Estuaries Vol. 24, NO;. 2, p. 285-298 Apti2001

Fish Assemblage Structure in Relation to Environmental

Variation in a Texas Gulf Coastal Wetland

FRANCES P. GELWICK*, SENOL AKIN. D. ~RF.Y ARRINGTON, and KIRK O. WINEMmLER

Department of Wildlife and Fisheries S~ Texas A&M University. 210 lfagi4 Hall, College

Station. Texas 77843-2258ABSTRACI': We described seasooal &h--~1ages in an estuarine marsh fringing Matagorda Bay. Gulf of Mexico.HabItat zones were identified by patterns of fish species ab1JDdaDce ~d indicator &pedes opdma along graments insaJinlty. dissolved oxygen (DO). and depth in our samples. Indicators of the lower brackish zone (lower Jake and tidalbayou d~ to the bay) were gulf !!!~den (~pGb'OnIU), bay anchovy ~ mitdrilIa), snver perch (BGirdi6ll4dl7:ys0llrG). and spotted seatrout (~ fteb,IIosUI) at salinity > 15roo. DO 7-10 mg I-I. and depth < 0.5 In. Indicatorsof the upper brackish zone (Jake and fringing salt marsh) were pinSsh (Lapdon momboida) and spot (l.8iostorIa&I-thurw) at saJiDity 10-20roo. DO > 10 mg I-I. and depth < 0.5 m. In the freshwater wedand zone (diked wedand,ephemeral pool, and pereDDial scour pool). indicators were sheepshead minnow (CypriI1odOrI ~), rainwater 1dDifish(LIIcaftiG par-). mosquito&sh (GGmhrma affinis), and saDfiD mony (PHc:ilia latipiIIna) at salinity < 5180. DO < 5 mg I-I.and depth ~ I m. In the freshwater dlaDDeHzed zone (slough-and irrigation ~). indiQtors were three sunfish species(LepomIs). white aappie (PorIIOIiI annulariI), and gizzard shad' (Doror- tep~) at Alinity < 5180. DO < 5 mg I-I.and depth > 1.5 m. In bnddsh zones, seasooal variation in species diversity among sites was positively correJated withtemperature, but assemblage 8tnIcture also was inftuenced by depth and DO. In the freshwater zones, seasooal ~tionin species diversity among sites was posi.tiveIy correlated ,nth depth, DO. and salinity. but assemblage structure wasweakly associated ,nth temperature. Species diversity and assemblage strucfure were strongly affected by the connectivitybetween freshwater wedand and brackjsh zones. Uncommon species in diked wedands, such as tarpon (Mega1op$ allan-Iiau) and fat sleeper (DonnitlJtDr ma!,,-!J~I.). indiQted movement of fish5 &om the brackish zone as the water levelrose during natunl floodiDg and scheduled (July) releases &om the diked wedand. From September to July. diversity inthe freshWater wedand zone de~ as recediDg waters left smaD isolated pooJs, and fish movement ~e blockedby a water-control stnicture. Subsequendy. diversity was reduced to a few speci5. ,nth opportunistic life histories and

tolerance to anoxic conditioDS that developed as flooded vegetation decayed.~

IntroductionOrganisms in coastal wetlands are affected by

spatial and temporal dynamics of salinity withinthese complex, terrestrial-freshwater-marine eco-tones (Bulger et al. 1993). Physiological toleranceto salinity and other physiochemical parameters,and life-history strategies determine how anima1srespond to habitat heterogeneity ~d dynamics.FISh communities of coastal ~t marshes generallyrepresent a mixture of ecological groupings thatcontain euryhalinespecies adapted to brackish en-vironments, plUs others more typical of either ma-rine or freshwater systems (Rozas and Hackney1983; Odum 1988; PetersOn and Meador 1994;Wagner and Austin 1999). Renfro (1959) conclud-ed that survival of freshwater fishes in saline habi-tats of the Aran~ River, Texas was generally short-term. He suggested that low abundance of fresh-water fishes was likely due to stress assoCiated withfrequent salinity fluctuations, and that freshwaterfishes either moved to less saline waters or died. In

other systems (Perry 1967; Peterson and Ross1991; Wagner and Austin 1999). freshwater fishcan withstand short-term exposure to ~ties >10%0. but occur most often at < 5CYoo. Freshwaterpredators such'as largemouth bass (Micropterus salrmoides) and longnose gar (Lepisosteus osseus) entertidal creeks to forage (Meador and Kelso 1989).However. different species and life-stages of fresh-water and estuarine fishes vary ~eatly in their sa-linity tolerancecS (Chipman 1959; Renfro 1959;Griffith 1974; Hollander and Avault 1975; Meadorand Kelso 1990a; Susanto and Peterson 1996).

Beyond iildividual" response to physiochemicalvariation. populations also respond through bioticinteractions that influence community strUctureacross these complex ecotones (Dunson and Travis1991; Wmemiller and Leslie 1992; Wagner andAustin 199:9). Habitat structure and the interactionof fishes and other aquatic fauna with terrestrialpredators. also influences composition and bi?"mass of aquatic biota (Vmce et ale 19,?6; .Knelb1987). Availability of aquatic prey for waterfowl,~g birds. and other animals depends o~ tem-poral change in habitat quality and quantlty for-.Corresponding author: tete: 979/845-6777; fax: 979/845-

4096; e-man: rgelwickOramu.edu.

285C 2001 Estuarine Reseald1 Federalian

MW 86"05'

Fig. 1. Mad Island Marsh Preserve. Site$ EI-E6 were brack.ish estUarine habitats. and sites FE and Fl-F3 were freshwaterhabitats. An irrigation canal diverted water from the ColoradoRiver, 20 bn east of Mad Island Marsh Preserve. Dotted linesindicate roads. The area between FE and El was dry in Marchand May, and Fl. was dry from March through July. A water-control structure at the road culvert betWeen Fl and FE wasdosed rrom Septemb.er tojuly.

evaluated across seasonal and spatial-gradients ofdepth. water quality (dissolved oxygen. salinity. andtemperature). and connectivity (diked and opensystems). In addition. habitat zones were identifiedbased on fish-assemblage composition. and char-acterized by indicator species and their optimaacross measured physiochemical factors. This studyalso provided baseline data and locations for mon-itoring and evaluating environmental conditionsand biotic integrity of MIMP.

