Determinism in Fish Assemblages of Floodplain Lakes of the Vastly Disturbed Mississippi Alluvial Valley

Introduction

The Mississippi Alluvial Valley between southern Illinois and southern Louisiana is a 40–200-km-wide area with hundreds of natural lakes (Baker et al. 1991). The lakes are sited along the Mississippi River and along tributaries of numerous major rivers, such as the White, Arkansas, Tensas, and Yazoo, that meander into the Mississippi Alluvial Valley. Connectivity between the rivers and valley has been altered by levees erected to reduce flooding in land that has been cleared to accommodate agriculture (Baker et al. 1991). Whereas lakes inside levees have often lost their connectivity to an expansive, forested floodplain, they retain various levels of connectivity to the river. Conversely, lakes outside levees have lost their connectivity to the river, yet retain their connectivity to an inactive floodplain that has been deforested to support intensive agriculture. Lakes outside levees are adversely affected by changes in hydrologic cycles attributable to landscape modifications (particularly deforestation and drainage alterations) and by agricultural runoff that commonly contains heavy sediment loads and pesticides (McHenry et al. 1982; Cooper et al. 1987). These disturbances affect all rivers flowing through the valley, although at different scales. Turbidity, sedimentation, and pesticides have been listed as key factors contributing to degradation of floodplain lakes in the region (Lucas 1985; Cooper and McHenry 1989).

Floodplain lakes are created by a variety of fluvial processes resulting in an assortment of shapes and sizes that influence lake abiotic characteristics and possibly play a role in patterning their fish assemblages. In the Mississippi Alluvial Valley, we have observed that the most distinctive abiotic characteristics of floodplain lakes are size, shape, transparency, and extent of connectivity to the river. These abiotic conditions have been shown to dictate the composition of fish assemblages in lakes of undisturbed floodplain ecosystems, either directly via limits on physiological tolerance or indirectly via constraints on biotic interactions. In the Orinoco River, Venezuela, fish assemblage structure in floodplain lakes was attributed to piscivory under the influence of transparency, as controlled by depth and area (Rodriguez and Lewis 1997; Lewis et al. 2000). In the Brazos River, Texas, assemblage structure showed large between-lake variation that was explained by depth, dissolved oxygen, dissolved nutrients, turbidity, and plankton densities (Winemiller et al. 2000). However, other authors have failed to find clear links between fish assemblage and abiotic characteristics of floodplain lakes, and have concluded that fish assemblages are simply random associations of species (e.g., Goulding et al. 1988; Saint-Paul et al. 2000).

Occurrence of repeatable fish assemblages in relation to abiotic conditions of lakes in the Mississippi Alluvial Valley can provide an empirical means for understanding assemblage patterns, forecasting assemblages of the many lakes for which surveys are not available, designing plans to restore and preserve fish assemblages, and developing lake management priorities. Although a few studies have successfully identified patterns in fish assemblages in relation to abiotic and biotic structuring mechanisms in floodplain lakes, available information is generally limited, and there is virtually no published information for lakes in the Mississippi Alluvial Valley (Crisman 1992). Hence, the purpose of this study was to describe fish assemblages in lakes of the Mississippi Alluvial Valley, to determine whether predictable patterns in fish assemblages developed in relation to prominent abiotic characteristics of lakes, and to explore whether the relevant abiotic factors could be linked to specific assemblage structuring mechanisms.

Study Lakes

The Mississippi Alluvial Valley is an area of little topographic relief, sloping towards the Gulf of Mexico an average of 10 cm/km (Baker et al. 1991). At the onset of the 19th century, the region was known as “The Wilderness” because it contained about 10 million hectares of nearly impenetrable bottomland hardwoods, half of which were flooded 4–5 months of the year (Smith 1954). Another distinct feature of the valley is its dense, alluvial clay and associated back-swamp deposits that historically supported extensive wetland areas. These clays have created low-permeability soils that limit the ability of rainwater to infiltrate the ground surface, and may cause runoff from agricultural fields to rapidly enter wetlands and streams (Chesters and Shierow 1985; Myers et al. 1985). Along with rivers and wetlands, the Mississippi Alluvial Valley contains scores of fluvial lakes (Figure 1) that have long been recognized as unique ecosystems in North America.

