2010 SOUTHEASTERN NATURALIST 9(4):649–672 Fish Assemblage Variability in a Florida Spring Kirsten Work1,*, Melissa Gibbs1, Brenda Peters1, and Laura French1 Abstract - Florida springs are generally characterized as static ecosystems with unique faunal and floral assemblages that persist under relatively stable physical and chemical conditions. We sampled the fish fauna of Volusia Blue Spring to determine whether this presumption would withstand scrutiny at a higher temporal resolution and over time. We sampled by seining or snorkeling at five stations along the 320-m run weekly or bimonthly from October 2000 to September 2004. This four-year study consisted of 1152 samples that produced approximately 164,000 observations of 30 species of fish on 72 sampling trips. Concurrent water quality samples were collected at 14 sites along the center of the run and at each of the seine sites. Virtually anoxic water discharged from the spring head, but this water accumulated oxygen as it traveled the run. Fish density and species composition also changed dramatically along the length of the run. Species that tolerate low oxygen concentrations, such as poeciliids, dominated the assemblage at the spring head. Species that use patches of algae or small backwater areas, such as fundulids, were prominent in the middle reach of the run. Larger species, such as centrarchids and Lepisosteus spp., were abundant in the lower reach of the run. Within these broad patterns, most species exhibited great variability in density, probably due to the influence of variable emigration of potential predators, and also perhaps smaller species, from the St. Johns River. Introduction In most streams, fish species are patchily distributed and segregate with respect to an array of abiotic and biotic factors, such as water depth and velocity, substrate type, dissolved oxygen concentrations, temperature, aquatic plants and other structure, and competitors or predators (Dibble and Harrel 2000, Gorman and Karr 1978, Kessler et al. 1995, McKinsey and Chapman 1998). These patches may shift with changes in season or with periodic disturbances, such as floods or droughts. Variation in water inputs often causes major changes in the morphometry and water chemistry of streams, which in turn affect stream fish assemblages. Seasonal and periodic changes in water input and associated physical and chemical parameters that are typical of streams are not present in most natural springs. Springs are comparatively stable environments, with little variation in temperature and water chemistry, primarily because spring water originates from large underground aquifers (Hubbs 1995, Whitford 1956). In contrast to the diversity of abiotic and biotic factors that influence stream fishes, low dissolved oxygen concentration has been considered one of the most important factors that control longitudinal 1Stetson University Biology Department, DeLand, fl32724. *Corresponding author - kwork@stetson.edu. 650 Southeastern Naturalist Vol. 9, No. 4 fish distribution patterns in springs (McKinsey and Chapman 1998). In many springs, water issuing from the spring head is anoxic or hypoxic. Dissolved oxygen concentrations in Florida springs often range from zero to 2.6 mg/L and average under 1 mg/L during the day throughout the year (McKinsey and Chapman 1998), whereas the dissolved oxygen concentrations of unpolluted temperate lakes and rivers average from 7 to10 mg O2 L-1 during daylight hours (Matthews 1998). Oxygen concentrations as low as those recorded in Florida springs have been associated with large fish kills in eutrophic lakes that experience a rapid depletion of oxygen to 3–4 mg L-1 (Bennett 1971). In springs, upstream fish assemblages exhibit more specialization to the oxygen-poor waters (e.g., upturned mouth, small body size, aquatic surface respiration) than the downstream assemblages, which experience a relatively oxygen-rich habitat (Hubbs 1995). Many of the habitats found in streams also are present in springs (highflow center-channel areas, low-flow backwater areas, snags, submerged vegetation); however, the constant flow and low dissolved oxygen of springs affect the suitability of these areas as fish habitat. In streams, center-channel, high-flow areas usually are more oxygenated than backwater areas that receive loads of detritus, as decomposition of this detritus may deoxygenate the water. During low water, streams may be reduced to center-channel pools and riffles that are only marginally or seasonally connected, restricting fish movement along the stream (Matthews 1998). During high water, fish may inhabit backwater areas to escape high, scouring flows. In contrast to streams, the center-channel high-flow areas of springs may possess the lowest dissolved oxygen; low-flow backwater areas may support greater algal growth and therefore higher oxygen concentrations. These low-flow areas also may allow fish to avoid the constant high flow of the center channel of springs. The constant flow and water chemistry of a spring provides a consistent habitat with little seasonal variability. This consistent habitat also can serve as a refuge for species or populations that typically inhabit more variable rivers. The goal of this study was to survey the fish assemblage of Volusia Blue Spring, flspatially and over time to identify factors that influence fish density and distribution in this well-delineated and stable habitat. Few ecological studies of cool-water springs have been published (e.g., Herald and Strickland 1949, Hubbs 1995, Hubbs and Allen 1943, McKinsey and Chapman 1998, Munch et al. 2006, Odum and Caldwell 1955, Walsh et al. 2009), and most studies have not investigated both inter-annual and seasonal variation in a cool-water spring. For Volusia Blue Spring, we hypothesized that fish would be segregated longitudinally on the basis of dissolved oxygen concentration and horizontally into oxygen-rich microhabitats along the shoreline. Finally, although springs are generally stable with consistent flow throughout the year, we expected seasonal variation in fish assemblages due to the influx of large species from the St. Johns River that use the run as a warm-water refuge during the winter. 