Methods- STUDy AREA AND SAMPLING PROTOCOL

The MIMP is located on the northwestern coastof the Gulf of Mexico, in Matagorda Bay (Fig. 1).Aquatic habitats in MIMP span physiochemical gra-dients, from freshwater Mad Island (MI) Sloughand diked wetlands, to oligo-mesohaline MI Lakeand salt marsh, and a polyhaline tidal bayou. Be-fore initial sampling, heavy rainfall caused floodingin October 1997 that over-topped diked roads andwedands throughout MI:MP, connecting all aquatichabitats. Wmter samples were made at four siteswhen they were accessible in December 1997, andother sites were added in March 1998 as access was

286 F. P. GeIwIck et II. . - .. .

'.

aquatic organisms (Kah11964; Kushlan 1976; WIl-lard 1977; Browder 1978). Submerged aquati~v~-etation influences prey availability and predation ..risk for fish and aquatic invertebrates (Rozas andOdum 1987a.b, 1988; Meador and Kelso 1990b).Also, schedules for accumulation of detritus in Saltmarshes and tidal creeks influences communitystructure, productivity, and transfer of energy be-tWeen freshwater and marine systems (Odum1984). Because of the natural Spatial and temporaldynamics in water level and habitat conditions incoastal srs:tems, the role of connectivity amonghabitats also should be understood. Diked wet-lands (often used for restoration) provide deep-water refugia for fishes (Toth et aI. 1998). Howev-er, when these reduce spatial connectivity amongaquatic habitats, dikes can isolate aquatic organ-isms, influence water quality, and affect communitydynamics Uohnson et aI. 1997). For example, de-spite identical water-level dynamics, diked wetlandsin l'1orida did not support a fish fauna as rich asthat of restored wetlands that were coD;nected tonatural open-water habitat (Toth et aI. 1998).

The use of terrestrial and freshwater resourcesby a growing h~an population has impacted theecological integrity and health of Texas coastal wet-lands (McKinney 1997). Reduction of freshwaterflow to estuaries has obvious effects on saliriity anda host of related abiotic environmental factors. TheNature Conservancy of Texas manages the Mad I&oland Marsh Preserve (MIMP) on the edge of Mat-agorda Bay (Gulf of Mexico). Management goalsinclude augmenting freshwater inflows and en-hancement of coastal salt-marsh and prairie. eco-systems through erosion control, prescribed burn-ing, mowing, grazing, and impoundment. Thoughfocused on benefits to terrestrial wildlife, wadingbirds. and migratory water fowl. management ac-tions also are relevant to habitat use and produc-tion of fishes. many of which are prey for birds.Connectivity among estuarine habitats in MIMPand management of freshwater inflows (e.g., dikesfor roads and rice fields. water-control structures.and irrigation canals) likely influence spatial dy-namics of physiochemical factors and habitat useby fishes.

Because fishes respond to both terrestrial andaquatic environmental factors, fish assemblagecomposition can indicate biological integrity ofaquatic systems (Karr 1991). In order to plan mon-itoring and management to restore natural ecosys-tems and their functions (e.g., flood control, nu-trient dynamics, food production for terrestrialwildlife) in MIMP, relationships between habitat at-tributes and assemblage structure of fishes werequantified. Assemblage structure (e.g.. speciescomposition, diversity. life-history strategies) was

287fish AuembIages of a Coastal ~eUand

tered in three consecutive hauls. Higher densitY- ofvegetation (Ruppia, spaTtinG) gen"erally red~~~ddle distance covered per seine haul. and total :di&-tance seined was reduced in very small habitatS.Across all seasons and sites.. total distance seine9 ata site ranged from 5 to 105 m. Captured fish wereanesthetized in MS-222. fixed in a 10% formalinsolution "in the field..th~n so~-and counted inthe la,boratory. Data wez:~ standardized to catch perunit of effort (CPUE. It:a1cu1ated as number permeter of distance seined) and log10 transformedfor analyses. .

,DATA ANALms

Shannon diversity (H') based on natural loga-rithm of (::PUE was calculated for each sample andtested for correlation with e;ach environmental var-iable (marginal effecr.,) and partial correlations(conditional, or independent additive effecr.,) ofeach environmental variable. Variables were testedfor fit to normal distributions using the univariateprocedure of SAS to test the Kolomogorov D sta-tistic (SAS 1996). Normally distributed variableswere tested using Pearson's correlation coefficient,and otherS were tested using Spearman's Rank cor-relation coefficient (SAS 1996).

Canonical Correspondence Analysis (CCA) , amultivariate method of direct gradient analysis, wasrun on species CPUE across all samples using CAN-aca (ter Braak and Smilauer 1998). To compare'the importance of the gradients~-tely in fresh-water and brackish zones identified by the initialCCA and cluster analyses (see below), additionalanalyses were run for MI Lake and salt marsh (E1-E6), and MI Slough, irrigation channel, and fresh-water wedand (FE, F1-S). To focus analyses on re-sponses by common species to environmental gra-dienr." species were restricted to those with totalCPUE ~ 1 % of total CPUE across all species andsamples. Each canonical axis was derived from aniterative reciprocal weighted averaging of speciesCPUE among samples, with the additional con-straint that ordination scores for species and siteswere linear combinations of environmental vari-ables. This met;l1od of ordination is particularlyuseful for spe~es that show a unimodal (humped)relationship to environmental gradients and zerovalues in many samples (ter Braak and Smilauer1998). Hill's scaling of scores with a focus on SJ?e-cies was chosen so that distances between speaeswere generalized Mahalanobis distances, and dis-tances betw~en sites represented species tUrnov.eracross samples (beta diversity) in standard dewa-tions (SD; ter Braak and Smila1,ler 1998). M°n.teCarlo simulations (199 iterations), tested the. S1~-nificance (p oS 0.05) of the variatic;>n in fish distn-bution among samples that was represented along

established. Spring, summer, and fall samples_~~!emade in May, July, and October 1998, respectively.Sites were located in the dominant aquatic habitatsof~ (Fig. 1). MI Bayou (E6) was a tidal sys~mcoimecting brackish MI Lake to the Intrac~~talWaterway. FIVe sites were located along MJ Lakeand fringing salt marsh (Fig. I, EI-ES). Amongthese, only E2 was sampled in December 1997, butit and other sites in MI Lake and Bayou were sam-pled in the four other months, representing thefour. seasons. An irrigation canal from the Colo-rado River (F2) and MI Slough (F3) representedchannelized open freshwater systems, but only FSwas connected to MI Lake (Fig. 1) while the Col-orado River entered the intracoastal waterwayabout 20 kIn east ofMIMP. Sampling at F2 beganin March, but the May sample was omitted by mis-take. Sampling at FS was possible in May, July, andOctober. A diked freshwater wedand (Fl) retainedwater from October 1997 through June 1998. InJuly 1998, scheduled releases of water drained Flas riser boards blocking a road culvert were re:.moved, and water flowed into the upper saltmarsh. This reduced the size ofFl from about 100mt to a 12-mt pool adjacent to the culverL In Sep-tember, scheduled closure of the culvert allowedrainfall and runoff to refill the wetland. A smallephemeral pool (Fl-a, ca. 6 mt; Fig. 1) was sam-pled when it retained overflow from F1 in Decem-ber 1997, and again in October 1998 after it haddried and refilled. A perennial scour pool (FE, ca.15 mt) had formed at the outflow of the culvertfrom the diked wetland. During March. and May(culvert closed), FE was sampled but the down-stream reach was dry, and thus, it was disconnectedfrom both the upper salt marsh and the diked wet-land. During July' and August (culvert open), flowreconnected FE to both F1 and E1.

