Using Fish Distributions and Behavior in Patchy Habitats to Evaluate Potential Effects of Fragmentation on Small Marsh Fishes: a Case Study

J. Great Lakes Res. 31 (Supplement 1):197–211 Internat. Assoc. Great Lakes Res., 2005

Using Fish Distributions and Behavior in Patchy Habitats to Evaluate Potential Effects of Fragmentation on Small Marsh Fishes: A Case Study

Jennifer Jacobus* and Paul W. Webb School of Natural Resources and Environment University of Michigan 430 East University Avenue Ann Arbor, Michigan 48109-1115

ABSTRACT. Habitat loss and fragmentation are global anthropogenic threats to aquatic ecosystems and biodiversity. Animal distribution and behavior are affected by changes in the geometry and structure of their environment. Therefore, predicting the effects of fragmentation on species assemblages requires a working knowledge of how individual species respond to habitat patches and boundaries. We measured the distribution and behavior of fishes in a northern Great Lakes coastal marsh in Mismer Bay, Lake Huron, a naturally patchy ecosystem. Using various sampling techniques, including mark-recapture and direct observation by snorkeling, we determined the relationship between fish species distribution, biotic and abiotic patch characteristics, and patch area. For dominant species we determined habitat utiliza- tion, mobility patterns, and behavioral patterns within patches and at boundaries. Fish abundance was related strongly to average water depth and the percent cover of floating and emergent macrophytes. Minimum patch area to maintain species assemblages was 128 m2 for individual patches and 400 m2 for major patches combined. Fishes that were widely distributed, highly mobile, and most abundant were expected to be least affected by increasing isolation and configuration. Given all results, we expected four species to be least affected overall by fragmentation in Mismer Bay: bluntnose minow (Pimephales notatus), brook stickleback (Culea inconstans), common shiner (Luxilus cornutus), and rock bass (Ambloplites rupestris). Although our conclusions are site-specific, the investigative processes we describe are applicable to fragmentation in other ecosystems. INDEX WORDS: Habitat fragmentation, coastal marsh fishes, patch dynamics, species-area relationship, isolation, configuration.

INTRODUCTION sizes separated by a matrix of a different, usually Landscapes and waterscapes are comprised of less hospitable, habitat type. Configuration de- patches that are natural, dynamic phenomena vary- scribes various forms of incision such that frag- ing in space and time. One of many anthropogenic ments are not specifically isolated, but remain threats to patches is fragmentation, which acceler- connected by corridors (Harrison and Bruna 1999, ates the rate at which patchiness is created and de- With et al. 1999). Isolation and configuration both stroyed. For example, shorelines are prime alter the geometric features of patches and frag- locations for residential and recreational develop- ments, which are measurable in terms of changes in ment, making them highly susceptible to fragmen- core area and edge-to-area relationships (Fagan et tation. Patches and fragmentation processes may be al. 1999, Rosenblatt et al. 1999, With et al. 1999). characterized by mapping spatial patterns on a Understanding the role of patchiness in ecology plane bounded by axes of increasing isolation and is of increasing concern for managers as anthro- increasing configuration (Fig. 1). Isolation leads to pogenic fragmentation events become more wide- the formation of inhabitable fragments of various spread. Management practices and restoration plans must be constructed from a comprehensive information base describing effects of patchiness on ani- *Corresponding author. Current contact information: Environmental Sci- ence Associates, 707 Wilshire Blvd., Suite 1450, Los Angeles, CA mal movements, population parameters, and 90017. E-mail: [email protected] assemblage composition (e.g., Haddad 1999, Han-

197 198 Jacobus and Webb

creases core and increases edge-to-area ratios. For example, some piscivorous fishes are ambush predators that gain greater access to prey at edges (Savino and Stein 1982, Killgore et al. 1989, Toepfer and Fleeger 1995). Alternatively, for prey fishes, loss of habitat may increase population densities in core areas, thereby increasing competition, reducing growth rate, recruitment, and survivorship (Werner et al. 1983). As a result of these factors, separating potential contributions of core and edge to assemblage composition is extremely difficult. A second issue pertains to complexities of experimental design. Empirical studies of patch dynam- FIG. 1. Fragmentation effects can be identified ics ideally involve many patches with varying by mapping spatial patch patterns on a plane characteristics. However, the creation of fragments bounded by axes of increasing isolation and for experimental purposes might be undesirable, increasing configuration. particularly in fragile or threatened habitat. There- fore, pre-existing anthropogenic fragmentation ap- pears to provide an ideal opportunity to study patch dynamics because often there are ranges in frag- ski 1999, Bélisle et al. 2001, Ricketts 2001). In ment characteristics. However, patch diversity general, patch dynamics can be quantified in terms might be found only at dispersed sites where site- of fluxes of materials, nutrients, energy, and organ- specific confounding factors may affect results. In isms within and among patches and across bound- aquatic systems, for example, human activities that aries (Cadenasso et al. 2003a). This topic is being create fragments are associated with other environ- studied at many scales to understand the distribu- mental impacts, such as pollution, boat traffic, and tion, persistence, and productivity of organisms dredging (Brazner 1997), diminishing the ability to (Belnap et al. 2003, Cadenasso et al. 2003b, Fagan separate fragmentation effects from other potential et al. 2003), especially economically important taxa effects. Barriers also cannot be used to create such as fishes (Frederick 1997b, Hastings and Bots- patches in aquatic systems due to the well-known ford 2003, Mumby et al. 2004). Theoretical ad- attraction of fishes to structures (FAD effect) (D’Itri vances in patch dynamics combine ideas from 1986), which prohibits the replication of edge-core island biogeography (MacArthur and Wilson 1967, relationships. Magnuson et al. 1998), metapopulation dynamics It is difficult to measure directly the effects of (Hanski and Gyllenberg 1993, Hanski and Gilpin patches on animals. Thus, indirect methods that 1997), and edge-mediated effects (Fagan et al. focus on behavior are often used to determine how 1999). animals move around patchy environments. Mark- Data collection to quantify effects of fragmenta- recapture techniques and radiotelemetry have been tion on organisms, test theoretical assumptions, and effective in determining spatial and temporal use of evaluate management plans provides special chal- habitat by fishes (Block et al. 2001, Sibert and lenges, such as quantifying the movement behav- Nielsen 2001, Diana 2004). Such studies are most iors of individuals in response to heterogeneous effective at coarse scales and with large-bodied environments (Ovaskainen 2004). Furthermore, ex- species, such as northern pike (Esox lucius). Differ- perimental research on patches and fragments is ent approaches are required at finer scales to eluci- complicated by several factors. First, patch proper- date patch use (e.g., McIvor and Odum 1988, Berry ties co-vary. Division of habitat into smaller and Bock 1999, Eggleston et al. 1999, Golden and patches and fragments not only reduces the core Crist 1999, Hargis et al. 1999), especially along area of each patch relative to that of the original shorelines where human activities abound and habitat, but also simultaneously increases the edge- threaten ecosystems. Studies at this scale are not to-area ratio. Some species and life history stages common in aquatic systems, although there are respond differently to core area versus edge some notable exceptions based on observations of (Toepfer and Fleeger 1995, Hager 1998). Taxa that fish behavior (Potts 1970, Helfand 1981, Bellwood utilize edge may benefit when fragmentation de- and Wainwright 2001). Fragmentation: Small Marsh Fishes in Patchy Habitats 199