2010 K. Work, M. Gibbs, B. Peters, and L. French 651 Study Area Volusia Blue Spring in central Florida (28°56'51.0"N, 81°20'22.5"W) is a first magnitude spring that has discharged water historically at an average of 4.6 m3 s-1 or 162 cubic feet per second (cfs) (range = 1.8–6.1 m3 s-1) daily from the Floridan aquifer into the St. Johns River (Scott et al. 2004). The run is 320 m long and approximately 20–30 m wide. Quercus virginiana Miller (Live Oak) and Sabal palmetto Walter (Sabal Palm) line the banks, and little submerged vegetation occurs in the run. The run possesses gently sloping banks with scattered woody debris at low water. During high flows, water reaches steep muddy banks, allowing organisms access to holes in the mud. The run possesses no physical longitudinal delineation (i.e., no riffle/ pool structure) other than a deep hole at the spring head and a slight slope to the river. The run provides habitat for more than 40 species of fish (Florida Division of Recreation and Parks 2005), including several nonindigenous species, such as Oreochromis aureus (Blue Tilapia) and two South American armored catfish species, Pterygoplichthys disjunctivus (Vermiculated Sailfin Catfish) and Hoplosternum littorale (Brown Hoplo). During summer, public recreation heavily affects the upper portion of the run; large numbers of people swimming, tubing, and wading in the run completely denudes the center channel of algae. Megalops atlanticus (Tarpon) and the threatened Trichechus manatus latirostris (Harlan) (Florida Manatee) use the run as winter habitat (Florida Division of Recreation and Parks 2005). Methods Sampling We sampled five stations for fish, weekly or semi-monthly, from October 2000 to September 2004 (Fig. 1). The stations were relatively evenly distributed along Volusia Blue Spring run. Each sampling effort began at the headspring (station 1) and continued downstream to station 5. Over the four years of this study, we conducted 72 surveys of the run, although on several occasions we were unable to sample the stations near the St. Johns River. We sampled stations 1–4 during every month of the year, whereas station 5 was under-sampled in fall due to backflow of tannin-stained river water and danger of alligators. Inclement weather also precluded complete surveys on some dates. The area of each station was approximately 500 m2, with the exception of station 1, which encompassed the entire headspring and so was much larger (2500 m2). At each of the five stations, we used a 6-m x 2.4-m seine with 3-mm mesh to sample three subsampling sites (each approximately 6 m long x 3 m wide, and 30 cm–1 m deep) for small fish. We identified these fish in the field and released them. The subsampling sites were located along the bank on both sides of the spring run, and we sampled these sites throughout the course of the study. We used the following criteria for initial selection of subsampling 652 Southeastern Naturalist Vol. 9, No. 4 sites: presence of fish, nearby cover (bushes, submerged limbs, algal beds), and relative freedom from obstacles within the site. We recognize that avoiding structure for logistical reasons biased our samples against species that were associated with structure (Angermeier and Karr 1984). We sampled the same sites with the same methods throughout the four-year study, so we consider our composite data to be a representative assessment of fish assemblage variability in Volusia Blue Spring. All sites had sandy bottoms and varying amounts of algae, with the exception of two sub-sampling sites at Station 1. Station 1 was rocky with little available cover, as it experienced the most human disturbance from swimmers frequently sitting along the banks. Stations 2 and 3 were exposed to moderate human disturbance from swimmers in the center of the run; these two stations had available cover in the form of algae and woody debris, particularly along the banks. Stations 4 and 5 experienced little or no human disturbance, as the public was not allowed to access these areas, and these stations contained moderate cover. Station 5 had more leaf litter than the other stations and was seasonally inundated by the St. Johns River during late summer and fall. Finally, after seining was completed, we surveyed the area encompassing all three sub-sampling sites at each station via snorkeling to identify and enumerate larger (>8 cm) fish species that easily evaded the seine. This snorkel survey represented a fourth sample of each station. We identified all fish to species, except Lepisosteus osseus (Longnose Gar) and L. platyrhincus (Florida Gar). Highly spotted L. osseus were common in the run, making the discernment of retreating Lepisosteus species difficult. Similarly, very small Lepomis macrochirus (Bluegill) and L. gulosus Figure 1. Sampling stations within Volusia Blue Spring, fl. Large dots along the periphery of stations represent subsampling sites, and small dots in the mid channel represent water quality sampling sites. 2010 K. Work, M. Gibbs, B. Peters, and L. French 653 (Warmouth) were difficult to differentiate while moving, so we also combined data for these two species. We recognize the possibility that we counted fish twice, but we consider the seine sites within a station far enough apart to have limited this problem for the density estimates of small fish. While snorkeling, we attempted to keep track of fish movements to reduce the problem of double counting larger species during snorkel counts. We collected concurrent water quality data at each of the subsampling sites and at 14 mid-channel stations along the length of the run (Fig. 1). At each of the subsampling sites, we measured surface water velocity and dissolved oxygen concentration after each fish collection. We anticipated that the subsampling sites near the bank would provide refuges of lower flow and higher dissolved oxygen from the center channel of the run. We estimated surface water velocity with the rate of travel of a float. We measured dissolved oxygen concentration with a handheld YSI 85 multiparameter meter (Yellow Springs, OH). At each of the 14 mid-channel stations, we measured water velocity as above; we measured temperature, dissolved oxygen, and specific conductance (± 1% accuracy for all measurements) with the handheld YSI meter; and we collected pH data with an Oakton pH test handheld meter (Vernon Hills, IL). In Volusia Blue Spring, surface