Lakes of the Mississippi Alluvial Valley exhibit a wide range of physical attributes. Surface areas range from about 10 ha to over 2,000 ha, and sizes may be up to 2 km wide and 35 km long (USACE 1975). These floodplain lakes are mostly arched and elongated, generally created when a meandering channel is cut off from the main channel at both ends. However, a lake may include only a fraction of a meander, giving it a slightly arched shape, have a full meander like an oxbow lake, or include several connected meander loops when a long section of channel has been abandoned. Morphology is further diversified by age, which differs by hundreds of years during which lakes surrender to successional processes that produce a continuum of predictable features. Older lakes are generally farther away from the river channel, and therefore may experience reduced frequency and intensity of flooding. Many of these lakes, except those inside river levees, are partly or entirely surrounded by agricultural lands that typically disturb drainage patterns, increase lake turbidity, and accelerate sedimentation (Cooper and McHenry 1989).

About 70 fish species reside in lakes of the Mississippi Alluvial Valley, at least 24 of which are common to abundant (Baker et al. 1991). Several of the abundant species are ubiquitous within the river ecosystem, including gars Lepisosteus spp., shads Dorosoma spp., buffalos Ictiobus spp., catfishes (Ictaluridae), and freshwater drum Aplodinotus grunniens. A number of species are characteristic of floodplain lakes, including largemouth bass Micropterus salmoides, crappies Pomoxis spp., and sunfishes Lepomis spp. Whereas few of the collections that are available for these lakes have targeted small fishes, species of the genera Fundulus, Labidesthes, Notropis, and Etheostoma are widespread (Ross 2002).

Within the Mississippi Alluvial Valley, we studied lakes of the Mississippi (N = 15) and Yazoo (N = 14) river basins (Figure 2). Lakes of the Yazoo River basin were sited in the state of Mississippi, whereas lakes of the Mississippi River basin were at or near the Tennessee–Arkansas, Mississippi–Arkansas, or Mississippi–Louisiana state boundaries. Lakes of the Mississippi River occurred either inside (N = 12) or outside (N = 3) the main-stem levee system. Lakes inside the levees are periodically connected to the river when the river rises, but the frequency and duration of connection varies depending on the elevation of the narrow channel that often joins each lake to the river. Lakes outside the Mississippi River levees flood less often, and receive water mainly from heavy rainfall and drainage ditches. Relative to the Mississippi River, the Yazoo River basin is only minimally controlled by levees. Lakes of the Mississippi River basin are generally wider than lakes of the Yazoo River basin. Aquatic macrophytes are uncommon in these lakes; nevertheless, in some of the least-turbid lakes, macrophytes can become abundant in some years (as much as 30–40% area coverage).

Methods

Results

In all, over 98,000 fish representing 61 species were collected during 331 h of electrofishing. Of these, 62,300 fish of 19 species pooled into 14 taxa were retained for analyses (Table 1). Other than shads, the species we excluded were collected in low numbers, or represented small-bodied species likely not adequately sampled by electrofishing. Of the species included, bluegills, longear sunfish, and largemouth bass were the most numerous, whereas redear sunfish and green sunfish were the least numerous. All gars (three species), buffalos (three species), and temperate basses (two species) were grouped into three genera for analysis. Although these species are adequately sampled by electrofishing, they were collected in low numbers, and because species within each genus generally have similar ecological roles, we considered it reasonable to combine them.