For each sample, water depth was measured(0.01 m), and temperature (OC), dissolved oxygen(DO, mg 1-1), and salinity (roo) were measuredwith a Hydrolab DatasondeS or a Yellow SpringsInstruments YSI Model 85 during daytime (0900-1600). Measurements were intended to representamong.,site and within-site variation at the time ofsampling, rather than diet extremes actually ex-perienced by fishes. Fishes were collected using acombinatiotr of active and passive gear to ~-mize number of species captured. Because seiningconsistently captured more species, as well as in-dividuals, than other methods, those data were se-lected for statistical analyses of assemblage struc-ture. Seines were 6-m long, 1.2-m tall, with 4.5-mmm~sh, and 1.2-m by 1.2-m bag. Brans were keptabout 3 m apart during seine hauls parallel toshore, and distance seined was recorded. Seiningcontinued until no new fish species were encoun-

fish ~es of a Coastal WeUand 289

TABLE 1; EXtended..

0.04'.

0.19 27..10.01 r1.19

25.51 0.050.010.41

20.67 0.120.010.76

0.080.020.150.16 0.07 0.30 0.08 0.07

0.210.65

0.6-'1.18 0.10

0.07o.os

0.02 0.08" 0.089O.0S

O.OS

2.0~0.15

0.020.03

0.03 0.050.250.04 0.03

0.180.02 0.10

0..050.860.84OM

0.050.02

0.050.050.01

0.05 0.050.2S

0.071.530.87

0.100.23 0.01 0.01 0.01

0.02O.ISO.OS

0.022.230.23

0.17 5.150.85

0.18 0.48 0.03 6.90 0.20 2.29 0.20 0.100.670.07 0.01

0.010.05 0.05

0.050.25

0.050.070.100.07

50

0.01 0.011-"" 0.55 0.450.02

0.050.49 0.05

0.070.05

0.010.02

45 SO0.01

750.03

30 400.01

10575 50 50 60 75 60 90 75 so 60. Dry area from El to FE.

vec~rs at acute angles to CA 1; Fig. S). Vectorspoint toward increasing values for an environmen-tal variable. with a longer vector indicating greaterrange of variation for a variable along the canoni-cal axes. Vectors point toward increasing val1:les foran environmental variable with a longer vector in-dicating greater range of variation in species dis-tributions for the variable. Thus. important vari-ables tend to be represented by. longer arrows. Themore acute the angle between the vector 3I1;d thecanonical axis, the greater the correlation betweenthem. Distribution of sites along this complex pri-mary gradient represented a complete turnC?ver (5SD) of species, ranging from those having optimain deeper water of the irrigation canal and MISlough (F2 and F3). and at lower DO and salinity.(Fig. S left side o:f CA .1)~ to those species withoptima at intermediate values for variables indiked wetland and scour pool (F1 and FE, Fig. 3middle of CA. 1), and Mally to those species in MILake with optima in shallower water and at higherDO and salinity. (Fig. S right side of CA 1).

Correlation of temperature to CA. 2, indicatedseasonal change in assemblage structure. which ex-plained 4.5% of species distribution. This repre-sented a less-extensi'¥e turnover of species (S SD)than that along the primaiily spatial gradient of CA.1 (Fig. S,bottom to top ofea~ panel). In Decem-ber and March 1998 (Fig. B lower center) after the

0.05; Table S and Fig. 2). Across samples only inMI Lake and salt-marsh, diversity was correlated totemperature (r = 0.865, P < 0.01). Among thesebrackish sites, the only significant partial correla-tion of diversity with any environmental variablewas at E2 for salinity (r = 0.99, P = 0.01 with De:cember data included; r = 0.99, P = 0.04 with De-cember omitted). At El and ES, partial correlationof salinity to diversity was at a lower significancelevel than at E2 (for each r = 0.99, P = 0.07). Thisreflected higher salinities in May and July (TableS) when salt-tolerant killifishes (Cyprinodontidae)entered uplake sites (Table I, EI-ES). AnJong sam-ples in MI Slough and the freshwat;er wedand, di-vers.ity was correlated to depth (r = 0.768, P <0.01), DO (r = 0.255, P = 0.05), and salinity (r -0.422, p - 0.01). At FE and Fl (each having datafor ~ five sample dates}, the only significant re-lationship with diversity was for partial correlationof DO (r - 0.96, P - 0.04) atFl, where lowest DOvalues were measured in July and October (TableS), when assemblages were reduc;ed to a few spe-cies tolerant of anoxia (Table 2).

Results for CCA with and without December(P9st flood) samples were similar, so Decembersamples are included for completeness. The stron-gest gradient (CA 1; eigenvalue - 0.565, P - 0.01)explained 15.1 % of species .variation related to sa-linity, DO, and depth (indica.ted by environmental

0.23'S.!O0.120.05

290 F. P. Gelwick et 8l

TABLE 2. FIShes in families and species that were cOllected by seining in Mad Island Slough from December 1997 to October 1998.Values are number of individuaJs m~1 of shoreline seined. Group membership was determined by clust.er analysis of sites using thesevalues and those for fishes from Mad Island Lake (Table 1).

F~ WctIaad H&bica& Zooc S

nJ'amny Spedea c- DecO Ma.- juJo Oc\"

LepisoSteidae AITtJdosWs $/JalidaQupeidae Dorwoma up.ai4mumC)'prinodontidae Adinia unica

CyJninodonVGri8gatusFu1ldulw pndisFundulus JIUZwrauLucania PaT-

Poedliidae Gambwia affinisPOIdlia latipimIa

Atherinidae Mmidia hlryUinaCentrarchidae I..J-is ~

I..J-is~I..J-is11lrlmltricwPomoxis -nularis

Mugilidae Mugil apllalluTotal dL1tance seined (m)

8.001.50

18.508.000.501.50

16.50

5.50 ~

13.0069.0096.5020.0013.50

0.50 25.00

11.50

alligator gargizzardahaddiamond kiJ1ifiJhsheepshead minnowguifkiJ1ifiJhbayou killifishrainwater killifishmosquitoftshIanfin moDyinland ~deblu~longear sunfishbantam sunfishwhite aappieatriped mullet

0.50

.5 5 5 5 5C culvert betWeen FE and Fl closed.. culvert open~. dry area from El to FE.