water velocity, discharge, and gage height vary somewhat independently of each other due to the influence of the St. Johns River’s large hydraulic head on the spring’s flow. When the St. Johns River gage height is high, river water can backflow into the spring, causing high gage height in the spring and sluggish flow, regardless of the discharge from the springhead. However, water quality can vary with discharge (K. Work, unpubl. data), so we considered discharge as a potentially important parameter in addition to flow rate and gage height. We acquired discharge and gage height data for the spring from the US Geological Survey online database of streamflow (station 02235500: http:// waterdata.usgs.gov/fl/nwis/rt). In our analyses, we only used data that had been collected within two days of our sample dates. Data analyses We evaluated annual, seasonal, and spatial variation in water velocity, dissolved oxygen concentration, total fish density, and densities of individual species with separate Kruskal-Wallis nonparametric multisample tests. The coefficient of variation (CV) was used to quantify the magnitude of variability in discharge, water velocity, specific conductance, and total fish density across all dates. Fish assemblage variability was evaluated with a principal components analysis (PCA). Principal components analysis creates new composite variables that are linear combinations of variables in the original dataset, thereby reducing the complexity of large data sets of many species or parameters. These new composite variables can be analyzed for large-scale shifts in assemblages, such as seasonal or annual changes. Natural log-transformed densities of all fish species in all samples (all dates 654 Southeastern Naturalist Vol. 9, No. 4 and stations) were used in the PCA, but we eliminated any species that occurred in fewer than 25% of the samples. We plotted the first two axes of the PCA; each data point represented the fish assemblage on a particular date at a particular station. We overlaid three sets of polygons on the same PCA data points in three separate graphs to represent samples grouped either by years, seasons, or stations to determine whether annual, seasonal, or spatial patterns were present in the fish assemblage data. For example, if polygons among years exhibited no overlap, then the fish assemblage differed in composition between years. If polygons were approximately the same size and had the same position in PCA space, then either the fish assemblage was similar between years or the variability was too great to discriminate between years. To examine whether variation in fish densities was related to physical or chemical characteristics of the run, we calculated Spearman rank correlations between water velocity, dissolved oxygen, and measures of fish species density. Where large numbers of correlations were calculated, the alpha level for significance was adjusted with a Bonferroni correction (Miller 1991). All statistical calculations were made using SPSS 13.0 (SPSS 2004). When correlations between water velocity and fish density appeared nonlinear, we determined whether each species typically minimized or maximized their exposure to high water velocities. For each station, we divided the water velocity of the subsampling site in which each species was observed (or the average of multiple subsampling sites if the species was present at more than one subsampling site) by the maximum water velocity available at a station. For example, for species i at station 1: % maximum water velocity available = water velocity in subsample used by species i / maximum water velocity in station 1. We repeated this process to assess fish associations with dissolved oxygen. Results Physical and chemical characteristics Discharge from the headspring was high (2.7–5.2 m3 s-1 or 63–214 cfs) and fairly consistent (CV = 12%), even though the study period included recovery from a drought. High discharge, however, did not correspond with high gage heights or with high water velocities, both of which correlate more with changes in water volume of the St. Johns River than with the discharge from the spring. The influence of the large and variable hydraulic head of the St. Johns River causes an increase in gage heights regardless of discharge after long periods of high rainfall. Seasonal changes in the hydraulic head of the St. Johns River produced significant variation in water velocity (χ2 = 7.9, df = 3, P = 0.05), with highest average velocities for the run in spring (0.31 m s-1 ± 0.10 SD) and lowest in fall (0.19 m s-1 ± 0.10 SD). Water velocity was predictable at large spatial scales, with the high2010 K. Work, M. Gibbs, B. Peters, and L. French 655 est velocities at station 2, due to the narrowing of the channel beyond the headspring, and lowest at station 5, due to the wider channel at that station (χ2 = 37.0, df = 4, P < 0.001; Fig. 2A). Velocity was consistently higher in the center of the channel than on the banks (center channel average velocity was ≥0.2–0.5 m s-1, while the velocity was often 0 m s-1 on the bank; χ2 = 58.8, df = 1, P < 0.0001). Despite these clear spatial patterns, water velocity was highly variable among stations (CV = 146%) and at a station over time (CV = 93% for one subsampling site at station 1 across all dates). Dissolved oxygen concentration varied seasonally (χ2 = 11.5, df = 3, P = 0.009), with highest averages of samples from the entire run in summer (1.6 mg L-1 ± 1.0 SD) and lowest in fall (1.1 mg L-1 ± 0.56 SD). Dissolved oxygen concentration also varied spatially (χ2 = 178.1, df = 4, P < 0.001), with lowest concentrations at the springhead and highest concentrations at the two stations near the confluence with the St. Johns River (Fig. 2B). Near the banks, dissolved oxygen concentration occasionally reached 4–6 mg L-1, but it was typically much lower (bank average = 1.26 mg L-1 ± 0.74 SD). Within a given site, the dissolved oxygen concentration measurements near the bank always exceeded measurements in the center of the channel (center channel average = 0.76 mg L-1 ± 0.80 SD; χ2 = 239.6, df = 1, P < 0.0001). Variability in dissolved oxygen among all samples was high (CV = 128%), although variability among dates for one station was much lower (CV = 44% for one subsampling site at station 