Trtmsoct;O/U of th~ Am~rican Fisheri~.. Soc;~1y 126:1012-1020. 1997C Copyrigbl by !be American Fisheries Society 1997

Sampling Fishes in Vegetated Habitats: Effects of HabitatStructure on Sampling Characteristics of the I-m2 Throw Trap

FRANK JORDAN1

Departments of Biology and Marine Science. Jacksonville University2800 University Boulevard North. Jacksonville. Florida 32211. USA,1

WC.,SEAN COYNE AND JOEL C. TREXLER

Deportment of Biological Sciences, Florida International UniversityUniversity Park Campus. Miami, Florida 33199, USA

Abstract.-Enclosure traps that quickly surround well-defined areas of habitat are perhaps themost widely used method for sampling fishes in vegetated habitats. However, relatively few dataare available to evaluate the effects of habitat structure on sampling characteristics of enclosuretraps. In this study, we determined how clearing efficiency and accuracy of l-m2 throw trapsvaried across a range of environmental conditions in the Florida Everglades by sampling withinenclosed areas of marsh habitat. Throw trap clearing efficiency and sampling accuracy did notdiffer among two widely separated locations and appeared to be unaffected by variation in waterdepth, canopy height, plant cover, plant stem density, and periphyton volume. Sampling accuracyaveraged 63% of fishes present after correcting for clearing efficiencies. On average, 83% of thefishes present in a throw trap were recovered. Therefore, it appeared that about 17% of the missingfishes may have burrowed into the substrate or been discarded with sorted detritus. In contrast,the remaining 20% of fishes probably avoided the throw trap. This is the first study to differentiatebetween potential sources of throw trap sampling errors. Importantly, density estimates obtainedby throw traps were positively correlated (r = 0.82) with actual population densities. Mean fishlengths and fish size distributions obtained by throw trapping usually did not differ from actualmean lengths or fish size distributions. Finally, high concordance of fish species ranks indicatedthat throw traps accurately described fish community structure. Throw traps appeared to providerelatively accurate estimates of fish density, fish size, and community structure across a range ofenvironmental conditions.

It is generally accepted that aquatic vegetationplays an important role in structuring aquatic com-munities, probably by providing prey species withrefuge from predators and increased foraging op-portunities (Downing 1991; Heck and Crowder1991). However, considerable gaps remain in ourunderstanding of the importance of aquatic veg-etation, including questions such as How muchvegetation is beneficial for aquatic systems? (Du-rocher et al. 1984; Hoyer et al. 1985; Hoyer andCanfield 1996a, 1996b; Maceina 1996); Do dif-ferent plant species and plant growth forms differin their importance to fishes? (Dionne and Folt1991; Chick and McIvor 1994); Does the impor-tance of aquatic vegetation change appreciablywith changing hydrologic conditions? (Loftus andEklund 1994). Advances in our understanding ofthe importance of aquatic vegetation have beenslow to emerge largely because of the difficulties

associated with accurately sampling fishes in veg-etated habitats.

Traditional sampling techniques such as seine-netting, electrofishing, and underwater observa-tion do not provide reliable estimates of fish den-sities in vegetated habitats (e.g., Freeman et at.1984; Dewey et at. 1989). In contrast. enclosuretraps that quickly surround a well-defined area ofhabitat have proven quite useful for sampling fish-es (Rozas and Minello 1997). Two general classesof enclosure traps are presently in use. The firstclass of enclosure traps includes drop traps (e.g.,Hellier 1958; Kushlan 1974, 1981; Kjelson et at.1975; Gilmore et at. 1978), pull-up nets (e.g., Hig-er and Kolipinski 1967; Kushlan 1974), and buoy-ant pop nets (e.g., Larson et ai. 1986; Serafy et at.1988; Connolly 1994). Use of these enclosure trapsusually requires erection of a sampling platformor alteration of habitat before use. Although thesedevices can provide quantitative data, habitat mod-ification may profoundly affect fish behavior andproduce spurious results (e.g., Loftus and Eklund1994; Rozas and Minello 1997; also see Petersonand Black 1994).