October 1997 flood, shallow upper reaches of MILake and salt marsh (EI-E3) "as well as freshwaterMI Slough, scour pool, and diked wetland (F3, FE,FI, FI-a, re$pectively) were nearly optimal fQr kil-lifish (Cyprinodontidae)~ In March, sites closer tothe intracoastal canal (E4-E6, Fig. S lower right)were nearly optimal for Gulf menhaden, croaker(Sciaenidae), and naked goby (Gohiosoma OO$c). In

May and July (Fig. 3 upper left), the irrigation ca-nal and MI Slough (F2 and F3, respectively) werenearly optimal for sunfishes (Lepomis) and whitecrappie (Pomoxis annularis), while shallower fresh-water sites (F1 and FE) were nearly optimal forlivebearers (Poeciliidae) and inland silverside(Menidia beryllina). The warmest temperatures andhighest DO values for our Sanlples in MI Lake oc-curred during May in the shallow salt marsh (Table3, E1-E3), and were nearly optimal for silver perch(Bairdiella chrysoura), red drum (Sciaenops ocellalus),and pinfish (IAgodqn rhomboides) (Fig. 3 upperright). In October, temperature and DO were lowin shallower freshwater sites (FE, F1, F1-a, Fig. 3left; Table 3) nearly optimal for killifish and live-bearers, while sites in MI Lake were nearly optimalfor inland silverside, spotted seatrout (C.)'noscion ne-bulosus), bay anchovy, and red drum (Fig. 3 right).

Overall, measur~d envirQnmental variables ex-plained 21;8% of spatial and temporal variation inassemblage composition (canonical eigenvalues =0.818; total eigenvalues = 3.747). Partial CCAshowed variation was correlated with additive ef-fects of salinity (5.0%, P = 0.01), temperature(4.2%, p = 0.02), and DO (3.7%, P = 0.04), aswell as interactions among environmental variables(7.6%), but not significantly with depth (1.4%, P= 0.75). Ranges for total species variation in sep-arate CCAoffreshwater (eigenvalues = 1.985) andbrackish samples (eigenvalues = 2.241) were small-er than for all samples combined but a higher per-centage of species variation was correlated with en-

Ii 2.5,- M8d I.IMd Slough 8nd Wetlands

2.0"

ccoc. .FE F1 F1.. F2 F3

EJ Dee D II. D !lay a JIA . Oct

Fig. 2. Shannon diversity indices for &hes seine samplesfrom December 1997 to October 1998 in brackj.,h estUarinehabitats (E) and freshwater habitats (F) of Mad Island ManhPreserve. Si~ codes follow Fig. I, and Tables 1 and 2. Lower-case letters indicate months when no sample was collected (n),when the watel'Conb"Olstructure was closed (c) or open (0),and when Fl-a was dry (d). Asterisks indicate when the areafrom El to FE was dry.

F8h A8861.~ fA . CoutaI Wel8r.t 291

TABLE2.I'.xten4cd.

vironmental variables in the separate anaJysis forbrackish sites. For brackish MI Lake (EI-E6). ca-nonical axes explained !2.5% (p - 0.01) of speciesvariation. which was primarily correlated with ad-ditive effects of depth (11.5%. p - 0.01) and DO(9.1 %. p = 0.02). but not significantly relatcd totemperature (5.5%. p = 0.80) or salinity (!.6%. P= 0.!9). and with the remaining correlated withinteractions (2.8%). For freshwater scour pool.diked wetland, irrigation canal. and slough (FE.F1-~) canonical axes explained 24.1 % of speciesvariation but significance was low for temperature(9.7%. p = 0.07). salinity (6.6%. P - O.!!). DO(4%. P = 0.64). and d~pth (!.4%. P = 0.81).

As in the CCA, clUSter analyses were mmi1:4r withand without December samples. and results using.December samples are presented. Four habitatzones were identified along th~ spatial gradientfrom intracoastal canal and lower MI Lake (Zone1. E5 and E6). to upper MI Lake (Zone 2. EI-E4).to freshwater wetlands (Zone !. FE. Fl. and Fl-a).and to MI Slough and the irrigation canal (Zone4. F2 and FS). The superscript on a species labelin the CCA plot (Fig. !) corresponds to the habitatzone indicated by a species. Probability of occur-rence plotted along gradients of measured envi-ronmental variables sh°'Ved significant optima forindicator species of the lower brackish zone (Fig.! right side. Zone 1). These spedes were small-bodied, ~d found along shallow « 0.5 m) shore-lines. or across wide ranges of salinity and DO inour samples but w.ere most likely to occur at salinity> 15~ and DO 7-10 mg 1-1 (Fig. 4). Two indi-cators of the upper bractL,h zone (Fig. S. Zon.e 2)were estuarine dependent, pinfish and spot (LIios-

tomw ~). and were most likely to occur atsalinity IG-20L. DO > 10 mg 1-1. and depth <0.5 m (Fig. 4). Indicators of the freshwater wetlandzone (Fig. 3. Zone 3) included mosq'uitofish. whichwere most likely to occur at lower salinity « 5%0)and DO « 5 mg 1-1). and three euryhaline spe-cies. bayou killifish (Fundulus pUlverftU). sheeps.-head minnow (~nodon variega~). and sai1finmolly (Poea1ia latipinna). which were most likely tooccur at DO < 5 mg 1-1 and depth> 1 m (Fig: 4).In the freshwater channelized zone (Fig. 3. Zone4). five indicator spedes. including three sunfishspecies (Lepomis macrochi1Us, L. megaiotis. L. .sym~trims). white crappie. and gizzard shad (Dorosomaap.dianum). occurred across.a wide range of depthand DO. but were most likely to occur at sa1inity< 5%0. DO < 5 mg 1-1. and depth> 1.5 m (Fig.4). .

n;-MI~ODA" expected. salinity had the strongest associa-

tion with fish assemblages along a spatial gradientfrom mesohaline MI Lake and fringing salt marshto oligohaline habitats of the freshwater dike:d wet-land and MI Slough. A study of Chesapeake andDelaware Bays wing multivariate analysis of pres-ence/absence data for fish and invertebrates de:-termined biologicany.based estuarine salinity zonesuseful for long-term monito~g of estuarine eco-system health (Bulger et a1. .1993).:0ur;analyseswere' based on seine CPUE across seasonal sam-ples; and identified indicator .pecies that had max-imal occurrence (in our samples) within three sa-linityzones «5%0. 10-20%0. and > 20%0). Thesecorresponded to zones 0-4%0. 11-18%0. and 16-

292 F. P. Gelwick et aI.

TABLE S. Values of environmental ftriables measured for fishamples at sites in Mad Island Marsh Preserve from December1997 to October 1998. .