1). Specific conductance was high compared to many freshwater streams (all sample average = 1450 μS cm-1 ± 370 SD). It was variable seasonally (χ2 = 45.8, df = 3, P < 0.0001), with highest values in summer (average of all center channel stations = 1650 μS cm-1 ± 400 SD) and lowest in winter (average of all center channel stations = 1230 μS cm-1 ± 180 SD), but not variable spatially (χ2 = 0.15, df = 4, P = 0.99). Variability was relatively low across all stations (CV = 26%) and within one station (CV = 26% for station 1). Variability in temperature and pH were negligible (all sample averages = 23.2 °C ± 0.71 SD and 7.67 ± 0.26 SD, respectively). Fish assemblage patterns Thirty-four fish species were observed during the course of this study (Table 1). Total fish density (all individuals of all species) declined down the length of the spring run (χ2 = 29.3, df = 4, P < 0.001; Fig. 3), primarily due to high densities of Gambusia holbrooki (Eastern Mosquitofish) at the headspring. Across all dates, 82.8% ± 18.6 SD of all fish collected at the headspring were G. holbrooki. Total fish density also varied seasonally (χ2 = 13.1, df = 3, P = 0.004; Fig. 3), again primarily due to high G. holbrooki densities in winter. Gambusia holbrooki was the most abundant fish species in the run on most sampling dates (58.6 ± 21.4% SD of total fish density across all dates and stations). The total density of fish was quite variable (CV = 133%). 656 Southeastern Naturalist Vol. 9, No. 4 Figure 2. Spatial variation in Volusia Blue Spring water velocity (A), dissolved oxygen concentration (B), and specific conductance (C) over the period of 2000–2004. Bars represent averages for each station across all dates with standard deviations. 2010 K. Work, M. Gibbs, B. Peters, and L. French 657 Most species exhibited high inter-annual variation in occurrence, and 31.4% of species were not observed every year (Table 2). Densities of all species, with the exception of Lepisosteus spp., Notemigonus crysoleucas (Golden Shiner), Notropis chalybaeus (Ironcolor Shiner), Piaractus brachypomus (Pirapatinga), and Ictalurus punctatus (Channel Catfish), varied significantly among years (P < 0.05). Much of this variation was sporadic; however, most centrarchids increased during the study period, and most fundulids declined (Fig. 4). The nonindigenous loricariid, Pterygoplichthys disjunctivus, also increased during the study period; at times, we observed hundreds of fish in groups scattered along the spring run. We observed large schools of the nonindigenous O. aureus in 2000, which had virtually disappeared by 2001, and a small school of the nonindigenous Table 1. Fish species observed in Blue Spring during 2000–2006 (* indicates nonindigenous species). Family Scientific Name Common Name Authority Lepisosteidae Lepisosteus osseus Longnose Gar (Linnaeus, 1758) Lepisosteus platyrhincus Florida Gar DeKay, 1842 Amiidae Amia calva Bowfin Linnaeus, 1766 Megalopidae Megalops atlanticus Tarpon Valenciennes, 1847 Cyprinidae Ctenopharyngodon idella Grass Carp* (Valenciennes, 1844) Notemigonus crysoleucas Golden Shiner (Mitchill, 1814) Notropis chalybaeus Ironcolor Shiner (Cope, 1867) Notropis petersoni Coastal Shiner Fowler, 1942 Catostomidae Erimyzon sucetta Lake Chubsucker (Lacepède, 1803) Characidae Piaractus brachypomus Pirapatinga* (Cuvier, 1818) Ictaluridae Ictalurus punctatus Channel Catfish (Rafinesque, 1818) Callichthyidae Hoplosternum littorale Brown Hoplo* (Hancock, 1828) Loricariidae Pterygoplichthys Vermiculated Sailfin (Weber, 1991) disjunctivus Catfish* Mugilidae Mugil cephalus Striped Mullet Linnaeus, 1758 Atherinopsidae Menidia beryllina Inland Silverside (Cope, 1867) Fundulidae Fundulus chrysotus Golden Topminnow (Günther, 1866) Fundulus seminolis Seminole Killifish Girard, 1859 Lucania goodei Bluefin Killifish Jordan, 1880 Lucania parva Rainwater Killifish (Baird & Girard, 1855) Poeciliidae Gambusia holbrooki Eastern Mosquitofish Girard, 1859 Heterandria formosa Least Killifish Girard, 1859 Poecilia latipinna Sailfin Molly (Lesueur, 1821) Cyprinodontidae Jordanella floridae Flagfish Goode & Bean, 1879 Centrarchidae Enneacanthus gloriosus Bluespotted Sunfish (Holbrook, 1855) Lepomis auritus Redbreast Sunfish (Linnaeus, 1758) Lepomis gulosus Warmouth (Cuvier, 1829) Lepomis macrochirus Bluegill Rafinesque, 1819 Lepomis microlophus Redear Sunfish (Günther, 1859) Lepomis punctatus Spotted Sunfish (Valenciennes, 1831) Micropterus salmoides Largemouth Bass (Lacepède, 1802) Pomoxis nigromaculatus Black Crappie (Lesueur, 1829) Percidae Percina nigrofasciata Blackbanded Darter (Agassiz, 1854) Cichlidae Oreochromis aureus Blue Tilapia* (Steindachner, 1864) Achiridae Trinectes maculatus Hogchoker (Bloch & Schneider, 1801) 658 Southeastern Naturalist Vol. 9, No. 4 Table 2. Occurrences of rare species in Blue Spring, fl. Many of these species were not collected during routine sampling, but were observed at other times. Species 2000 2001 2002 2003 2004 2005 2006 Amia calva X Ctenopharyngodon idella X Notropis petersoni X Hoplosternum littorale X X Menidia beryllina X X X Jordanella floridae X Enneacanthus gloriosus X Percina nigrofasciata X Oreochromis aureus X X X X X X Trinectes maculatus X Figure 3. Spatial and seasonal variation in density of the fish fauna of Volusia Blue Spring over the period of 2000–2004. Bars represent averages for each station across all dates with standard deviations. P. brachypomus sporadically. However, due to high variation in counts within a season, fewer than half of the species (8 out of 21) exhibited statistically significant seasonal variation, and in most cases, inter-annual 2010 K. Work, M. Gibbs, B. Peters, and L. French 659 Figure 4. Temporal variation in large predators (A), centrarchids (B), and small fundulids and poeciliids (C) over the period of 2000– 2004. Values represent averages for each date across all stations, and error bars are standard deviations. 