The nature of patches in aquatic systems, con- cause vegetation zones in the aquatic bed were eas- founding effects of anthropogenic and natural di- ily identified to even the casual observer. versity factors, and lack of information on patch use in threatened shoreline systems argues for increased PATCH CHARACTERIZATION multiple-technique empirical studies on species as- The marsh waterscape of Mismer Bay is a patch- semblages to understand and predict consequences work of discrete habitats. Preliminary observations of patchiness and fragmentation on fishes (Sven- of five marshes in Les Cheneuax showed that Mis- ning 1998, Berry and Bock 1999, Golden and Crist mer Bay was most extensive and included all types 1999). Here we extend these theoretical and practi- of patches found among Les Cheneuax marshes. A cal ideas using a northern Great Lakes coastal wet- transect was selected in Mismer Bay, covering ap- land as a case study for addressing questions at an proximately 300 m from the shoreline to open water organismal level. We draw on new information and traversing the full range of patch features. We from our current studies to illustrate how appropri- sampled macrophytes along the transect using five ate information can be obtained and how basic data 0.25 m2 quadrats every 30 to 50 m along the tran- on fish distribution, mobility, habitat fidelity, and sect during May through August of 2002 and 2003. behavior in patchy habitats might inform problems We used quantitative measurements of macrophyte of fragmentation. richness; percent cover of emergent, floating, and We focused on the aquatic beds of well-estab- submerged macrophytes; emergent stem density; lished, relatively pristine coastal marshes. We first depth; and temperature to delineate patches within the bay (Table 1). A polythetic agglomerative hier- identified the spatial pattern of habitat patches and archical clustering analysis (PAHC) confirmed the boundaries in order to define and compare their presence of four patches corresponding to those we physical characteristics. We then used exemplary observed, which also agreed with Albert’s (2003) patches and critical boundaries to compare the resi- description of marsh zonation in a Great Lakes pro- dent forage fish assemblage composition and move- tected embayment. We defined these patches as per- ments and behavior of assemblage members manent mixed (PM), seasonal mixed (SM), between patches. The marsh, located within Mismer seasonal emergent (SE), and open submerged (OS) Bay in the Les Cheneaux Islands of northern Lake (Jacobus and Ivan 2005). Permanent marsh was Huron (approximately 46°00′ N, 84°28′ W), was closest to the shoreline, and hence, it was the shal- particularly suitable for studying patchiness be- lowest patch type. It was characterized by high den-

TABLE 1. Characteristics of four patches in Mismer Bay identified by polythetic agglomerative hierar- chical clustering analysis using measurements of physical properties along a transect from the shoreline to open water (Jacobus and Ivan 2005). Average Maximum Seasonal or Dominant Water Dominant Plant Growth Percent Cover Patch Permanent Vegetation Depth Species Form by Species PM Permanent Mixed 11–48 cm Eleocharis smallii Emergent 40% Potamogeton crispus Emergent 50% Schoenoplectus tabernaemontani Emergent 40% Nuphar variegata Floating 85% Scirpus subterminalis Submerged 100% Utricularia vulgaris Submerged 40%

SM Seasonal Mixed 18–67 cm Schoenoplectus tabernaemontani Emergent 15% Nuphar variegata Floating 25% Utricularia vulgaris Submerged 5%

SE Seasonal Emergent 40–93 cm Schoenoplectus tabernaemontani Emergent 30%

OS Open water Submerged 62–105 cm Myriophyllum spp. Submerged 80% 200 Jacobus and Webb

sities of emergent macrophytes, primarily softstem used throughout the remainder of the study. In one bulrush (Schoenoplectus tabernaemontani), with portion of our study we made direct observations of substantial development of floating and submerged fishes using snorkeling techniques, as well. plants, primarily yellow pond lily (Nuphar varie- To determine the relationship between fishes and gata) and common bladderwort (Utricularia vul- patch characteristics, gangs of five traps were lo- garis), respectively. The transition from permanent cated at right-angles to the same transect where to seasonal marsh was readily distinguishable dur- physical characteristics and vegetation were mea- ing spring months by the excision of all stems at a sured. Gangs were comprised of five traps spaced depth of about 5 cm below the water surface by evenly along a 30 m line. There were nine gangs in winter ice. In turn, seasonal marsh was divided into 2002 and eight gangs in 2003, for a total of 45 and two patches. An inshore seasonal mixed habitat 40 traps in each year, respectively. Fish were sam- contained macrophyte species also common in per- pled continuously from all traps from May through manent marsh, although they occurred at compara- August in both years. All minnow traps were tively lower densities. Emergent softstem bulrush checked every 48–72 hours, and fishes were identi- dominated the deeper, offshore seasonal emergent fied to species, counted, and released into the patch, which abruptly ended at open water. Open marsh. This protocol was applied to all aspects of water contained large areas of submerged vegeta- our study involving minnow traps. tion, primarily milfoil (Myriophyllum spp.), cover- Fish distribution varied by species among ing the bottom. patches as expected (see Table 2). Some fishes, A stepwise discriminant analysis (p = 0.05 to such as bluntnose minnow (Pimephales notatus), enter, p = 0.10 to remove) of all habitat variables brook stickleback (Culaea inconstans), and rock showed that patches were primarily characterized bass (Ambloplites rupestris), were widely distrib- by depth, which was not surprising because patches uted. Others were restricted to one patch, such as were located along the depth gradient from onshore northern redbelly dace (Phoxinus eos) in permanent to offshore. (2002: n = 34, df = 12, Wilks’ Lambda marsh. Adjacent patches were most similar in fish = 0.021, p = 0.000; 2003: n = 40, df = 12, Wilks’ assemblage composition. For example, in 2003, Lambda = 0.043, p = 0.000). Patches also differed nearshore PM and SM patches shared 25 out of 31 significantly in the percentage of surface area cov- species (81%), and offshore SE and OS patches ered by submergent and emergent plants (Jacobus shared 14 out of 15 species (93%). The largest dif- and Ivan 2005). ferences occurred at patch extremes, so that PM and OS shared only 13 out of 31 species (42%) in 2003. Although habitats differed largely in species that FISH ASSEMBLAGES IN PATCHES were rare in abundance (e.g., single samples of Fish species richness and abundance are affected muskellunge (Esox masquinongy) in PM, and bur- by habitat complexity, primarily macrophyte diver- bot (Lota lota) in OS), they also differed in locally- sity in both species and associated growth forms abundant species that were constrained in range (Tonn and Magnuson 1982, Rahel 1984, Randall et (e.g., 1,000+ northern redbelly dace in PM) as al . 1996, Weaver et al. 1997). We sampled fishes in noted above. the four patches identified in Mismer Bay (PM, We used Canonical Correspondence Analysis SM, SE, OS) to determine the relationship between (CCA) (Ter Braak 1986) to relate the distribution of presence-absence and abundance of species and fishes to patch characteristics. All measured macro- patch characteristics. We captured fishes using phyte and physical habitat variables were entered baited Gee™ minnow traps. These proved to be the into the CCA, along with abundances for individual most efficacious gear for sampling nearly all fish species and abundances and average CPUE for species (M.R. Conlon, P.W. Webb, and J.A. Teeri, all fishes combined. We only included twelve indi- unpublished observations), supporting other studies vidual fish species that had a total relative abun- that confirm minnow traps as an efficient, practical dance 3% in the analysis so that rare species method for collecting data in freshwater wetlands would ≥not influence results (Wang et al. 2003). and salt marshes and detecting small-scale habitat CCA results indicated that abundances of six fish use by fishes (Rozas and Odum 1987, Hettler 1989, species had strong correlations with habitat vari- Snodgrass et al. 1996, Halpin 1997). This method ables, primarily average depth and the percent focused our study on the small-bodied, forage fish cover of emergent or floating macrophytes (Jacobus portions of the overall fish community and was and Ivan 2005) (Fig. 2). Species differed in the di- Fragmentation: Small Marsh Fishes in Patchy Habitats 201