:JN

.1

~TrcfSol(.)

DOt-. 1-')a,-./S- Moam 3,6 2,6

361 20 FE 1:A 0-0

~ 30 186 DEPtH---

10

FE0

Brackish Habitats

.1'.6 Mar

MayJul .Oct

E5 Mar

MayJulOct

1'.4 Mar. May

JutOct

E3 Mar

MayJulOct

E2 DecMar

MayJulOct

1'.1 Mar4

Ma".sJulOct

Freshwater HabitatsFE Deco

~

Mayc*Jul°Oct"

11 DecoMar"'

MaY"Jul°Oct<

Flea DecOct

12 May

JulOct

13 Mar

JutOct

20.0029.0532.4028.5019.7431.5128.0028.5019.1034.6729.3027.8018.8833.26SO.10SO.SO13.5018.8832.1731.5030.6016.4856.0857.20SO.30

12.015.529.412.812.012.512.012.012.012.115.97.65.6

15.515.19.70.55.5

14.918.29.85.7

15.021.1.4.4

7.175.704.956.506.557.525.674.977.019.664.516.897.10

12.855.029.555.707.529.865.~9.657.85

12.752.858.67

o~0.500.410.740.500.400.200.600.270.400.210.560.290.400.210.540.350.250.500.170.480.500.500.140.45

N

c1

cCI).

10

f?

EOctFOct

.2.0 SD CA 1

E Dec+ E Mar. E May. EJul8FDec. FMarO FMayO FJulO

WIMJC\

eLmegfe Lme",

eLsym'e Pann"

e DCep4

Gaff' Aspe4.. . PIsI' DEP'T1i--15.50

15.0028.0928.7027.2015.5015.SO28.1528.7027.!O14.0027.5028.85SO.1028.4015.0051.7029.40

0.5

0.5

0.0

0.4

0.2

0.5

0.0

0.9

0.4

0.2

0.0

0.6

0.0

0.5

0.5

0.0

0.5

0.2

4.505.204.827.720.204..505.505.410.502.404.502.521.071.504.544.002.985.15

1.001.00 cot0.40 ~0.401.000..501..500.150.251.500..50G..502.002.000.50 -2.0 SD CA 1 +3.0 SD

1.000.45 Fig. 3. Plot of sample acofes (symbols in upper panel) and1.00 spedes centroids (circles in lower panel). on the first tWO (a.:A.VV ivnical) axes (Ca'..}., c...o:.} ~vl' meeo:-:-_yond8Dce anal)'S!s of

fishes in seine samples. from brackish estuarine habitats (E) andfreshwater habitats (F) of Mad Island Marsh Preserve in Decem-ber 1997 (E = +, F - .) and March (E = 8, F - 0), May (E= +, F = 0), July (E - 8, F = 0), and OctOber (E - A, F= 6,) of 1998. Arrows indicate direction of blaeasing value forenvironmental wriables. More acute angles betWeen arrow andaxis bldicate stronger correlation of variables with the axis.Numbers in upper panel correspond to site codes in Fig. 1. Inthe lower panel, centtoids indicate approximate centers of spe-des' distributions relative to environmental 'Variables. Speciescodes are the first letter of the genus and first three letten ofthe specific name. Supersaipts identify the habitat zone (Tables1 and 2) that was best indicated by a species' occurrence andabWldance, and centtoids indicate significant (filled circle) and

nonsignificant (shaded circle) indicator spedes.

0(I).In..,

~ cu1ven betWeen FE and n dosed.. culvert o~n.. dry area from El to FE.

27%0 identified for Chesapeake and Delaware Bays(Bulger et al. 1993). As previously suggested (Bul-ger et al. 1993), gradients for variables other thansalinity further differentiated our assemblages. Oc-currence of species along combined gradients ofDO, depth, and salinity delineated four habitatzones. Salinity and DO explained the greatest per-centage of variation in overall species distributionacross these zones, and is consistent \Ofith findingsfrom studies of other North American coastal salt

10

2~,:FE 0 1"1!

+3.050A1:.

.~FISh Assemblages oIa CoutaI WeUand

~.

'ZJ

range of temporal dynamia characterized assem-blages across this physiochemical ecotone. Overall. '

asse~b1ages were most simn~r when saltwedge in-trusion was minim~1 (December 1997 and March1998, following flood conditions of October 1~7),and were least similar when it was maximal Uu!r1998). Overall dive;rsity was positively correlatedwith temperature and dePth; both~-'Which tendedto be positively correlatep With seasonal changes insalinity in the upper brackish zone .and with DOand salinity in the freshwater zones. In the lowerbrackish zone, sa1inity was less variable and diver-sity declined with seasonal dominance by gulfmen-haden and bay anchovy. In freshwater zones, di-versity was positively correlated with depth, DO,and salinity. In a study of a similar coastal saltmarsh on the Gulf coast of Mississippi, Petersonand Ross (1991) found that tidal freshwater andoligohaline habitats had greater fish diversity thana mesohaline site located on the edge of.Biloxi Bay.They concluded that species richness at low-salinitycoastal sites was enhanced by the presence of fresh-water species, as well as species which tolerate lowsalinity, and euryhaline marine species with broadsalinity tolerance. .

Cit;ing findings by Hackney et al. (1976), -Peter-son and Ross (1991) also suggested that habitatsexperiencing frequent and large fluctuations in sa-linity also nlight support £ewer .fish.species..For ex-ample, if sites are predoIi11nan-ny"fresnwater, thendownstre~ mesohaline sites would have lowerspecies richness if obligate freshwater species areexcluded, whereas euryhaline marine specieswould not be excluded from upstream sites. Alter-natively, if si.tes are predominantly mesohaline,then lowest species richness would be associatedwith upstream sites if they experience grea.t envi-ronmental fluctuations. In our study, additive andsynergistic effects of salinity, depth, and DO onspecies richness and diversity similarly dependedon site location and dynamics of environmentalvariation across the landscape. Variation among as-semblages in the freshwater zones was related tospecies tolerances or physiological optima alonggradients of DO and depth; among these sitesmore species and familieS occurred at the most in-land freshwater site (MI Slough), which was con-tinuo~ connected to MI Slough. The lake con-tained a range of primary.division freshWater fishes(Ostariophysians) including Cyprinidae, Catast°D;l-idae, and Ictaluridae {unpublished data thISstudy), as well as marine species (striped mullet)that penetrate Gulf coast rivers hundreds of kilo-meters inland (Riggs 1957). The lake'~ upperbrackish zone contained euryhaline marme spe-cies, pinfish, sciaenids, striped mullet, and south-ern flounder (Paralichthys leth~), as well as