Table 1. Percentage composition of the 62,300 fish of 19 species pooled into 14 taxa groupings and retained for analyses of the relation between environmental variables and fish assemblages in Mississippi Alluvial Valley lakes

Assemblage composition and physical characteristics varied several-fold among the study lakes (Table 2). Bluegills, largemouth bass, and gars exhibited the greatest ranges in percent frequency of occurrence, and warmouths and green sunfish the smallest. Of the physical variables, A exhibited the greatest amount of variability and Secchi visibility the least. Maximum depth of the lakes averaged 5.7 m (range, 3–11 m, N = 14), and mean depth averaged 3.1 m (1.5–4.3 m, N = 8). These variables were not included in the CCA because they were not available for all lakes; in lakes with available depth records, correlation between A and maximum depth was high (r = 0.84, N = 14). Many of the physical variables included in the CCA were correlated, but correlations between lake morphology variables and water clarity were low (Table 3).

Table 2. Minimum, mean, and maximum percentage frequency of occurrence of the fish assemblage and description of physical variables for 29 lakes of the Mississippi Alluvial Valley

Table 3. Pearson product–moment correlations among physical variables of 29 Mississippi Alluvial Valley lakes included in the canonical correspondence analysis

The M based on percentage composition revealed no significant correlation (M = −0.12, P = 0.57) between collection year and assemblage structure among lakes. Thus, proximity between collection years was a poor predictor of similarity in assemblage structure: in other words, lakes were temporally independent. Moreover, results from this test suggested that no increasing or decreasing trends were detectable during the study period.

Among the six physical variables included in the analyses (Table 2), three were retained by the forward selection procedure (A, L–W ratio, and Secchi visibility). None of the three physical variables exhibited a dominant influence over fish assemblage, and the correlations among these variables were generally low (r ≤ 0.21, Table 3). The correlations (ter Braak 1986) between species assemblages and physical variables were significant for axis 1 (r = 0.83, P = 0.01) and axis 2 (r = 0.80, P = 0.03), but not for axis 3 (r = 0.46, P = 0.65). Axis 1 (28% of the total variance) represented a gradient that contrasted temperate basses, redear sunfish, gars, black crappies, largemouth bass, and warmouths against white crappies, orangespotted sunfish, channel catfish, buffalos, and freshwater drum with regard to Secchi visibility and lake elongation. Correlations between axis 1 and Secchi visibility, A, and L–W ratio were −0.85, −0.31, and 0.62, respectively. Axis 2 (17% of the total variation) represented a gradient that contrasted redear sunfish, gars, buffalos, and warmouths against temperate basses, channel catfish, orangespotted sunfish, longear sunfish, and freshwater drum with regard to A. Correlations between axis 2 and Secchi visibility, A, and L–W ratio were −0.03, 0.94, and −0.06, respectively.

The CCA ordination diagram revealed various patterns of fish assemblages relative to environmental variables (Figure 3). Redear sunfish, warmouths, and gars were positioned at the high end of the Secchi visibility line, the low end of the lake L–W ratio line (if that line is extended in the opposite direction of the arrow), and the low end of the A line, and tended to be better represented in fish assemblages of clear, short, small lakes. Similarly, temperate basses tended to have higher incidence in clear, short, large lakes, while black crappies and largemouth bass tended to be better represented in clear, short, mid-sized lakes. Buffalos were more common in assemblages of turbid, long, small lakes, whereas white crappies were better represented in turbid, long, mid-sized lakes, and channel catfish and orangespotted sunfish were better represented in turbid, long, large lakes. Freshwater drum and longear sunfish occurred in moderately turbid, mid-length, large lakes. Bluegills and green sunfish were positioned near the centroid or mean of the physical variables, indicating that these species had higher occurrences near the average for all three environmental variables considered, or perhaps that these species are habitat generalists.

Lakes were also examined in the CCA ordination diagram relative to the environmental variables (Figure 4). Lake scores represent the weighted mean taxa scores (ter Braak and P. Šmilauer 1998). Lakes positioned to the right of the ordination diagram are more turbid and elongated than those located to the left, and those located near the top of the diagram are larger than those located near the bottom. The diagram indicated that, with exceptions, lakes within the Mississippi River levees tended to be clearer than those outside the levees, and lakes of the Yazoo River basin tended to be more turbid, smaller, and more elongated than those of the Mississippi River basin. Nevertheless, the lakes ranged widely in relation to the three environmental variables.