TABLE 2. Inventory of fishes in Mismer Bay, Lake Huron, during May through August of 2002 and 2003. Data were obtained with minnow traps and direct observation by snorkeling in four marsh patches defined as permanent mixed (PM), seasonal mixed (SM), seasonal emergent (SE), and open water with submerged vegetation (OS). Marsh Patch PM SM SE OS Petromyzontidae Northern brook lamprey Lampetra appendix X Amiidae Bowfin Amia calva XXX Umbridae Central mudminnow Umbra limi XXXX Esocidae Muskellunge Esox masquinongy X Northern pike Esox lucius XXX Cyprinidae Blackchin shiner Notropis heterodon X Blacknose dace Rhinichthys atratulus X Blacknose shiner Notropis heterolepis X Bluntnose minnow Pimephales notatus XXXX Brassy minnow Hybognathus hankinsoni XXX Common carp Cyprinus carpio XX Common shiner Luxilus cornutus XXXX Creek chub Semotilus atromaculatus XX Fathead minnow Pimephales promelas X Finescale dace Phoxinus neogaeus XXX Golden shiner Notemigonus crysoleucas XX Hornyhead chub Nocomis biguttatus XX Longnose dace Rhinichthys cataractae XX Northern redbelly dace Phoxinus eos X Pearl dace Margariscus margarita XX Sand shiner Notropis stramineus XXXX Spottail shiner Notropis hudsonius XXXX Catastomidae White sucker Catostomus commersoni X XXX Ictaluridae Brown bullhead Ameiurus nebulosus XX XX Fundulidae Banded killifish Fundulus diaphanus XX Gadidae Burbot Lota lota X Gasterosteidae Brook stickleback Culaea inconstans XXXX Ninespine stickleback Pungitius pungitius XX XX Threespine stickleback Gasterosteus aculeatus XX X Percidae Iowa darter Etheostoma exile XXXX Johnny darter Etheostoma nigrum XXXX Logperch Percina caprodes XXX Yellow perch Perca flavescens XXXX Centrarchidae Largemouth bass Micropterus salmoides X Pumpkinseed Lepomis gibbosus XX Rock bass Ambloplites rupestris XXXX Smallmouth bass Micropterus dolomieu XX XX Cottidae Mottled sculpin Cottus bairdi XXXX 202 Jacobus and Webb

TABLE 3. The species-area relationship (SAR) for small marsh fishes in three patches in Mismer Bay, Lake Huron. All species-area equations were generated using a power function and were signif- icant at the p < 0.001 level. a SAR Maximum Patch zcrichness Permanent Mixed 0.20 6.36 19 Seasonal Mixed 0.22 6.01 22 Seasonal Emergent 0.20 6.92 21 aData are slopes (z), constants (c), and maximum cumulative species richness. FIG. 2. Biplot of the first two axes from the canonical correspondence analysis for fish met- ing from 16 m2 to 400 m2. Matrices were installed rics (response variables) and habitat metrics (pre- and checked continuously during the summer of dictor variables) (adapted from Jacobus and Ivan 2002. 2005). Fish species with canonical functions   We described the SAR using a power function to greater than 0.5 were considered strongly corre- z lated to habitat variables: smallmouth bass relate richness, S, and area, A, where S = cA , (Micropterus dolomieu) = mido; sand shiner (Gleason 1922, MacArthur and Wilson 1967, (Notropis stramineus) = nost; central mudmin- Rosenzweig 1995, Neigel 2003); c and z were fitted now (Umbra limi) = umli; pearl dace (Mar- coefficients. SAR patterns were most apparent on a gariscus margarita) = mama; northern redbelly seasonal basis, probably because fish mobility can dace (Phoxinus eos) = pheo; brown bullhead bias measures over short time periods. The expo- (Ameiurus nebulosus) = icne. Only habitat vari- nents (z) of the SARs ranged from 0.20 to 0.22 for ables that strongly correlated with species’ abun- SARs based on 90-day sampling, independent of dances are shown: average depth (DEPTH); per- patch type (Table 3), similar to values expected cent cover of floating macrophytes (%F); and from biogeographic theory (Rosenzweig 1995, percent cover of emergent macrophytes (%E). Brown and Lomolino 1998), and similar to values found in marine ecosystems (see Neigel 2003). As with other SAR studies, species lost with declining area were less numerous (rarer) while abundant rection (positive or negative) of their relationship species were found in subplots of all sizes. The with each variable. For example, smallmouth bass species-area curves in the PM and SE patches also (Micropterus dolomieu) and sand shiners (Notropis appeared to plateau at approximately 128 m2, stramineus) were positively associated with depth, reaching maxima of 19 and 21 fish species (Fig. 3). while northern redbelly dace, pearl dace (Mar- gariscus margarita), and central mudminnows (Umbra limi) were negatively associated with PATCH FIDELITY AND depth. BOUNDARY TRANSIT Boundaries are recognized by spatial changes in SPECIES-AREA RELATIONSHIP (SAR) habitat parameters. The rates of change of these parameters as measured along the transect showed While patches are recognizable by distinct struc- that boundaries among patches in Mismer Bay tural characteristics, they also vary in size. In order marshes were mainly gradual dimensional bound- to determine the effect of patch size on fish assem- aries (Strayer et al. 2003); i.e., between patches, blages, we determined the species-area relationship multiple habitat characteristics varied in small in- (SAR) using square matrices of 25 minnow traps crements over distances of up to 60 m. An excep- constructed within PM, SM, and SE patches. Min- tion was the boundary between SE and OS patches, now traps were set 4 m apart and each trap was as- which was a sharp dimensionless edge. Here the sumed to sample the 4-m diameter area in its percent cover of emergent vegetation changed from vicinity. Thus nested subplots covered areas rang- 30% to 0% and the percent cover of submerged Fragmentation: Small Marsh Fishes in Patchy Habitats 203