3

iOS~=.a

iO'

+0.0 Deplil (m) +2.0

Fig. 4;. Pr9babUity Of occurrence for indicator species thathad significant relationships to measured gradients for salinity(upper panel). DO (middle panel). and depth (lower panel) inlamples of fishes at Mad ~d Manh Preserve. Species arethose that indicated habitat zones identified by duster analysisof fishes. Labels indicate location Of maximum probability ofoccurrence for each species along gradients. Codes for specieslabels are those in Fig. S.

marshes (Keup and Bayless 1964; Parrish and Yer-ger 1974; Weinstein et al. 1980jPetei'Son and Ross1991).

Although transition between freshwater and ma-rine faunas is bound to occur. the temporal andspatial scale of the transition will vary dependingon geography. topography. and composition of re- .gional faunas. Within the relatively small spatialscale of the MIMP (ca. 6 km from F3 to E6). aconsiderable turnover from freshwater to marinefaunas was observed across habitat zones. The onlyspecies in seine samples occurring at each end ofthe salinity gradient were gizzard shad. sheepsheadminnow. bayou killifish. and striped mullet. A

294 F. P. Gelwlck et aI.

freshwater species that tolerate low salinity, alliga-tor gar, an.d gizzard shad. No species in MI Lakeand its fringing salt marsh had maximal occur-rence at salinity .5-10%0. Indicator species of theupper brackish zone (where salinity fluctuatedacross this range).had optima> 10%0. Our resultsare $imUar to those for summer fiSh assemb~esacross longer reaches (up to 150 kin) of lowerChesapeake Bay tributaries (Wagner 1999), wherea species-richness minimum occurred at 8-10%0,across a range of 0-181foo. In contrast, Petersonand Ross (1991) sampling seasonally at three sitesacross a longer distance (17 kIn) but along a sim-ilar salinity gradient to that in MIMP, r~rtednosignificant relationship between ~ombined fresh-water-estuarine/marine species richness and salin-ity at each site. They related this to compensatoryshifts in assemblage structure from primarily fresh-water to increasing estuarine/marine species rich-ness with increasing salinity.

Aquatic macrophytes provide cover and a sourceof invertebrate prey for salt marsh fishes (Rozasand Odum 1987b, 1988). ~ergent shoreline mac-rophytes (e.g., SpaTtina spp., Typha latifolia) andsubmergent forms (e.g., Ruppia maritima, CJIaTasp.) occurred in MIMP, although dominant plantspecies varied over the topographic/salinity gradi-ent. In freshwater tidal salt marshes in VIrginia,fish abundance was greater at low-order sites, ap-parently associated with greater macrophyte den-sity (Rozas and Odum 1987a). Temporal as well asspatial patterns determine how a given habitat fea-ture influences biota. For example, greater ampli-tude and duration of water-1evel fluctuation in amarsh should lead to extended periods of emer-gent macrophyte production, followed by periodsof hypoxia associated with plant decay during re-flooding. IIi tropical-river floodplains, this cycle isdriven by seasonal flood pulses, followed by fishemigration and mortality (Howard-Williams andJunk 1976; Wmemiller 1996). In shallow MI Lake,DO was sufficiently high during most of our sam-pling to support even those species most sensitiveto hypoxia. In contrast, low DO « -' mg 1-1) wasrecorded for several samples in the freshwaterchannelized zone and in the October 1998 samplein the freshwater wetland zone, as macrophyte de-cay proceeded. Dominating those samples weresmall-bodied fishes (killifishes, liverbearers) thatefficiently use aquatic-surface respiration as com-pensation for hypoxic conditions (Lewis 1970). Im-poundment of freshwater runoff in the diked wet-land when the culvert was closed (September toJuly) likely increased the amplitude and durationof normal flood conditions, and consequently de-creased habitat quality for organisms sensitive toaquatic hypoxia. This also was illustrated in July,

after the culvert was open to drain the diked wet-land, and more fishes were present in the SC9urpool (DO = 7.7 mg 1-1) than in the resid~.')V.et-land (DO = 0.5 mg 1-1).

Despite the fact that shallow coastal salt marshesserve as important nursery habitats fot numerousmarine-estuarine dep~ndent &hes and inverte-brates, high predation .~orta1i~ occur whenwater level declines. ne abundant small fishes ofcoastal salt marshes ptovide a food resource foralligators, wading birds, and piscivorous fishes,such as gar and red drum. Periodic concentrationsof sman fishes, such as those we encountered inthe ephemeral wetland pool, can be an importantresource for nesting piscivorous birds (Kahl1964;Kushlan 1976; Willard 1977; Browder 1978). Such .conditions apparently favored small-bodied cypri-nodontiform fishes with an. opportunistic life ~tory strategy that allows them to quickly increasepopulation sizes and recolonize disturbed habitats(Wmemiller and Rose 1992). Killifishes and live-bearers (Poeciliidae) dominated all freshwatersites, except the irrigation' canal. In our study. 7killifish and livebearer species were collected in theephemeral pool in December 1997 (after the Oc.:tober 1997 flood), and 4 of these were again col-lected in October 1998, despite the fact that it haddried and refilled in the interim. Even in deeperhabitats of the channelized freshwater zone, thesesmall fishes were abundant in shallow marginswhere they could probably consume zooplanktonand ~a1ler macroinvertebrates. May and July sam-ples in MI Slough contained 6 and 8 species re-spectively, but only 5 in October. Five species pre-sent in the earlier samples, most of them relativelylarger-bodied (alligator gar, gizzard shad, andthree sunfishes), were absent in the latter sampleswhen water was shallower. In the irrigation canal,Bayou killifish, another small-bodied euryhalinespecies that moves inland with rising floodwaters.was sampled only in October.

Movement of fishes in the lower, brackish lakezone shifted assemblage composition, but thischange was uncorrelated with salinity. Lar.8'eschools of marine fishes, such as juvenile gulf men-haden, seasonally migrate into and out of shallowestuarine habitats, and likely influenced correla-tion of measured environmental variables to assem-blage composition and diversity. Although not asnumerous other marine species moved in and outof MI Lake on variable time scales. For example.large piscivorous red drum and spotted seatroutentered MI Lake from the bay in pursuit of abun-dant shrimp, crabs, mullet, and small fishes (Akinunpublished data). Densities of red drum werehighest in October when water temperature wasmoderate and prey were near their peak densities

Ash AS88ntIIages of a Coastal WeUand 295

after the summer growing season (Akin unpub-lished data, Dumesnll personal communication).