Discussion

Patterns of fish assemblages in lakes of the Mississippi Alluvial Valley were linked to environmental variables, suggesting an environmental basis for understanding fish assemblages and for developing management and conservation plans. Lakes within the Mississippi River levees generally had different fish assemblages and physical characteristics than lakes outside the levees or lakes in the Yazoo River basin, but assemblages were similar when physical characteristics of the lakes were alike. The distributions of the 14 taxa in the 29 lakes were governed primarily by two gradients that contrasted assemblages relative to A, lake elongation, and water clarity. Thus, knowledge of whether a lake was clear or turbid, large or small, and long or short helped to predict the predominant taxa in the lake.

The habitat and water characteristics of floodplain lakes are implied by their sizes and shapes, which in turn reflect characteristics such as relative amounts of pelagic and littoral habitats, depth, and water chemistry. Lake area, relative L, and water clarity are interrelated, and are proximal factors for a variety of environmental variables that may structure fish assemblages. For example, lake depth is highly correlated with A, and depth dictates daily and seasonal patterns in water temperature and dissolved oxygen. At Eagle Lake (included in our study), depth was inversely related to the contribution of sediment respiration to dissolved oxygen concentration in the water column (Miranda et al. 2001). Because small lakes typically experience more extreme fluctuations in physicochemistry, they tend to support fish assemblages with more tolerant species (Rahel 1984). Additionally, the lake L–W ratio has been linked to water clarity of fluvial lakes through sediment resuspensions produced by the interaction of fetch (distance over which wind blows) and depth (Hamilton and Lewis 1990). Fetch determines the size of wind-driven waves that form at the lake surface, whereas depth determines the degree to which the waves agitate the sediment surface. Fetch and depth are usually large in large lakes and small in small lakes, and therefore moderate the effect that waves exert on sediment resuspension. However, elongated fluvial lakes have a higher fetch-to-depth ratio (Hamilton and Lewis 1990), and thus are more affected by resuspension-induced turbidity.

In the Mississippi Alluvial Valley, lake size and shape also reflect the intensity of interactions with disturbed watersheds. Bottomland hardwood forests that once covered the region have been cleared, and wetlands have been drained to facilitate agricultural use of the rich alluvial soils. Agricultural development has since reduced the forest to 20% of its original area, with half of the remaining forests lingering inside the Mississippi's main-stem levee system (Forsythe 1985; Gore and Shields 1995). Most forests remaining in the former floodplain are small, fragmented blocks left within patches too wet for farming. Only narrow or nonvegetated riparian zones have been left to buffer lakes, and therefore eroded soils and particulate organic matter may enter lakes throughout their perimeters and through drainage ditches directed into lakes. Before deforestation, the forests slowed runoff after rains, and lakes were fed slowly from woodlands and wetlands (Kleiss 1996; Lowrance et al. 1997). Small lakes are more likely to be affected by their watersheds than large lakes. Indeed, sedimentation rates are reportedly higher in small lakes in the region (Ritchie et al. 1979; McHenry et al. 1982; Cooper and McHenry 1989). Moreover, watersheds potentially affect elongated lakes more severely because of their higher shoreline-length-to-area ratios.