660 Southeastern Naturalist Vol. 9, No. 4 variation was as great as or greater than seasonal variation (Fig. 4). Only G. holbrooki, which produced large numbers of juveniles in winter, and M. atlanticus, which used Volusia Blue Spring as a warm-water refuge from cool river or ocean temperatures, exhibited a strong and predictable seasonal pattern of high densities in winter. All fish species varied spatially in density (number of individuals per square meter, P < 0.001; Fig. 5). With the exception of poeciliids, the majority of small species (fundulids, cyprinids, small centrarchids) were most abundant midway down the run (stations 2 and 3), although Lucania parva (Rainwater Killifish) densities also were high at the lower end of the run (stations 4–5). Lepomis microlophus (Redear Sunfish) were most abundant at station 4. However, most individuals of L. microlophus that occurred at sites 4 and 5 were large (estimated total length > 20 cm versus ≈5–15 cm upstream). Larger species, such as Micropterus salmoides (Largemouth Bass), Lepisosteus spp., and M. atlanticus, generally were most abundant at the lower end of the run (stations 4 and 5), although M. salmoides also occurred at stations 2 and 3. As a result of the large variation and differing patterns of density among species, the first three axes of the PCA only explained 39.7% of the variation in fish density. The first two axes explained the majority of this variation (16% and 14%, respectively). Fundulids and poeciliids correlated most with the first axis, and centrarchids correlated most with the second axis (Table 3). Clusters of samples from different years and seasons overlapped greatly (Fig. 6A, B). The polygons circumscribing PCA points for stations also overlapped, although stations 1 and 5 were nested within a smaller portion of the PC space than any of the other stations (Fig. 6C). Despite the prevalence of large predatory species at station 5 that were virtually absent at station 1, these two stations overlapped in the PC space because fundulids and centrarchids were relatively scarce at both stations. Table 3. Loadings of the fish variables on the first two principle components. Species Axis 1 Axis 2 Lepisosteus osseus/platyrhincus -0.165 -0.115 Notemigonus crysoleucas -0.016 0.161 Pterygoplichthys disjunctivus -0.231 0.143 Mugil cephalus 0.208 0.355 Fundulus chrysotus 0.461 0.142 Fundulus seminolis 0.281 0.335 Lucania goodei 0.740 0.160 Lucania parva 0.525 0.079 Gambusia holbrooki 0.652 -0.284 Heterandria formosa 0.751 -0.075 Poecilia latipinna 0.697 -0.191 Lepomis auritus -0.067 0.482 Lepomis macrochirus/gulosus 0.065 0.778 Lepomis microlophus -0.095 0.351 Lepomis punctatus 0.053 0.648 Micropterus salmoides -0.092 0.357 2010 K. Work, M. Gibbs, B. Peters, and L. French 661 Figure 5. Spatial variation in large predators (A), centrarchids (B), and small fundulids and poeciliids (C) over the period of 2000–2004. Values represent averages for each date across all stations, and error bars are standard deviations. 662 Southeastern Naturalist Vol. 9, No. 4 Figure 6. Plots of the first two axes of principle components analyses of the Volusia Blue Spring fish fauna showing annual variation (A), seasonal variation (B), and spatial variation (C). The first two axes accounted for only 30% of the variation in the fish fauna. 2010 K. Work, M. Gibbs, B. Peters, and L. French 663 Relationships with water velocity and dissolved oxygen Most species were affected by water volume in the run (gage height), discharge, or water velocity; only seven species were unaffected by one of these measures of water volume or flow rate (Table 4). However, none of these relationships were linear; species’ densities typically were high at low velocities and low to zero at moderate and high velocities (Fig. 7). Even large species that were commonly found in the center channel exhibited this type of distribution. Fish typically selected much lower water velocities than the maximum available (average = 29% of maximum available; Table 5). Only the largest species (Lepisosteus spp. and M. atlanticus) and benthic species (I. punctatus and P. disjunctivus) selected greater water velocities (average = 57% of maximum available). Although we observed a shift in the assemblage from low-oxygen tolerant poeciliids at the headspring to a much more diverse assemblage at the lower end of the run, correlations with dissolved oxygen were weak (Table 4). Plots of the distributions of most species relative to oxygen concentration yielded unimodal or bimodal curves rather than positive linear relationships (Fig. 8). However, all species controlled their oxygen exposure by selecting areas with 70–80% of the maximum oxygen concentration available at a station (Table 5). Discussion The stable and dramatic gradients in dissolved oxygen and water velocity for Volusia Blue Spring supported the fish distribution patterns predicted by previous studies of Florida springs (McKinsey and Chapman 1998, Odum and Caldwell 1955). The ability of poeciliids to use the airwater interface for respiration in hypoxic water (McKinsey and Chapman 1998, McLane 1955, Odum and Caldwell 1955) allowed G. holbrooki and Poecilia latipinna (Sailfin Molly) to occur at the headspring. Most other small species, such as the fundulids, require higher dissolved oxygen concentrations and are usually associated with algal beds and woody structure (McLane 1955). These species’ densities typically were highest downstream of the spring head, but upstream of the highest densities of the larger potential predators, such as M. salmoides and Lepisosteus spp. Lucania parva provided an exception to this pattern; the distribution of this species greatly overlapped with potential predators, perhaps due to its ability to change color for camouflage (Cox et al. 2009). The smaller species (poeciliids, fundulids, cyprinids, small centrarchids) typically occurred along the banks, areas that possessed significantly higher oxygen concentrations and lower water velocities than the center of the channel. Finally, low dissolved oxygen concentrations likely contributed to the virtual exclusion of larger species from the spring head. Although Lepisosteus spp., M. atlanticus, and P. disjunctivus gulp air at the surface (Graham 1997; Wootton 1990; K. Work, pers. observ.), areas with higher oxygen concentrations 664 Southeastern Naturalist Vol. 9, No. 4 Table 4. Spearman rank correlations (rs) of fish species with dissolved oxygen and flow in Blue Spring, Volusia County, fl. Fish density data were collected by seine (small fishes) or by snorkel survey (large fishes) from October 2000 to September 2004. Significant (P < 0.05) correlations shown in bold. Species