Synergism of environmental variability and hab-itat connectivity undoubtedly influences spatiotem-poral assemblage patterns (Gelwick et al. 1997;Johnson et al. 1997; Toth et al. 1998). Reductionof aquatic ~abitats that periodically concentratesfishes, can increase their competition for resources(Lowe-McConnell 1987; Wmemiller 1989) and fa-cilitate predation by wading birds (Kushlan 1976;Browder 1978). During low-water periods in thefreshwater wetland zone. rooting by feral hogs (Susscrofa) disturbed the damp soil and rem~;n;ng veg-etation. and wading birds were common. The pe-rennial scour pool below the water-control culvertretained fishes that might otherwjse have movedtoward the upper brackish zone as water level re-ceeded across the fringing salt marsh. For-exam-ple. a juvenile tarpon (Megalops allantialS) collect-ed in the diked wetland in October, and fat sleep-ers (Donnitator maculatus) collected in the scourpool and diked wetland in July. indicated use ofthe freshwater zones by fishes from brackish ~ones.However, even when water level would otherwisehave allowed movement of fishes between thesezones, the nearly year-round (September to July)closure of the culvert restricted fish movement.During years without pervasive flooding to connecthabitats. water level fluctuations at MIMP were in-fluenced by management primarily for productionof rice and waterfowl forage (Le.. draining. burn-ing. irrigation. and mowing). Alternative scenariosfor the management of surface hydrology and veg-etation-growth at MIMP and other coastal marsh~scould be explored to increase connectivity a]pongwetlands. and synchronize water-level manipula-tions to natural flood regimes in favor of enhancedaquatic. as well as terrestrial, biodiversity.

ACKNOWLEDG~

We thankJim Bergan and Mark DumesnD for their assistancewith locating sampling lites and allowing us to use the MadIsland Manh Preserve headquarters and visitors center duringour study. We also thank Dan Roelke for the use of a hydroJab,and for helpful ~ons on the limnology of coastal ecosys-ten1:5- Selahatdn Aydin, \\1ksel Bolek, Archis Grubh, Todd Lantz,Hemin L6pez-Femindez. Mike R.o~rtson, Soner Tarim. andJeremy Walther assisted with fish samples. This study was fundedunder the Applied Coastal Conservation Research Program ofThe Nature Conservancy of Texas.

LrrERATUREClTED

BR.OWDER,J. A. 1978. A modeling stUdy of water, wedands, andwood stOrks, p. 325-!46.lnA. c. Sprunt IV,J. c. Ogden, andS. Wmckler (eds.), Wading Birds. Research Report 7. NationalAudubon Society, New York.

BULGER, A. J., B. P. HAmEN, M. E. MONACO, D. M. N~N, ANDM. G. McCo~CK..RAY. 1995. Biologicany.based estuarine sa-linity zones derived from a multivariate analysis. Estumia 18:311-322.

QmNAN, R. K. 1959. StUdies of tol~ceof certain freshwaterfishes to hrine water from on wells. EeD1Dg, 40-.299-!02.

DUFKiNE. M. AND P.l.WENDu. 1997. Species assemblages andindicator species: The need for a flexible asymmetrical ap-proach. ECtJlDgjt:tll MonogrD/1/u 67:Ms-S66. -.

DUNSON, W. A. AND J. TRAVIS. 1991. The role of abiotic faCtorsin community organization. Amm"ean Natu1Ulist 188:1067-1091. .

GELWJa, F. P., M. S. Sroa, AND .W. J. ~!HEWS- 1997. Effectsof fish, water depth, and p~n risk ~ patch dynamics ina north-temperate river ecO17S"'tem. 000s 80:882-398.

GmmH. R. W. 1974. Environ1nent and salinity toierance in thegenus Fundulus. Copaa 1974:319-351.

HAcKNEY, c. T., W. D. BURBANa, AND O. P. HACKNEY. 1976. Bi-ol~cal and physical dynamics of a Georgia tidal creek. ChIS-

~ Sdma.17:271-280.HOUANDu. Eo Eo ANDJ. W. AVAULT,JR. 1975. Effects ofsaJinity

on survival of buffalo fish eggs through yearlings. PnIgrusiwFish a&lhnist 37:47-51.

HoWARD-~, c. AND W.J.JUNL 1976. The decomposidonof aquatic maaophytes in the floating meadows of a Centr31Amazonian WrZea lake. ~ica 7:115-123.

JOHNSON, D. L, W. E. L~QJ, JR., AND T. W. MORJUSON. 1997.FISh communities in a diked Lake Erie wetland and an adja-cent undiked area. WItlandl17:4S-54.

KAHI. M. P. 1964. Food ecology of the wood stork (M]a.riG-iea1ID) in F1orida. Eeological MD1U>gTrJ/1/u 34:97-117.

KARR,J. R. 1991. Biological integrity: A long-neglected aspect ofwater resource management. E&Ological AppliaJtiDn.S 1:66-84.

K!UP, L AND J. BA~. 1964. FISh distribution at 'Var}'iDg salin-ities in Neu"e River basin, North Carolina. ChGaP1DJr4 Sdmu5:199-12.5.

KNEm. R. T. 1987. Predation risk and u"e of intertidal habiratsby young fishes and shrimp. EaJ/I)g, 68:379-386.

K1IS1n.AN,J. A. 1976. Wading bird predation in aseasooany fluc-. tuating pond. Aui 93:464-476. .

I.EWls, JR., W. M. 1970. Morphological adaptations of cyprino-dontoids for inhabiting oxygen deficient waten. CoJ1eia 1970:519-526.

LoWE-McCoNNEu., R. H. 1987. Ecological StUdies in TropicalFISh Communities. Cambridge University PreS$, Cambridge,

England.McCtc1Nr;. B. AND M. J. MErroRD. 1997. PCORD. Multivariate

Analysis of Ecological Data, Version 5.0.lYljM Software Design,Gleneden Beach, Oregon.

McKINNEY, L 1997. Troubled Waters Part n. Texas Parks andWildlife, February:47, Austin, Texas.

MEADOR, M. Eo AND W. E. ~. 1989. Behavior and movementsof largemouth bass in response to salinity. 1T~ oj theAmm'canFuAIriG S«i4tJ 118:409-415. .

MEADOR, M. Eo AND W. E. KELso. 199Oa. Growth of largemouthbass in low-salinity environmenb. Thmuzctioru of the AmnicanFisI..riIS S«i4tJ 119:545-552.

