-2006), the hydroperiod averages 332 d of flooding (± 8.3 d) in marshes upstream of RB creeks, and 305 d (± 13.5 d) in marshes upstream of NW creeks. Nutrient concentrations are similar between regions: relatively high nitrogen (approximately 1 mg L^-1) and low phosphorus concentrations (below 0.02 mg L^-1) are characteristic of both RB and NW waters (Levesque, 2004).

Sampling Effort. --All sampling was conducted in the main channel of headwater creeks and in the uppermost boat-accessible 600 m reach of each creek. Sampling included only first and second order creeks (Strahler, 1957). Creek shorelines were vegetated by riverine mangrove forests dominated by red mangrove, Rhizophora mangle Linnaeus, 1753 (Lugo and Snedaker, 1974). Creek depth at sampling locations averaged 1.37 m (± 0.03 m, n = 108); width averaged 10.8 m (± 1.0 m, n = 107). Sampling was conducted during November 2004, February 2005, and April 2005, corresponding to the wet season, the transition between wet and dry seasons (hereafter "transition"), and dry season. Daily marsh water level and creek salinity measurements were obtained from the nearest National Park Service (NPS) hydrologic stations (Fig. 1 and Fig. 2). While sampling, we measured salinity at creek sites with a YSI® 85 unit (Fig. 3).

Large-fish Sampling. --Large fishes (55-750 mm standard length, SL) were sampled using a boat-mounted electrofishing unit (two-anode, one-cathode Smith-Root® generator-powered pulsator 9.0 unit rated to a maximum salinity of 15). Electrofishing has been shown to be an effective method for sampling larger fishes in other Everglades habitats (Nelson and Loftus, 1996; Chick et al., 1999). At each creek, sampling was conducted in three 5- min (pedal time) bouts (three bouts * six creeks * two regions * three seasons = 108 electrofishing samples). For all bouts, electrofishing power was standardized to 1500 watts according to temperature and salinity conditions (Burkhardt and Gutreuter, 1995). On average, each bout sampled 122.6 m (± 2.8 m, n = 108) of creek shoreline. Bouts were distributed evenly over the 600-m segment of creek, so that each bout was considered an independent sampling unit. For each bout, we randomly selected a creek shoreline and made a single pass with the electrofishing boat. All fish captured were identified to species, measured to nearest mm SL, weighed to nearest g, and released after full recovery. Non-indigenous species were collected and brought to the laboratory for processing.

Figure 2. (A) Mean daily salinity and (B) water levels collected by the nearest four NPS monitoring stations to study headwater creeks. See Figure 1 for approximate location of stations in reference to sites. Bold lettering indicates sampling months. Dotted line indicates 5-cm water depth cutoff used for calculation of marsh hydroperiod (see text for explanation). [larger image]

Small-fish Sampling. --Small fishes (< 50 mm SL) were sampled with 3-mm, metal-mesh minnow traps (25.4 mm opening) deployed unbaited, overnight along creek banks. Minnow traps are a commonly used and easily replicable sampling device, but it suffers from several sampling biases (Rozas and Minello, 1997), one of which is trap placement. Minnow traps are typically set on the substrate, where they are unlikely to be encountered by water-column or surface dwellers (Layman and Smith, 2001). In this study, we deployed minnow traps in pairs; one set on the substrate and a second suspended just beneath the water surface, secured to mangrove prop roots. In each creek, we deployed three pairs of traps during the November 2005 sampling event, but increased effort to five pairs for subsequent sampling events (November: six traps * six creeks * two regions = 72 samples, February and April: 10 traps * six creeks * two regions x two seasons = 240 samples; total sample size is 312). Fish captures from minnow traps were preserved in 10% formalin and brought to the laboratory for processing.

Statistical Analyses. --We examined variation in the abundance of fishes among regions and creeks and as a function of season with nested, repeated-measures ANOVA or ANCOVA models. Season was the repeated measure in our analyses, and nesting allowed us to account for spatial variation among regions (RB and NW); creeks were nested within regions. Focal response variables included: CPUE for the large fishes caught in electrofishing (number 5-min^-1 pedal time) and gill nets (number 30-min^-1 soak time), CPUE for the small fishes caught in minnow traps (number 24 h⁻¹), and the proportion of CPUE that was freshwater in electrofishing and minnow trap samples (CPUE was too low in gill net samples). Species were classified as either marine, estuarine, or freshwater (Table 1) based upon their habitat occurrence (per Loftus and Kushlan, 1987; Loftus, 2000). Preliminary analyses examined seasonal and spatial variation in the number of species caught in all gears, but results were indistinguishable from analyses of CPUE; and thus, are not presented here.

Table 1. [click here to see entire table]

---

Figure 3. Salinities over the three sampling seasons in the two study regions: the North and Watson rivers and Rookery Branch. Shown are means ± 1 standard error (SE). [larger image]

A two-way ANOVA was used to examine seasonal and spatial variation in salinity levels. Salinity was used as a covariate in analyses of the large fish data; no salinity measurements were made at the time of minnow trap deployment. To better satisfy assumptions of parametric tests, CPUEs were ln (observed value + 1)-transformed and proportions were subject to angular transformations prior to analyses. Post-hoc pairwise comparisons were performed using Tukey-corrected contrasts. If salinity was a significant covariate, simple linear regressions were used to examine the relationship between response variables and salinity. All analyses were performed using Proc Mixed in SAS Version 9.1.3^®.

We used analyses of similarity (ANOSIM) based on Bray-Curtis similarity matrices to test for effects of region, season, and gear (electrofishing vs gill net, and top vs bottom minnow trap) on fish community structure (Clarke and Warwick, 2001). Dissimilarity matrices were constructed based on ln (observed value + 1)-transformed estimates of the relative abundance of all taxa in samples, except for the gear comparison of gillnets and electrofishing, where a presence/absence matrix was used. Analyses included 28 taxa from electrofishing samples, 10 taxa from gill nets, and 22 taxa from minnow traps (Table 1). We followed ANOSIM analyses with percentage of similarity analyses (SIMPER) to determine which taxa contributed most to groupings observed among samples. We constructed non-metric multi-dimensional scaling (NMDS) plots to illustrate dissimilarity among groups. In these plots, the distance between data points is proportional to the degree of similarity between samples. All community structure analyses were conducted using Primer© Version 5.2.9.