FIG. 3. Individual patch and multi-habitat species-area relationships (Tjørve 2002) for Mismer Bay, Lake Huron. Original data are shown for three marsh patches, × = permanent mixed (PM), ■ = seasonal mixed (SM), and ● = seasonal emergent (SE). These data show that the number of species reached maximum values for areas of 128 m2 in PM and SE patches. The species-area curves for each patch are added together for all combinations of patches using the percentage of overlap in species between patches. The results of these additions are shown by the continuous curves, which are the best fit power functions for each combination. The underlying data for the top curve representing all three patches reaches a plateau at approximately 400 m2.

vegetation changed from 0% to 79% over a distance multiple locations within the PM, SM, and SE of approximately one meter. patches between May and August 2002. All fishes We used mark-release-recapture methods to de- were captured and recaptured in the 45 minnow termine mobility across boundaries and habitat fi- traps deployed in gangs along the transect and in delity for patches for the more numerous species. traps used to create matrices for the species-area (Guy et al. 1996). At total of 1,119 individual fish analysis. Representatives of nine species were re- representing 22 species were marked with visible captured, representing 13% of all marked fishes implant fluorescent elastomer (VIE) (Northwest (Table 4). Results were consistent with previous ob- Marine Technology, Inc.). This method has been servations of fish distributions in the various used with fishes as small as 8 mm (Frederick patches of marsh. Species varied in distance trav- 1997a), and age-0 Salmo trutta (brown trout) eled, rate of travel, and their fidelity to specific (Olsen and Vollestad 2001). A study of small gob- habitats. Travel distances ranged from a minimum ies (Coryphopterus glaucofraenum, C. nicholsii, of 0 m for rock bass and northern redbelly dace to a and Lythrypnus dalli) found no effect of VIE tags maximum of 272 m for an individual bluntnose on growth or predation (Malone et al. 1999). minnow traveling from the PM to the SE patch Fishes were captured, tagged, and released at (Table 4). Rock bass traveled at the fastest rate, 204 Jacobus and Webb

TABLE 4. Mobility of nine fish species marked and recaptured in Mismer Bay, Lake Huron. Distances traveled (m) are the maximum, minimum, and mean net distances traveled from point of release to recapture for individuals of each species. Net rates of travel (m/day) account for the number of days it took individuals to cover the distances traveled. The percentage of total individuals for each species that were recaptured in different habitats from the release habitat is expressed in terms of switching habitats (i.e., Switch Patches).

Number Number Distance traveled (m) Rate of travel (m/day) Switch Species Tagged Recaptured Max Min Mean Max Min Mean Patches Pimephales notatus 218 33 272 8 70 108 1 10 31% Ambloplites rupestris 88 29 249 0 112 124 0 36 37% Pungitius pungitius 164 4 218 18 68 15 2 8 25% Phoxinus eos 130 68 45 0 35 22 0 9 0% Culaea inconstans 250 7 181 12 68 91 3 20 43% Luxilus cornutus 20 1 — — 251 — — 18 100% Notropis stramineus 94 1 — — 226 — — 8 100% Phoxinus neogaeus 13 1 — — 20 — — 2 0% Amia calva 57 1 — — 26 — — 13 0% other 85 0

ALL SPECIES 1,119 145 22%

moving 247 m in 2 days and averaging 36 m per likely to affect fish mobility. Therefore, it is most day. Northern redbelly dace, bowfin (Amia calva), likely to indicate how boundaries affect fish behav- and finescale dace (Phoxinus neogaeus) were re- iors and distributions. Snorkeling methods allowed leased and recaptured only within the PM patch. us to determine behavior for fishes of all sizes, not Twenty-two percent of fishes were recaptured in just small-bodied individuals. Behavioral observa- patches different from those in which they were orig- tions at the boundary were compared and contrasted inally caught and released. These fishes therefore with data collected in the same manner within the crossed boundaries when moving among patches. core of adjacent SE and OS patches approximately The bluntnose minnow, ninespine stickleback (Pun- 10 m from the edge. Recorded behavior included for- gitius pungitius), rock bass, common shiner (Luxilus aging, schooling or individual swimming, location in cornutus) and sand shiner all crossed the gradual water column, and direction. Foraging behavior was boundaries that separate PM, SM, and SE patches. In recognizable by darting, mouth opening, pecking, addition, rock bass were recaptured in the OS patch, biting, sucking in sand, algae, or particles, or force- and thus crossed the sharp SE-OS boundary as well. fully filtering water through gills. Only untagged individuals of the other four species A total of 24 fish species were observed at the were captured in the OS patch. edge and in cores of neighboring patches. Nineteen species were observed at the edge, 16 in SE and 11 in OS. Some species were found in one or two of BEHAVIOR: PATCHES VERSUS EDGES these habitats, and overlap between each core habi- Understanding how boundaries affect species dis- tat and the edge was about 50%. Fish abundance tributions requires knowledge of their behavior at also varied among habitats. A total of 1,843 indi- boundaries and within patches. Fish behavior is in- viduals were observed at edge sites during the en- fluenced by macrophyte growth form, stem and plant tire data collection period (700 minutes) compared densities, and phenology of these features (Savino to a combined total of 1,058 fishes in the two adja- and Stein 1982, Wiley et al. 1984, Engel 1988, Kill- cent patch cores. The greater richness and abun- gore et al. 1989, Weaver et al. 1997). We snorkeled dance of fishes at the edge is consistent with at the sharp seasonal emergent/open water (SE/OS) observations in other systems and is attributed to boundary recording fish behavior for 10–15 minutes the creation of habitat diversity by edges (Peterson at dawn, midday, and dusk during the months of and Turner 1994, Kunin 1998). June, July, and August of 2003. We focused on this Behavior varied among species, with edge/core boundary because it is a sharp edge and is most location, and with time of day (Table 5). Benthic Fragmentation: Small Marsh Fishes in Patchy Habitats 205

TABLE 5. Summary of observations made by snorkeling at dawn, mid-day, and dusk at the boundary between the seasonal emergent marsh (SE) and open water with submerged macrophytes (OS) and in the core of the adjacent SE and OS patches. Only 16 of the 24 fishes observed are listed here. The eight miss- ing species were only observed once or twice in one patch. Fishes are categorized as swimming individu- ally or in pairs (I) or in shoals of 3 or more fishes (S). The presence of these individuals or shoals in the three habitats is indicated by “p,” and presence while feeding as “F.” Fish movements along the edge in both directions are indicated by ↑. Movements from the core to the edge and vice versa are shown by ← and →, with arrows in both directions indicating fish passage across the boundary.