Sedimentation of the colloidal clay particles responsible for turbidity is probably the most important process contributing to changes in the physicochemistry of shallow lakes surrounded by disturbed watersheds (Cooke et al. 1993). Sedimentation rates averaging 3–7 cm/year have been reported in some of the study lakes (Ritchie et al. 1979; McHenry et al. 1982; Cooper and McHenry 1989). Sedimentation reduces lake depth and water storage capacity, and thereby affects the distribution of heat, dissolved oxygen concentration, and other vital water quality parameters (Scheffer 2001). Imported sediments are often enriched with nutrients and organic matter that become available to stimulate new primary production, resulting in further eutrophication of the lake. The mineralization of allochthonous organic matter results in increased water column and sediment respiration. The incidence and extent of low dissolved oxygen concentration is exacerbated by decreased depth, because there is less oxygen in the overlying water column to support benthic respiration (Miranda et al. 2001). Loss of depth further promotes turbidity, primarily through the resuspension of sediments induced by wave action (Hamilton and Lewis 1990), as well as through the stirring action of benthivorous fishes searching for food (Scheffer 2001). The environmental changes stimulated by turbidity and sedimentation not only influence structure and function of the biotic community, but also the esthetic and recreational values of floodplain lakes.

In the Mississippi Alluvial Valley, the fate of lakes appears to be intricately linked to inputs from the watershed due to intensive agriculture, highly erodible soils, and rainfall that averages 125–150 cm annually. Turbidity is maintained most of the year by high sediment loading from the watershed and wind-induced resuspension, although in late summer and fall, turbidity also may be contributed by high algal biomass resulting from excessive nutrient loads (Kleiss et al. 2000). Sustained high levels of turbidity reduce the photic zone and can interfere with the foraging efficiency and behavioral adaptations of some species (Bruton 1985; Miner and Stein 1996; Sweka and Hartman 2001). In turbid lakes, the occurrence of schooling predators (e.g., yellow bass, white bass, and largemouth bass) might be hampered by weakening of school cohesiveness and coordination under poor light conditions (Vandenbyllaardt et al. 1991). The schooling yellow bass once occurred in streams and lakes of the Yazoo River basin (Hildebrand and Towers 1927; Hildebrand 1933), but have not been collected in recent decades. A reduction in predator representation was correlated with increased turbidity and increased representation of small white crappies, some lepomids, buffalos, and catfishes that exhibited slow growth. In turbid waters, buffalos and catfishes further benefit from their tactile bottom-feeding adaptations. Reduced benthic production associated with excessive turbidity and sedimentation was reflected by diminished representation of the redear sunfish, which thrives in relatively clear waters, where it feeds on snails and other benthos (Huish 1958). Similarly, black crappies were replaced in turbid lakes by white crappies. Although no major differences are apparent between these two species in their ability to detect prey in turbid water (Barefield and Ziebell 1986), black crappies reportedly depend on benthic invertebrates more than white crappies (Ball and Kilambi 1973). Sedimentation in the Yazoo River basin has rendered substrates unsuitable for many benthic invertebrates (Cooper 1987), and has contributed to the demise of fish species that rely on benthic insect larvae and crustaceans during at least part of their life cycles. In fact, the redear sunfish was once abundant in lakes of the Yazoo River basin, but has now virtually vanished (Hildebrand 1933; Lucas 1985). Furthermore, turbidity was inversely correlated with fish species that associate with aquatic vegetation (e.g., gars and black crappies).

Our analyses suggest a conceptual model for fish assemblages in lakes of the Mississippi Alluvial Valley, wherein assemblages are structured by abiotic factors through limitations on foraging and on physiological tolerances. Foraging by visual piscivores is expected to be limited in turbid lakes, such that their representation in the fish assemblage declines. Conversely, prey species would then expand through decreased vulnerability to predation. Via sedimentation, substrates would become loose and no longer suitable for the diverse assemblage of invertebrates that attach to plants and detritus, and fish species that forage on benthic invertebrates would become less abundant. Tactile, riverine, nonvisual species that forage by sucking on sediment would be predisposed to benefit. High nutrient loadings and sediment oxygen demand, combined with reduced depth, should contribute to high diurnal fluxes in dissolved oxygen and temperature that tend to select against less-tolerant organisms such as insect larvae (other than chironomids) and piscivorous fishes. In advanced stages of sedimentation, fish assemblages in lakes of the Mississippi Alluvial Valley would be expected to include largely species that thrive in turbid, shallow systems with few predators and low oxygen contents, such as buffalos, catfishes, orangespotted sunfish, and white crappies.