MEADOR, M. E. AND W. Eo ~. 1990b. PhJlSiolQgical responsesof largemouth bass, Microptlrus sabMid£S, eXpoIed to salinity.Canadian Journal of FuJI.riu and Aquatic Scimea 47:2S58-2!63.

OOUM, W. Eo 1984. Dua].gradient concept of detritUS transportand processing in estUaries. B~ oj Marini Scima 35:510-521. .

OOUM, W. Eo 1988. Comparative ecology of tidal freshwater andsalt marshes. Anmlal ~ oj EcJJlDg, and SJstImatics 19:147-176.

PARRISH, P. R. AND R. W. YERGER. 1974. Ochlockonee river fishes:SaIinit)'-temperature effects. Florida Scimtisl 56:179-186.

PERB.Y, W. G. 1967. Distribution and relative abundance of bluecatfish, 1ctal1n'W juTcatu.s and channel catfish, 1ctal1n'W ptmc-talus. with relation to salinity. PnIc.Iing of tAl A.nml4l Confmnuof the SDII.I1II4st#7I AssDdation ofGQ.".. andFash Com1nimonm21:436-444.

~i~~":::;.~..,"!t" ..':.~.:';'~~~ "

,~~ k~,~~::-"-'

296 F. P. Gelwlck et aI.

~~~~~

-- ,- - - -. ---

WAGNER. c. M. 1999. Expression of the estuarine species mini-mum m Uaoral fish assemblagea of the lower Ch~~ Baytnoutaries. Estv4riG 22:504-512. - -- -

WAGNER. C. M. AND H. M. AumN. 1999. Correspondence be-tWeen environmental pients and summer littoral fish u-semblages in low Ia1iDity reaches of the Chesapeake Bay. USA.Marim EI:t1lDgJ P1ogr&fS s.riG 177:197-212.

~. M. P.. S. 1.; WEISS, AND M.F. WALTERS. 1980. MUltiplede~ants of commwlity structure in. shaUow marsh habi-tats, Cape Fear River Eltuasy. North"&roUDa. USA. MGri1I4BiDlOD 58:227-24!. !

VINCE, s.. L VALIELA. N. BAai1s.ANDJ. M. TEAL 1976. Predationby the salfl'marsh ki1lifisb F1mdulw Att#oditus (1-) in relationto prey Size and habitat structure; con.sequences for prey dis-tribution and abundance. Jw1'7I4l oJE1L1I'ri-ntGl MminI BiQl()D~ Ecoiog, 25:255-266.

~, D. E. 1977. The feeding ecology and behavior of fivespecies of herons in southeastern New Jersey. Condor 79:462-470.

WJN!:!6n~~ K. O. 1989. Ontogenetic diet shifts and resourcepartitioning among pisdVorous fishes in me Venezuelan Da-nos. E,,~ BiolDD oj FISlIM 26:177-199.

WINW-~~ K. O. 1996. Dynamic diversity: FISh communitiesof a-opical riven. p. 99-154. In M. 1.; Cody and J. A. sman-wood (eds.), Long-Term Studies ofVertcbrate Communities.A~demic Preu, Orlando. Florida.

WINL~~~ K. O. AND M. A. I.DLIL 1992. FISh aS1embiagesaaoas a complex, tropical freshwater/marine ecotone. EmIi-ronmml4l BiDlDD of Fis1&cs 54:29-50.

WJNEMn.LD. K. o. AND K. A. ROSE. 1992. Patterns oflife-bistorydiversification in North American fishes: ImpUtations for pop-ulation regulation. CanadianJoumal ofFisllma and Aquatie sa-mas 49:2196-2218.

SOURCE OF UNPUBLISHED MATERIAlSD~ M. Personal Communication. The NatUre Conser-

vancy of Texas. P. O. Box 165, CoUegepon. Texas 77428-0165.

RafJId. for coruidcr'Gtion. Slptm«r 7, 1999Acapt«lforJlu1Jlicstilm. N~ 17,2000

PETUSON, M. S. AND M. R. MEADOR. 1994. Effects of salinity onfreshwater fishes in coastal plain drainages in the lOuth~-ern U.s. R.uiIrD$ in Fr.slIm4f Sd6IuG 2.-95-121.

PEI'DSON, M. s. AND S. T. ~ 1991. ~a of littoral fishesand decapods along a coastal rl\'er.Cstu2rine gradienL ~n,.,. CMstaZ aM SlI4l/ Scima M:467-485. -

REi-aR.o. w. c. 1959. SUr.viva1 and migration of freshwater fishesin salt water. T-. Journal of Sdma 11:172-180. .

"R.Iccs, c. D. 1957. Mup tePlaaZtu in Oklahoma and northernTexas. CopM 1957:158-159.

RoZAS. L P..AND Co T. HACKNEY. 1985. The bnportante of on-gobaline estUarine wedand habitats to fisheries ~urces. W.l4fI4s 5:77-89.

Rp'ZAS, L P. AND W. EoODUM. 1987-. Use of tidal freshwatermanhes by fishes and maaofaunal a"i~C~~S along a marshstream-order gradienL EsIuarics 10:56-4-'.

RoZAS, L P. AND W. Eo anUM. 1987b. Fash and maaoaustaceanuse of submerged plant beds in tidal &eshwater marsh aeeks.'MariJW EtOlog, PrrJgrISS s.ri4S 58:101-108.

RoZAS. L P. AND W. Eo anUM. 1988. The role of submergedaquatic vegetation in influencing the abundance of nektonon contiguous tidal fresh-water marshes. Journal of ~tal Marim BiDIOD aM ECD1l1f1114:289-S00.

SAS INsmvn INc. 1996. SAS/STAT User'. Guide, Version 6,4th edition, Volume 2. Carr, North Carolina, USA.

SMnAuu, P. 1992. CanoDraw. Miaocomputer Power. Ithaca.New York. . .

SUSANro, G. N. AND M. s. PETERSON. 1996. Survival, osmoregu-lation and oxygen consumption of my coastal largemouthbass, Miaoptm£S S4l8IOida ~epede) exposed to saline me-dia. HJIlTOIIiD!ogi4 52S:119-127.

TER. BR-'Ax, c. J. F. AND P. s.InAUD.. 1998. Canoco ReferenceManual and User's Guide to Canoco for Wmdows: Softwarefor Canonical Community Ordination (version 4). Microcom-puter Power, Ithaca, New York.

TOTH, L A, S. L MELVIN, D. A ABRJNCTON, AND J. QW&Bn-LAIN. 1998. Hydrologic manipulations of the channelized Ki5osimmee 1Uver: Implications for restoration. BioSdma48:751-164.