Discharge Gage height Water velocity Dissolved oxygen Lepisosteus osseus/platyrhincus rs = 0.17, P = 0.23 rs = -0.07, P = 0.65 rs = 0.03, P = 0.68 rs = 0.21, P = 0.0005 Megalops atlanticus rs = 0.20, P = 0.15 rs = 0.07, P = 0.64 rs = 0.21, P = 0.002 rs = 0.05, P = 0.39 Notemigonus crysoleucas rs = 0.35, P = 0.01 rs = 0.24, P = 0.09 rs = -0.07, P = 0.30 rs = -0.02, P = 0.69 Notropis chalybaeus rs = -0.05, P = 0.74 rs = 0.15, P = 0.30 rs = 0.04, P = 0.58 rs = -0.03, P = 0.63 Piaractus brachypomus rs = -0.08, P = 0.68 rs = 0.49, P < 0.0001 rs = 0.09, P = 0.18 rs = 0.07, P = 0.27 Ictalurus punctatus rs = 0.50, P < 0.0001 rs = 0.22, P = 0.13 rs = 0.01, P = 0.86 rs = -0.002, P = 0.98 Pterygoplichthys disjunctivus rs = 0.57, P < 0.0001 rs = 0.56, P < 0.0001 rs = 0.16, P = 0.019 rs = -0.15, P = 0.014 Mugil cephalus rs = -0.18, P = 0.21 rs = -0.40, P < 0.0001 rs = -0.02, P = 0.79 rs = 0.43, P < 0.001 Menidia beryllina rs = -0.04, P = 0.75 rs = 0.49, P < 0.0001 rs = 0.01, P = 0.88 rs = 0.02, P = 0.79 Fundulus chrysotus rs = -0.21, P = 0.13 rs = -0.36, P = 0.01 rs = 0.12, P = 0.046 rs = 0.12, P = 0.046 Fundulus seminolis rs = -0.06, P = 0.69 rs = 0.02, P = 0.90 rs = -0.10, P = 0.14 rs = 0.07, P = 0.26 Lucania goodei rs = -0.17, P = 0.23 rs = -0.59, P < 0.0001 rs = -0.07, P = 0.28 rs = 0.57, P < 0.001 Lucania parva rs = -0.16, P = 0.26 rs = -0.65, P < 0.0001 rs = -0.12, P = 0.063 rs = 0.41, P < 0.001 Gambusia holbrooki rs = 0.16, P = 0.27 rs = -0.24, P = 0.10 rs = -0.09, P = 0.16 rs = -0.17, P = 0.005 Heterandria formosa rs = -0.27, P = 0.05 rs = -0.60, P < 0.0001 rs = -0.18, P = 0.008 rs = 0.31, P < 0.001 Poecilia latipinna rs = 0.14, P = 0.33 rs = -0.08, P = 0.57 rs = -0.11, P = 0.11 rs = 0.008, P = 0.90 Lepomis auritus rs = 0.30, P = 0.03 rs = -0.23, P = 0.11 rs = 0.02, P = 0.74 rs = 0.37, P < 0.001 Lepomis macrochirus/gulosus rs = 0.52, P < 0.0001 rs = -0.32, P = 0.02 rs = 0.18, P = 0.006 rs = 0.29, P < 0.001 Lepomis microlophus rs = 0.004, P = 0.98 rs = -0.24, P = 0.09 rs = -0.07, P = 0.32 rs = 0.24, P < 0.001 Lepomis punctatus rs = 0.46, P < 0.0001 rs = -0.12, P = 0.39 rs = 0.03, P = 0.69 rs = 0.24, P < 0.001 Micropterus salmoides rs = 0.18, P = 0.21 rs = -0.18, P = 0.21 rs = -0.004, P = 0.95 rs = 0.41, P < 0.001 Pomoxis nigromaculatus rs = 0.03, P = 0.82 rs = 0.02, P = 0.91 rs = -0.03, P = 0.68 rs = 0.08, P = 0.17 Oreochromis aureus rs = -0.35, P = 0.01 rs = 0.30, P = 0.04 rs = -0.01, P = 0.87 rs = -0.001, P = 0.98 2010 K. Work, M. Gibbs, B. Peters, and L. French 665 Figure 7. The relationship between fish density and water velocity in Volusia Blue Spring from 2000– 2004. 666 Southeastern Naturalist Vol. 9, No. 4 are less stressful to large fish (Wootton 1990). Even at the lower end of the run, all three species were commonly observed gulping at the surface. Interestingly, during periods of extremely high gage height during floods, we observed a shift in these larger species toward the spring head, possibly in response to the backflow of tannic-stained water from the river. Although we could not seine at high gage heights, during these periods we also observed a paucity of small fish, particularly Lucania spp. and Heterandria formosa (Least Killifish), perhaps due to the proximity of large predators. Despite the support for the predictable trends of species’ distributions observed in other studies (McKinsey and Chapman 1998, Odum and Caldwell 1955), the more extensive dataset of our study highlighted the high degree of variation that underlies these predictable patterns, as indicated by high coefficients of variation for fish densities relative to stream fish populations and the low explanatory power of the PCA. Clearly, the fish fauna did not fall into neat groupings based exclusively on water velocity or oxygen tolerance, despite the high flow rates, extremely low oxygen concentrations, and the strong and consistent oxygen gradient. Most of the smaller (poeciliids, fundulids, small centrarchids) or benthic (P. disjunctivus) species typically selected lower water velocities than the maximum available at a station, but their nonlinear distributions relative Table 5. Water velocity and dissolved oxygen preferences as a percentage of the maximum available at a station (with standard deviation) in Volusia Blue Spring, flduring 2000–2004. % max = percent maximum available (± SD). Water velocity (m s-1) Dissolved oxygen (mg L-1) Species Range % max Range % max Lepisosteus osseus/platyrhincus 0-1.00 48.5 ± 39.7 0.12–4.86 74.7 ± 20.7 Megalops atlanticus 0–0.42 61.1 ± 39.8 0.51–2.19 71.4 ± 17.6 Notemigonus crysoleucas 0–0.56 24.4 ± 27.1 0.17–3.08 76.4 ± 21.3 Notropis chalybaeus 0–0.20 20.2 ± 30.4 0.18–2.48 72.5 ± 28.2 Ictalurus punctatus 0–0.36 79.1 ± 39.6 0.22–2.58 64.6 ± 23.3 Pterygoplichthys disjunctivus 0–0.57 41.0 ± 38.4 0.09–4.86 76.9 ± 20.3 Mugil cephalus 0–1.00 35.0 ± 34.8 0.18–6.04 71.5 ± 23.4 Menidia beryllina 0–0.17 23.2 ± 26.9 0.39–1.80 86.4 ± 13.9 Fundulus chrysotus 0–0.50 33.4 ± 29.4 0.19–6.04 72.6 ± 19.6 Fundulus seminolis 0–0.56 27.3 ± 28.5 0.18–5.23 75.4 ± 21.6 Lucania goodei 0–0.59 37.0 ± 26.9 0.09–6.04 69.6 ± 19.3 Lucania parva 0–0.72 38.3 ± 27.6 0.20–6.04 70.3 ± 16.5 Gambusia holbrooki 0–0.72 35.0 ± 19.4 0.06–6.04 70.1 ± 15.7 Heterandria formosa 0–0.59 32.3 ± 24.7 0.09–6.04 73.4 ± 21.2 Poecilia latipinna 0–0.56 33.4 ± 27.6 0.07–6.04 71.5 ± 19.0 Lepomis auritus 0–0.57 25.2 ± 26.7 0.22–4.57 79.6 ± 19.4 Lepomis macrochirus/gulosus 0–0.95 27.1 ± 23.6 0.14–6.04 76.1 ± 18.5 Lepomis microlophus 0–0.57 25.5 ± 27.6 0.16–5.23 79.4 ± 15.0 Lepomis punctatus 0–0.57 28.1 ± 28.5 0.22–4.97 77.6 ± 18.1 Micropterus salmoides 0–0.57 28.6 ± 25.8 0.16–4.97 74.6 ± 19.8 Pomoxis nigromaculatus 0–0.19 30.4 ± 40.4 0.55–4.51 81.6 ± 17.9 Oreochromis aureus 0–0.15 13.7 ± 14.0 1.31–1.58 98.4 ± 3.3 2010 K. Work, M. Gibbs, B. Peters, and L. French 667 Figure 8. The relationship between fish density and dissolved oxygen concentration in Volusia Blue Spring from 2000–2004. 