Determinism in assemblage organization of floodplain lakes relative to recurrence in physicochemical features has been documented in the floodplains of two major, unaltered rivers in South America. Tejerina-Garro et al. (1998) concluded that transparency and depth were significantly related to fish assemblage structure in lakes of the Amazon River floodplain. Visually-oriented fishes had highest abundances in clear lakes, whereas fishes with adaptations for low visibility were most abundant in turbid lakes. Lewis et al. (2000) concluded that, despite their vast spatial distribution in relation to the Orinoco River, lakes in the Orinoco floodplain showed a surprising degree of order and repetition in their physicochemical environment, principally transparency, which in turn led to order and predictability in their biotic components. Annual repetition of the major hydrological events forced a strong degree of repetition in biotic events on the floodplain. Whereas the Mississippi Alluvial Valley has been the subject of major anthropogenic interventions and is not entirely functional like the floodplains of South American rivers (Baker et al. 1991), its fish assemblages apparently remain deterministic and are influenced by the same key factors, although now shaped by different circumstances. Nevertheless, we recognize that the manifestation of deterministic or stochastic processes may depend on how one defines the assemblage (Rahel 1990). Sampling limitations generally require assemblages to be defined in terms of species and life-stage subsets; depending on how the subsets are constructed, assemblages could possibly be characterized as either deterministic or stochastic. Our analyses were limited to 14 relatively long-lived taxa and to individuals 7 cm and longer, potentially focusing on the least stochastic portion of the assemblage.

The fish assemblage patterns we identified relative to physical characteristics of lakes in the Mississippi Alluvial Valley suggest three general conservation and management foci. Because turbidity and sedimentation are the leading factors affecting fish assemblages, and the land surrounding many of the lakes belongs to the private sector, lake and fishery managers must collaborate with farmers and organizations involved in watershed management in the region. Large-scale changes in the environment require an expanded level of human resources involved in management, and therefore the need for partnering between government agencies, nongovernmental organizations, and the public (Miranda 2003). Established approaches to control erosion include slowing and diverting runoff, vegetative filter strips, development of riparian filtration zones and wetlands, reduced or no-till agriculture, and reforestation (Hairston et al. 1984; Ewing 1991; Higgins et al. 1993; Mitsch 1994; Haycock et al. 1997). A second focus is restoration of lake habitats. The major habitat issue in lakes of the region is sedimentation, which affects substrate characteristics and causes depth reductions that, in turn, affect physicochemistry of the water. Removal of sediments alleviates shoaling and increases substrate firmness, whereas increasing water level eases only shoaling; however, both approaches will step back succession and restore earlier fish assemblage stages (Peterson 1982; Filipek et al. 1993). The observed reductions in density of piscivorous species are undesirable from fish community structure and recreational fishery perspectives. Thus, the third focus for management is manipulating fish populations through harvest regulations to protect key predators and spawners, stocking to supplement year-classes weakened by disrupted environments, and removing fish to adjust assemblages that have lost their natural evenness (Kohler and Hubert 1993). This step, however, is merely a temporary fix unless watershed and habitat disorders are addressed. Small, elongated lakes are the ones most seriously impacted and should be a restoration priority.

An important limitation of our study was that electrofishing provided a biased representation of the suite of species present in the study lakes. Such an incomplete view of the fish assemblage is a common weakness of field studies, as all collection gears are species selective and size selective to various extents. However, we suggest that the relationships identified here are fundamental to other segments of the fish assemblages in lakes of the Mississippi Alluvial Valley. Nevertheless, we caution that it is also plausible that abiotic factors not considered in our study could be equally or more relevant to species not considered in our analyses. Consequently, restoration and conservation efforts directed at other segments of the assemblages may benefit from further studies.

Acknowledgments

We thank D. Jackson, B. Hubbard, and H. Schramm for helpful reviews, and the MDWFP for its generous support of this research.