Vertical SE Core SE/OS Edge OS Core Species location Group Dawn Day Dusk Dawn Day Dusk Dawn Day Dusk Etheostoma exile benthic I F F F F F F F F p ↑↑← Cottus bairdi benthic I F F p ↑↑→ Rhinichthys cataractae benthic S P ↑ Etheostoma nigrum benthic I, S F F F F F F F F p Percina caprodes benthic I p P p p

mixed shoals mid-water S F F F F F F ←↑→ ←↑→ ←↑→ Notropis stramineus mid-water I, S p p p F Fppp ←↑→ ←↑→ Pimephales notatus mid-water I, S F F F F F F p p F ↑ ↑← Luxilus cornutus mid-water S p F FF ← → ← → ← → Micropterus dolomieu mid-water I p p p p F F p p p ← → ← → ← → Perca flavescens mid-water I F F F p → Culaea inconstans mid-water IF pp ppp pFF ←↑ ← ← Pungitius pungitius mid-water I p p p ↑ ↑

Phoxinus neogaeus mid-water I, S p Fp p p → ← Hybognathus hankinsoni mid-water I F p p → Catostomus commersoni mid-water I F p p ← → ←

Notropis hudsonius mid-water SF 206 Jacobus and Webb

fishes were primarily sessile, occasionally swim- not remaining habitat is sufficient or substitutable ming off the bottom to feed or to cover small dis- for altered patches. Species-area curves can provide tances of a few body lengths. Thus, benthic fishes an initial estimate of habitat area requirements. (e.g., Iowa darters (Etheostoma exile) and the rare Curves obtained for habitat components, such as mottled sculpin (Cottus bairdi)) moved infrequently identifiable marsh patches, can be used to obtain a among patches and traveled across the edge only at more comprehensive relationship, combining patch- dusk (Table 5). Johnny darters (Etheostoma nigrum) specific SARs into one multi-habitat species-area fed as widely as conspecific Iowa darters, but were curve (Tjørve 2002). We constructed a multi-habitat not seen to move out of the feeding areas at the species-area curve for our patchy marsh waterscape times sampled. in an additive fashion, combining equal patch areas Mid-water fishes spent most time in the water while accounting for species overlap (Fig. 3). column, although at times fed on bottom materials. We combined the species-area curves from PM, Mid-water species were more mobile than benthic SM, and SE patches in Mismer Bay in all possible species, moving along and across the edge at all orders (Fig. 3). The two seasonal patches alone ac- times of day (Table 5). Shoals of multiple species, counted for almost all species present in all three mainly cyprinids, also commonly displayed such patches, as illustrated by the slight difference in the behavior while feeding at the edge. Shoals of com- respective curves (Fig.3). This analysis implies that mon shiners crossed and fed at the boundary all if only the PM patch were altered, species loss day, but they were rarely observed in core areas. would be minimal since most species were also pre- This use of the edge was largely paralleled by sent in the other two patches. In contrast, losses in smallmouth bass, although this species was also the SM patch would be expected to have the great- seen at all times in the two adjacent cores. Schools est impact on species richness. This conclusion fol- of bluntnose minnows, the most abundant species, lows because the (SE+PM) curve is the lowest of fed at all times at the edge and in SE core, moving all dual patch combinations and is least similar to continuously along, but not really across the bound- the top curve depicting all three patches (Fig. 3). ary. Sand shiners were regularly seen as both indi- Analysis of SARs for areas from 16 m2 to 400 m2 viduals and in shoals in the cores, but only fed at in each patch suggested a minimum patch size of the edge at dawn and dusk. Yellow perch (Perca approximately 128 m2. The multi-habitat SAR sam- flavescens) foraged at the edge, but did not move pled combined areas from 48 m2 to 1,200 m2, tak- along the boundary, behaving much like benthic ing into account habitat structure in all three darters. patches. The minimum area required to maintain Other species foraged only in core habitats at the marsh fish assemblage in Mismer Bay was ap- specific times of day. For example, finescale dace, proximately 400 m2, determined from the plateau in brassy minnows (Hybognathus hankinsoni), and the underlying data for the 3-patch curve white suckers (Catostomus commersoni) fed in SE (PM+SM+SE; Fig. 3). However, two lines of argu- core at dusk, a common observation among fishes ment suggest even the multi-habitat SAR is inade- (Hobson 1974, Helfand 1981), after making diurnal quate alone to estimate minimum habitat migrations across the edge at dawn and dusk. Brook requirements for fishes. First, the number of species sticklebacks and spottail shiners (Notropis hudso- in certain patches, notably PM, determined from the nius) were seen feeding at dawn in SE core, al- SAR is lower than that found by direct sampling of though the former species also fed at mid-day and that patch. The SAR samples were taken deliber- dusk in OS core. ately in marsh areas that were as homogenous as possible, whereas the transect gangs sampled DISCUSSION greater habitat variability. For example, water ex- change between land-marsh-open water involves Evaluating Fragmentation Using sheet flows throughout the entire bay plus localized Species Distributions stream inputs and outputs (Mitsch and Gosselink Determining minimum area requirements and 2000). Transects traversed such features, whereas evaluating the effects of habitat loss on mobile sampling areas for determining the SAR sought to species, such as fishes, in patchy landscapes re- avoid features that could not be included over the quires an assessment of resources in all habitats. full range of areas. Thus, a species-area analysis When patches are removed or reconfigured, species complements patch analysis, and resource managers are affected differentially, depending on whether or should consider that species richness depends upon Fragmentation: Small Marsh Fishes in Patchy Habitats 207