668 Southeastern Naturalist Vol. 9, No. 4 to water velocity indicated that individuals selected higher velocities under some conditions. Most species selected 70–80% of the maximum oxygen concentration available at a station. For many species, the unimodal relationships of fish density against dissolved oxygen concentration were skewed toward lower oxygen concentrations, further suggesting that fish were not selecting the maximum available oxygen concentrations as might be expected in this system with extremely low dissolved oxygen concentrations. These relationships suggest that although relatively constant flow and oxygen gradients affect fish distributions in Volusia Blue Spring, other factors, some of which may be biotic, modify the responses of fish to these parameters. Why is the Volusia Blue Spring fish fauna variable? Several factors likely contribute to the variability of the fish fauna. First, although Volusia Blue Spring differs from the St. Johns River in hydrology and chemistry, the run is short and wide with a broad connection to this large river. The St. Johns River contains a high diversity of organisms, many of which use the run as a periodic resource rather than as a permanent habitat, such as Trichechus manatus latirostris, M. atlanticus, Mugil cephalus (Striped Mullet), and Callinectes sapidus (Blue Crab), which are each present in the river on a seasonal basis (McLane 1955). Several studies of stream fish assemblages have indicated that the location of a site and the connectivity among sites within a landscape have significant effects on site-level fish diversity (Erős and Grossman 2005, Magalhães et al. 2002, Taylor 1997, Taylor and Warren 2001). Therefore, it is likely that “stream level variability” (Dunham and Vinyard 1997), or the variability of the St. Johns River, affected the variability of the fish fauna in Volusia Blue Spring. The magnitude of the influx of species, either as numbers of species or as numbers of individuals, has not been quantified, but assemblage composition is likely to be influenced at the very least by species that use the run as a thermal refuge in winter. Large piscivorous species, like M. atlanticus, are predaceous on smaller species, whereas others may alter the availability and distribution of resources. For example, manatee movements and the feeding activity of P. disjunctivus can strip attached algae off the bottom of the run, thereby removing habitat for smaller fish species (K. Work, unpubl. data). We did not quantify these activities, so we can only speculate as to their effects on fish variability. However, the influx of more than thirty M. atlanticus (>2 m long), for example, to a 620-m run likely would affect the densities and distributions of other fish species in the run. Furthermore, less visible species that move into and out of the run from the St. Johns River may have more subtle effects on the fish fauna. Large species like M. atlanticus may have home ranges larger than the spring run, but smaller species with smaller home ranges still may make forays out of the run (Hill and 2010 K. Work, M. Gibbs, B. Peters, and L. French 669 Grossman 1987). As a result, some of the variability that we observed likely was not true population variation, but rather a snapshot of a subset of the population in the run at any given time. A second potential factor in the high variability of the Volusia Blue Spring fish fauna is that, although the run provides a harsh oxygen environment, species can move easily throughout the run due to its morphology. The run is comparatively wide and deep with no riffles, so even large predators occasionally reach the spring head. Small headwater streams often have riffles that preclude movement of larger predators between pools during low-water periods. Pools with few or no large predators may serve as refugia for smaller prey species. Matthews et al. (1994) observed significant differences among pools in a small, variable Oklahoma stream, and some of the differences were attributed to the presence or absence of M. salmoides and large L. macrochirus. Although we did not observe M. atlanticus at the spring head during this study, individuals of 60% of all species were observed at the spring head on at least one occasion during the four years of our study, including predators such as M. salmoides and Lepisosteus spp. Occasional forays of predators into lowoxygen, but prey-rich habitat near the spring head may increase variability in the distribution and densities of smaller species. Finally, the fish fauna of Volusia Blue Spring may be inherently stochastic. Environmental instability can cause variability in fish populations (e.g., Humpl and Pivnička 2006, Marchetti and Moyle 2001, Matthews et al. 1994, Oberdorff et al. 2001, Pearsons et al. 1992, Pusey et al. 2000), and such environmental instability varies in magnitude. Highly controlled experimental streams should be comparatively stable and, therefore, a good indicator of the potential for inherent variability in a stream fish assemblage. Matthews and Marsh-Matthews (2006) followed seven assemblages of stream fishes in experimental streams over the course of a year. These assemblages were stocked with the same species at the same densities and life stages and yet, at the end of the year, the assemblages differed substantially. Some of this variability may have been due to differences in resources between pools that developed over the course of the year, but much of the variation appeared to be unpredictable change. It is likely that the variability of the Volusia Blue Spring fish fauna is a combination of this type of inherent variability overlain with St. Johns River-induced variability. This study indicated that the Volusia Blue Spring fish fauna was more variable than expected from the results of previous studies on Florida springs (McKinsey and Chapman 1998, Odum and Caldwell 1955), particularly given the consistency of the physical and chemical parameters of the spring. This consistency provides a predictable physical and chemical habitat, but the close connection with the St. Johns River undoubtedly magnifies inherent fish assemblage variation. Volusia Blue Spring is likely more variable than most springs in Florida due to this river connection. However, all springs 670 Southeastern Naturalist Vol. 9, No. 4 may be more variable than indicated by studies of short duration or low temporal resolution (i.e., infrequently sampled). Most Florida springs are threatened by reductions in flow, increases in nutrient inputs, and invasions of nonindigenous species (Scott et al. 2004). Therefore, a better understanding of variability in the ecology of these springs is crucial for