available area within a habitat and the structure of or fragmentation alter edge and core geometry. that habitat. Boundaries may restrict mobility, even for species Second, our results reflect an implicit assump- that are widely distributed. As a result, a broad dis- tion that fragmentation, or an increase in habitat tribution is not necessarily synonymous with large patchiness, is undesirable. However, habitat suit- travel distances or frequent boundary crossings. Fur- ability indices widely used in stream assessment thermore, species that do not engage in long-dis- (Baker and Coon 1995, McMahon et al. 1996) mea- tance travel and resulting high dispersal behaviors sure habitat complexity, and more diverse, some- incur the most risk from isolation. For example, times patchy, conditions generally earn higher marsh populations of bluntnose minnows and north- scores. Similarly, numerous studies of shoreline ern redbelly dace would be threatened by patch iso- structure and restoration practices show the desir- lation because neither species was observed crossing ability of diverse, patchy habitats (Engel and the sharp edge into open water, a passage that might Nichols 1994, Engel and Pederson 1998, Jennings be necessary if isolation also coincides with a reduc- et al. 1999). Fragmentation also increases habitat tion in available habitat area. Bluntnose minnows complexity due to increased isolation and configu- have some advantage, however, because they utilize ration, but it is considered to reduce “suitability” more than one patch type and travel longer distances compared to the original unfragmented habitat. This (up to 272 m) than northern redbelly dace (up to 45 apparent discrepancy probably reflects area/scale m). In contrast, rock bass cross all marsh boundaries effects. Relatively large patches, such as beach, and traverse all patches, including those in open cobble-boulder, or marsh, are desirable, while frag- water. Thus, rock bass dispersal behavior makes this mentation within each patch is undesirable. Thus, species less likely to be affected by isolation. These there are minimum sizes for continuous, multi- examples suggest that evaluating the effects of isola- patch habitats to support fishes (Lewis et al. 1996, tion requires more than distribution data. Ideally, With et al. 1999). Therefore, the multi-patch SAR such evaluations should include specific mobility as- approach of species-area analysis and direct patch sessments for each species and their range of travel analysis generally suggests that management of distances. marshes to minimize effects of fragmentation In addition to isolation, fragmentation can create should maximize the area and/or choose the area to complex habitat configurations with isolated patches be protected to include rarer habitat features and connected by corridors and incisions that result in a unique finer-scale patches that are not readily incor- greater proportion of edge habitat and less core porated into determinations of the SAR. habitat (Fig. 1). In a highly configured landscape, fish species with high fidelity to one patch would need corridors in order to move between patches Evaluating Fragmentation Using (e.g., northern redbelly dace). However, as with po- Species Mobility and Behavior tential effects of isolation, if individual fish do not Evaluating the total effects of fragmentation on travel long distances, then corridors might not be species requires detailed information not only about useful to range-restricted species. Fish that swim fish distribution, but also mobility, habitat fidelity, along edges (e.g., bluntnose minnows and sand shin- and behavior patterns, including foraging and ers) are more likely to use corridors to locate iso- schooling behavior at different times of day. We ap- lated patches of habitat than fishes that avoid edges. proached these issues using mark-release-recapture In contrast, fishes that utilize all patches and freely and direct observations of fish behavior. cross boundaries (e.g., rock bass) would likely not Fragmentation events change the geometry of be affected by configuration. landscape patchiness, which can result in isolated In addition, configuration will differentially af- patches. These are delineated, and hence partitioned, fect fish species as core area decreases and the ratio by boundaries or edges, features that serve as “con- between edge and core habitat increases. As edge trol points” to species behavior (Cadenasso et al. habitat increases, fishes that feed at edges, such as 2003a). Macrophyte boundaries and patches within the piscivorous yellow perch, might experience coastal marshes provide important cues to predatory greater access to prey items because predation has and anti-predatory behavior, feeding behavior, and been shown to increase with edge length (Smith mobility patterns (Savino and Stein 1982, Engel 1995). However, species that forage within patches 1988, Killgore et al. 1989). Hence, fishes are af- would likely experience a reduction in available fected when naturally occurring changes in patches food resources as patch area decreases. Fishes that 208 Jacobus and Webb

migrate diurnally into and out of patches might and abundance of fishes. Second, richness and seem resilient to configuration, but these fishes abundance effects fall differentially on various might respond to a decrease in core habitat. If species and can be attributed in at least some part to fishes return to patch cores to feed (e.g., brook fish mobility and behavior. In this second area, we sticklebacks) and patch area is reduced, then suggest two major categories of species to help species abundances might decrease as competition evaluate assemblage changes. One group of species for food resources increases. is narrowly distributed within Mismer Bay, does not A major factor affecting fish behavior at bound- cross boundaries, and individuals travel relatively aries is probably associated with predation risk, and short distances within their resident habitats. An ex- our results are consistent with this idea. Samples ample is the northern redbelly dace, which only oc- collected using gill nets and fyke nets showed that cupies the PM patch and would be highly affected, large piscivorous fishes, such as bowfin, northern perhaps lost from the fish community, if that patch pike, various salmonids, largemouth bass (Mi- were altered. Another group of species is widely cropterus salmoides), and yellow perch, were most distributed, generally is not restricted by patch prevalent at the SE-OS boundary (M.R. Conlon, boundaries, and travels long distances as individu- P.W. Webb and J.A. Teeri, unpublished observa- als. An example is the rock bass, which likely tions), attesting to the general elevated predation would be unaffected by alteration to the PM patch risk at boundary sites. Both rock bass and brook because it utilizes all patches. Species such as the sticklebacks have spined dorsal fins that serve as central mudminnow do not fit into either group, and predator defenses, and both species are prevalent at the effect of alteration on the PM patch to this and across boundaries. Thus, brook sticklebacks species would be uncertain due to its restricted mo- were sampled from all patch types and recaptures bility yet ubiquitous distribution. indicated that they cross the sharp SE-OS edge. Initially, the SAR indicates that species richness Rock bass were the most mobile and widely distrib- declines as area is lost. The species eliminated first uted species, utilizing all patches and crossing all are rare species, as shown by the underlying SAR boundaries. In contrast, predation may make the data. Thus, the composition of the fish community boundary too risky for soft-rayed fishes, such as is expected to change into an assemblage of com- dace, minnows, and shiners (Family: Cyprinidae), mon (abundant) species. In addition, decreases in to cross (Savino and Stein 1982, Killgore et al. habitat area also may increase fish density, and with 1989, Smith 1995). this, competition for space and resources also in- Species’ behaviors also reflected different strate- creases. These biotic factors ultimately reduce over- gies that would minimize predation risks. Benthic all fish abundances. From the abundant species that species are camouflaged, have spiny rays, and are remain given area effects, an increase in configura- less likely to move and attract predators. One ex- tion and isolation is expected to further select ception, the longnose dace (Rhinichthys catarac- among species. As patch geometry and edge-to-core tae), a benthic cyprinid, was observed swimming ratios change, widely-distributed species are ex- along the edge. In lakes, this species has been ob- pected to persist at the expense of range-restricted served moving from shallow water to deeper water species. Thus, based on our observations in Mismer as temperature rises during the summer (Hubbs and Bay, we anticipate communities with greater pres- Lagler 1967) and thus may have been following the ence of the bluntnose minnow, brook stickleback, edge to deeper water. Finally, the most numerous common shiner, and rock bass following a fragmen- fishes, cyprinids, were found most often in shoals, tation event. Our conclusions are speculative and which are known to reduce predation risks for indi- specific to Mismer Bay. However, the investigative vidual group members. Spiny-rayed bass and perch processes we have described here can be applied to were observed at the edge only as individuals, per- systems in other locations in order to assess the ef- haps a less-risky endeavor due to their dorsal spine fects of habitat loss and fragmentation on assem- defenses. blages of mobile species.

CONCLUSIONS ACKNOWLEDGMENTS Our observations can be summarized in terms of This project was supported by the National two main effects. First, reduced area that typically Oceanic and Atmospheric Administration’s Michi- results from fragmentation changes species richness gan Sea Grant Program, Project R/GLF-5; The Na- Fragmentation: Small Marsh Fishes in Patchy Habitats 209 ture Conservancy, Michigan Chapter; the Univer- Diana, J.S. 2004. Biology and Ecology of Fishes, Second sity of Michigan Biological Station; and the Uni- edition. Carmel, Indiana: Cooper Publ. Group. versity of Michigan School of Natural Resources D’Itri, F.M. 1986. Artificial Reefs. Chelsea, Michigan: and Environment. These studies could not have Lewis Publishers, Inc. been accomplished without the dedicated work of Eggleston, D.B., Elis, W.E., Etherington, L.L., Dahlgren, C.P., and Posey, M.H. 1999. Organism responses to numerous students: Katie Boyer, Lori Ivan, Tracy habitat fragmentation and diversity: habitat coloniza- Jones, Jen Metz, Liza Ray, and Meghan Shilts; and tion by estuarine macrofauna. J. Exp. Mar. Biol. Ecol. high school students Aaron Bright, Esther Loeffler, 236:107–132. Phillip Preston, and Nick Schaedig. We also thank Engel, S. 1988. The role and interactions of submersed three anonymous reviewers for their constructive macrophytes in a shallow Wisconsin lake. J. Fresh- critiques of this manuscript. water Ecology 4:329–341. ——— , and Nichols, S.A. 1994. Restoring Rice Lake at REFERENCES Milltown, Wisconsin. Wisconsin Department of Nat- ural Resources, Madison, WI. PUBL-RS186-94. Albert, D.A. 2003. Between land and lake: Michigan’s ——— , and Pederson, J.L. 1998. The construction, aes- Great Lakes coastal wetlands. Michigan Natural Fea- thetics, and effects of lakeshore development: a litera- tures Inventory, Michigan State University Extension, ture review. Wisconsin Department of Natural East Lansing, MI. Extension Bulletin E-2902. Resources, Madison, WI. PUBL-RS177-98. Baker, E.A., and Coon, T.G. 1995. Development and Fagan, W.F., Cantrell, R.S., and Cosner, C. 1999. How evaluation of alternative habitat suitability criteria habitat edges change species interactions. Am. Nat. for brook trout Salvelinus fontinalis. Michigan 153:165–182. Department of Natural Resources. Research Rept. ———, Fortin, M.-J., and Soykan, C. 2003. Integrating 2017. edge detection and dynamic modeling in quantitative Bélisle, M., Desrochers, A., and Fortina, M.-J. 2001. analyses of ecological boundaries. Bioscience 53: Influence of forest cover on the movements of forest 730–739. birds: a homing experiment. Ecology 82:1893–1904. Frederick, J.L. 1997a. Evaluation of fluorescent elas- Bellwood, D.R., and Wainwright, P.C. 2001. Locomo- tomer injection as a method for marking small fish. tion in labrid fishes: implications for habitat use and Bulletin of Marine Science 61:399–408. cross-shelf biogeography on the Great Barrier Reef. ——— . 1997b. Post-settlement movement of coral reef Coral Reefs 20:139–150. fishes and bias in survival estimates. Mar. Ecol. Prog. Belnap, J., Hawkes, C.V., and Firestone, M.K. 2003. Ser. 150:65–74. Boundaries in miniature: two examples from soil. Bio- Gleason, H.A. 1922. On the relation between species and science 53:739–749. area. Ecology 3:158–162. Berry, M.E., and Bock, C.E. 1999. Effects of habitat and Golden, D.M., and Crist, T.O. 1999. Experimental landscape characteristics on avian breeding distribu- effects of habitat fragmentation on old-field canopy tions in Colorado foothills scrub. Southwestern Natu- insects: community, guild and species responses. ralist 43:453–461. Oecologia 118:371–380. Block, B.A., Dewar, H., Blackwell, S.B., Williams, T.D., Guy, C.S., Blankenship, H.L., and Nielsen, L.A. 1996. Prince, E.D., Farwell, C.J., Boustany, A., Teo, S.L. Tagging and marking. In Fisheries Techniques, eds. H., Seitz, A., Walli, A., and Fudge, D. 2001. Migra- B.R. Murphy and D.W. Willis, pp. 353–383. tory movements, depth preferences, and thermal biol- Bethesda, MD: American Fisheries Society. ogy of Atlantic bluefin tuna. Science 293:1310–1314. Haddad, N.M. 1999. Corridor and distance effects on Brazner, J.C. 1997. Regional, habitat, and human devel- interpatch movements: a landscape experiment with opment influences on coastal wetland and beach fish butterflies. Ecological Applications 9:612–622. assemblages in Green Bay, Lake Michigan. J. Great Hager, H.A. 1998. Area-sensitivity of reptiles and Lakes Res. 23:36–51. amphibians: are there indicator species for habitat Brown, J.H., and Lomolino, M.V. 1998. Biogeography, fragmentation? Ecoscience 5:139–147. Second edition. Sunderland, Massachusetts: Sinauer Halpin, P.M. 1997. Habitat use patterns of the mummi- Associates. chog, Fundulus heteroclitus, in New England. I. Intra- Cadenasso, M.L., Pickett, S.T.A., Weathers, K.C., Bell, marsh variation. Estuaries 20:618–625. S.S., Benning, T.L., Carriero, M.M., and Dawson, T.E. Hanski, I. 1999. Metapopulation Ecology. Oxford: 2003a. An interdisciplinary and synthetic approach to Oxford University Press. ecological boundaries. Bioscience 53:717–722. ———, and Gilpin, M. 1997. Metapopulation biology: ——— , Pickett, S.T.A., Weathers, K.C., and Jones, C.G. ecology, genetics, and evolution. New York: Acade- 2003b. A framework for a theory of ecological bound- mic Press. aries. Bioscience 53:750–758. ———, and Gyllenberg, M. 1993. Two general metapop- 210 Jacobus and Webb