developing effective strategies to protect them. Acknowledgments We thank personnel of Blue Spring State Park in Volusia County for granting access to the spring and for logistical aid. We also acknowledge the people who made substantial contributions as field assistants: Corey Green, Sabrina Krisberg, Kara Moore, Aaron Odegard, Kevin Palmer, Tessa Payne, Alicia Schultheis, Kira Smedley, and Trevor Tyner. Finally, we thank Terry Farrell for consultation and Cindy Bennington, Stephen Walsh, and two anonymous reviewers for editing the manuscript. Literature Cited Angermeier, P.L., and J.R. Karr. 1984. Relationships between woody debris and fish habitat in a small warmwater stream. Transactions of the American Fisheries Society 113:716–726 Bennett, G.W. 1971. Management of Lakes and Ponds, 2nd Edition. Van Nostrand Reinhold Co., New York, NY. 375 pp. Cox, S., S. Chandler, C. Barron, and K. Work. 2009. Benthic fish exhibit more plastic crypsis than non-benthic species in a freshwater spring. Journal of Ethology 27(3):497–505. Dibble, E.D., and S.L. Harrel. 2000. Microhabitat selection by three stream-dwelling cyprinids: Blacktail Shiner (Cyprinella venusta), Bluntface Shiner (C. camura), and Striped Shiner (Luxilus chrysocephalus). Journal of Freshwater Ecology 15(4):455–463. Dunham, J.B., and G.L. Vinyard. 1997. Incorporating stream level variability into analyses of site-level fish-habitat relationships: Some cautionary examples. Transactions of the American Fisheries Society 126:323–329. Erős, T., and G.D. Grossman. 2005. Fish biodiversity in two Hungarian streams: A landscape-based approach. Archiv für Hydrobiologie 162:53–71. Florida Division of Recreation and Parks. 2005. Blue Spring and Hontoon Island State Parks Unit management plan, Florida Department of Environmental Protection,Tallahassee, fl. Gorman, O.T., and J.T. Karr. 1978. Habitat structure and stream fish communities. Ecology 59:507–515. Graham, J.B. 1997. Air-breathing Fishes: Evolution, Diversity, and Adaptation. Academic Press, San Diego, CA. 299 pp. Herald, E.S., and R.R. Strickland. 1949. An annotated list of the fishes of Homosassa Springs, Florida. Journal of the Florida Academy of Sciences 11(4):99–109. Hill, J., and G.D. Grossman. 1987. Home-range estimates for three North American stream fishes. Copeia 1987(2):376–380. Hubbs, C. 1995. Springs and spring runs as unique aquatic ecosystems. Copeia 1995(4):989–991. 2010 K. Work, M. Gibbs, B. Peters, and L. French 671 Hubbs, C.L., and E.R. Allen. 1943. Fishes of Silver Springs, Florida. Proceedings of the Florida Academy of Sciences 6(3/4):110–116. Humpl, M.. and K. Pivnička. 2006. Fish assemblages as influenced by environmental factors in streams in protected areas of the Czech Republic. Ecology of Freshwater Fish 15:96–103. Kessler, R.K., A.F. Casper, and G.K. Weddle. 1995. Temporal variation in microhabitat use and spatial relations in the benthic fish community of a stream. American Midland Naturalist 134:361–370. Magalhães, M.F., D.C. Batalha, and M.J. Collares-Pereira. 2002. Gradients in stream fish assemblages across a Mediterranean landscape: Contributions of environmental factors and spatial structure. Freshwater Biology 47:1015–1031. Marchetti, M.P., and P.B. Moyle. 2001. Effects of flow regime on fish assemblages in a regulated California stream. Ecological Applications 11:530–539. Matthews, W.J. 1998. Patterns in Freshwater Fish Ecology. Chapman and Hall, New York, NY. 756 pp. Matthews, W.J., and E. Marsh-Matthews. 2006. Temporal changes in replicated experimental stream fish assemblages: Predictable or not? Freshwater Biology 51:1605–1622. Matthews, W.J., B.C. Harvey, and M.E. Power. 1994. Spatial and temporal patterns in the fish assemblages of individual pools in a midwestern stream (USA). Environmental Biology of Fish 39:381–397. McKinsey, D.M., and L.J. Chapman. 1998. Dissolved oxygen and fish distribution in a Florida spring. Environmental Biology of Fish 53:211–223. McLane, W.M. 1955. The fishes of the St. Johns River system. PhD. Dissertation. University of Florida, Gainesville, fl. 361 pp. Miller, R.G., Jr. 1991. Simultaneous Statistical Inference. Springer Verlag, New York, NY. 299 pp. Munch, D.A., D.J. Toth, C. Huang, J.B. Davis, C.M. Fortich, W.L. Osburn, E.J. Phlips, E.L. Quinlan, M.S. Allen, M.J. Woods, P. Cooney, R.L. Knight, R.A. Clarke, and S.L. Knight. 2006, Fifty-year retrospective study of the ecology of Silver Springs, Florida. St. Johns River Water Management District Special Publication SJ2007-SP4. Palatta, fl. Oberdorff, T., B. Hugueny, and T. Vigneron. 2001. Is assemblage variability related to environmental variability? An answer for riverine fish. Oikos 93:419–428. Odum, H.T., and D.K. Caldwell. 1955. Fish respiration in the natural oxygen gradient of an anaerobic spring in Florida. Copeia 1955(2):104–106. Pearsons, T.N., H.W. Li, and G.A. Lamberti. 1992. Influence of habitat complexity on resistance to flooding and resilience of stream fish assemblages. Transactions of the American Fisheries Society 121:427–436. Pusey, B.J., M.J. Kennard, and H. Arthington. 2000. Discharge variability and the development of predictive models relating to stream fish assemblage structure to habitat in northeastern Australia. Ecology of Freshwater Fish 9:30–50. Scott, T.M., G.H. Means, R.P. Meegan, R.C. Means S.B. Upchurch, R.E. Copeland, J. Jones, T. Roberts, and A. Willet. 2004. Springs of Florida. Florida Geological Survey, Bulletin No. 66. Tallahassee, fl. SPSS. 2004. SPSS, Inc. Chicago, IL. Taylor, C.M. 1997. Fish species richness and incidence patterns in isolated and connected stream pools: Effect of pool volume and spatial position. Oecologia 110:560–566. 672 Southeastern Naturalist Vol. 9, No. 4 Taylor, C.M., and M.L. Warren. 2001. Dynamics in species composition of stream fish assemblages: Environmental variability and nested subsets. Ecology 82:2320–2330. Walsh, S.J., L. Knowles, Jr., B.G. Katz, and D.G. Strom. 2009. Hydrology, water quality, and aquatic communities of selected springs in the St. Johns River Water Management District. USGS Scientific Investigations Report 2009–5046. Reston, VA. Whitford, L.A. 1956. The communities of algae in the springs and streams of Florida. Ecology 37:433–442. Wootton, R.J. 1990. Ecology of Teleost Fishes. Chapman and Hall, London, UK. 404 pp.