ulation models and the core-satellite hypothesis. Am. McIvor, C.C., and Odum, W.E. 1988. Food, predation Nat. 142:17–41. risk, and microhabitat selection in a marsh fish assem- Hargis, C.D., Bissonette, J.A., and Turner, D.L. 1999. blage. Ecology 69:1341–1351. The influence of forest fragmentation and landscape McMahon, T.E., Zale, A.V., and Orth, D.J. 1996. pattern on American martens. J. Appl. Ecol. Aquatic habitat measurements. In Fisheries Tech- 36:157–172. niques, eds. B.R. Murphy and D.W. Willis, pp. Harrison, S., and Bruna, E. 1999. Habitat fragmentation 83–120. Bethesda, MD: American Fisheries Society. and large-scale conservation: what do we know for Mitsch, W.J., and Gosselink, J.G. 2000. Wetlands. New sure? Ecography 22:225–232. York: John Wiley & Sons, Inc. Hastings, A., and Botsford, L.W. 2003. Comparing Mumby, P.J., Edwards, A.J., Arias-Gonzalez, J.E., Lin- designs of marine reserves for fisheries and for biodi- deman, K.C., Blackwell, P.G., Gall, A., Gorczynska, versity. Ecological Applications 13:S65–S70. M.I., Harborne, A.R., Pescod, C.L., Renken, H., Wab- Helfand, G.S. 1981. Fish behavior by day, night, and nitz, C.C.C., and Llewellyn, G. 2004. Mangroves twilight. In The behavior of teleost fishes, ed. T.J. enhance the biomass of coral reef fish communities in Pitcher, pp. 479–512. London: Chapman and Hall. the Carribbean. Nature 427:533–536. Hettler, W.F. 1989. Nekton use of regularly flooded salt- Neigel, J.E. 2003. Species-area relationships and marine marsh cordgrass habitat in North Carolina, USA. Mar. conservation. Ecological Applications 13:S138–S145. Ecol. Prog. Ser. 56:111–118. Olsen, E.M., and Vollestad, L.A. 2001. An evaluation of Hobson, E.S. 1974. Feeding relationships of teleostean visible implant elastomer for marking age-0 brown fishes on coral reefs in Kona, Hawaii. Fish. Bull. trout. N. Am. J. Fish. Manage. 21:967–970. (U.S.) 72:915–1031. Ovaskainen, O. 2004. Habitat-specific movement para- Hubbs, C.L., and Lagler, K.F. 1967. Fishes of the Great meters estimated using mark-recapture data and a dif- Lakes region. Cranbrook Institute of Science, Bloom- fusion model. Ecology 85:242–257. field Hills, MI. Bulletin No. 26. Peterson, G.W., and Turner, R.E. 1994. The value of salt Jacobus, J., and Ivan, L.N. 2005. Evaluating the effects marsh edge vs interior as a habitat for fish and deca- pod Crustaceans in a Louisiana tidal marsh. Estuaries of habitat patchiness on fish assemblages in a Great 17:235–262. Lakes coastal marsh. J. Great Lakes Res. 31:466–481. Potts, J.A. 1970. The schooling ethology of Lutianus Jennings, M.J., Bozek, M.A., Hatzenbeler, G.R., monostigma (Pisces) in the shallow reef environment Emmons, E.E., and Staggs, M.D. 1999. Cumulative of Aldabra. J. Zool. Lond. 161:223–235. effects of incremental shoreline habitat modification Rahel, F.J. 1984. Factors structuring fish assemblages on fish assemblages in north temperate lakes. N. Am. along a bog lake successional gradient. Ecology J. Fish. Manage. 19:18–27. 65:1276–1289. Killgore, K.J., Morgan, III, R.P., and Rybicki, N.B. Randall, R.G., Minns, C.K., Cairns, V.W., and Moore, 1989. Distribution and abundance of fishes associated J.E. 1996. The relationship between an index of fish with submersed aquatic plants in the Potomac River. production and submerged macrophytes and other N. Am. J. Fish. Manage. 9:101–111. habitat features at three littoral areas in the Great Kunin, W.E. 1998. Biodiversity at the edge: a test of the Lakes. Can. J. Fish. Aquat. Sci. 53:35–44. importance of spatial “mass effects” in the Rotham- Ricketts, T.H. 2001. The matrix matters: effective isola- sted Park Grass experiments. In Proceedings of the tion in fragmented landscapes. Am. Nat. 158:87–99. National Academy of Sciences of the United States of Rosenblatt, D.L., Heske, E.J., Nelson, S.L., Barber, America 95, pp. 207–212. D.M., Miller, M.A., and Macallister, B. 1999. Forest Lewis, C.A., Lester, N.P., Bradshaw, A.D., Fitzgibbon, fragments in east-central Illinois: islands or habitat J.E., Fuller, K., Hakanson, L., and Richards, C. 1996. patches for mammals? Am. Mid. Nat. 141:115–123. Considerations of scale in habitat conservation and Rosenzweig, M.L. 1995. Species diversity in space and restoration. Can. J. Fish. Aquat. Sci. 53:440–445. time. Cambridge, UK: Cambridge University Press. MacArthur, R.H. and Wilson, E.O. 1967. The Theory of Rozas, L.P., and Odum, W.E. 1987. Use of tidal fresh- Island Biogeography. Princeton, NJ: Princeton Uni- water marshes by fishes and macrofaunal crustaceans versity Press. along a marsh stream-order gradient. Estuaries Magnuson, J.J., Tonn, W.M., Banerjee, A., Toivonen, J., 10:36–43. Sanchez, O., and Rask, M. 1998. Isolation vs. extinc- Savino, J.F., and Stein, R.A. 1982. Predator-prey interaction in the assembly of fishes in small northern lakes. tion between largemouth bass and bluegills as influ- Ecology 79:2941–2956. enced by simulated, submerged vegetation. Trans. Malone, J.C., Forrester, G.E., and Steele, M.A. 1999. Am. Fish. Soc. 111:255–266. Effects of subcutaneous microtags on the growth, sur- Sibert, J.R., and Nielsen, J.L. 2001. Electronic Tagging vival, and vulnerability to predation of small reef and Tracking in Marine Fisheries. Dordrecht: Kluwer fishes. J. Exp. Mar. Biol. Ecol. 237:243–253. Academic Publishers. Fragmentation: Small Marsh Fishes in Patchy Habitats 211

Smith, K.D. 1995. Vegetation-open water interface and the species composition and richness of fish assem- the predator-prey interaction between largemouth blages in northern Wisconsin lakes. Ecology bass and bluegills: A field experiment. Michigan 63:1149–1166. Department of Natural Resources, Ann Arbor, MI. Wang, L., Lyons, J., Rasmussen, P., Seelbach, P., Simon, Fisheries Research Report Number 2000. T., Wiley, M., Kanehl, P., Baker, E., Niemela, S., and Snodgrass, J.W., Bryan, J.A.L., Lide, R.F., and Smith, Stewart, P.M. 2003. Watershed, reach, and riparian G.M. 1996. Factors affecting the occurrence and influences on stream fish assemblages in the Northern structure of fish assemblages in isolated wetlands of Lakes and Forest Ecoregion, U.S.A. Can. J. Fish. the upper coastal plain, U.S.A. Can. J. Fish. Aquat. Aquat. Sci. 60:491–505. Sci. 53:443–454. Weaver, M.J., Magnuson, J.J., and Clayton, M.K. 1997. Strayer, D.L., Power, M.E., Fagan, W.F., Pickett, S.T.A., Distribution of littoral fishes in structurally complex and Belnap, J. 2003. A classification of ecological macrophytes. Can. J. Fish. Aquat. Sci. 54:2277–2289. boundaries. Bioscience 53:723–729. Werner, E.E., Gilliam, J.F., Hall, D.J., and Mittelbach, Svenning, J.-C. 1998. The effect of land-use on the local G.G. 1983. An experimental test of the effects of pre- distribution of palm species in an Andean rain forest dation risk on habitat use in fish. Ecology 64: in northwest Ecuador. Biodiv. Conser. 7:1529–1537. 1540–1548. Ter Braak, C.J.F. 1986. Canonical correspondence analy- Wiley, M.J., Gorden, R.W., Waite, S.W., and Powless, sis: a new eigenvector technique for multivariate T. 1984. The relationship between aquatic macro- direct gradient analysis. Ecology 67:1167–1179. phytes and sport fish production in Illinois ponds: a Tjørve, E. 2002. Habitat size and number in multi-habitat simple model. N. Am. J. Fish. Manage. 4:111–119. landscapes: a model approach based on species-area With, K.A., Cadaret, S.J., and Davies, C. 1999. Move- curves. Ecography 25:17–24. ment responses to patch structure in experimental Toepfer, C., and Fleeger, J.W. 1995. Effects of marsh- fractal landscapes. Ecology 80:1340–1353. edge habitat variables on feeding success by juvenile bay whiff, Citharichthys spilopterus (Teleostei: Both- Submitted: 11 March 2004 idae). Southwestern Naturalist 40:297–300. Accepted: 18 July 2005 Tonn, W.M., and Magnuson, J.J. 1982. Patterns in Editorial handling: John Janssen