~

I Present address: Department of Biological Sciences,

Loyola University New Orleans, 6363 St. Charles Av-enue, New Orleans, Louisiana 70118, USA.

1012

FISH SAMPLING IN VEGETATED HABITATS 1013

The second class of enclosure traps includesthrow traps of varying sizes and construction (We-gener et at. 1974; Chick et at. 1992; Rozas andMinello 1997). Throw traps are portable, do notrequire habitat modification before sampling, andhave been used to sample fishes and other organ-isms in shallow habitats such as subtidal fresh-water creeks (Rozas and Odum 1987), vegetatedfreshwater marshes (Kushlan 1981; Loftus and Ek-lund 1994), dense littoral vegetation (Miller et at.1991; Chick and McIvor 1994), estuarine seagrasses (Sogard et at. 1987; Sogard and Able1991), and montane ponds (Pfennig et at. 1991).The versatility of the throw trap provides a richdatabase for comparison of fish assemblagesacross a range of habitat types and aquatic systems.However, little effort has been made to determinethe effects of variable habitat structure on the sam-pling characteristics of throw traps; that is, are datafrom different habitats and systems comparable?

The comparative sampling efficacy of differenttypes of throw traps has been well studied (e.g.,Kushlan 1981; Chick et at. 1992). Additional re-search has measured clearing efficiency of throwtraps (i.e., the proportion of fishes removed fromwithin a throw trap; also referred to as recoveryefficiency; Rozas and Minello 1997). For example,Freeman et at. (1984) and Rozas and Odum (1987)recovered between 90-100% of the marked fishesthat they released into throw traps. However, ofmuch greater concern to aquatic ecologists is howclosely estimates of fish densities derived fromthrow trap samples match the actual populationdensities of fishes in the habitats being sampled(i.e., accuracy; also referred to as catch efficiency;Rozas and Minello 1997). Kushlan (1981) foundthat throw traps were biased and only capturedabout 73% of the fishes present in emergent wetprairies. Jacobsen and Kushlan (1987) later hy-pothesized that evasion or flushing of fishes aroundthe edges of a falling trap may reduce (~8l %) theeffective sampling area of throw traps. Kushlan(1981) did not assess clearing efficiency, whichcould have affected his estimates of fish density(e.g., Freeman et at. 1984). Finally, Kushlan(1981) held conditions ''as constant as possible"in his studies. Therefore, how clearing efficiencyand accuracy of throw traps are affected by vari-ation in habitat structure remains unclear.

In the present study, we assessed the effects ofhabitat structure on clearing efficiency and accu-racy of I-m2 throw traps at two wet prairie sitesin the Florida Everglades. This research comple-ments earlier research on the sampling efficacy of

throw traps (e.g., Kushlan 1981; Freeman et al.1984; Jacobsen and Kushlan 1987) by differenti-ating between clearing efficiency and sampling ac-curacy and by sampling across a wide range ofenvironmental conditions (water depth, periphytonvolume, plant stem density, plant cover, and can-opy height).

Methods

Study area.-Research was carried out in emer-gent wet prairies of Water Conservation Area 3Ain the Florida Everglades during September 1996when water temperatures ranged between 20°C and25°C. We focused on wet prairies because thesehabitats are readily used by most Everglades fishes(Loftus and Kushlan 1987; Jordan 1996a) and be-cause most long-term studies of marsh fish ecologyhave been performed in this type of habitat (e.g.,Kushlan 1976; Loftus and Eklund 1994; Jordan1996a, 1996b). Sampling characteristics of throwtraps were examined at two wet prairie sites thatrepresented a wide range of plant stem densitiesand environmental conditions (Table 1). More im-portant, we feel that the results obtained from wetprairies are valid for other types of vegetated hab-itats amenable to throw trap sampling (e.g., seagrass meadows, littoral zones).

Throw trap methodology.- The 1-m2 throw trapused in this study was identical to that describedby Kushlan (1981). Copper pipe was used to con-struct a rectangular frame, and 1.5-rom-mesh net-ting was used to cover the sides of the frame. Tocollect fishes, the trap was thrown into positionand then quickly pressed into the substrate. Waterdepth and canopy height were measured to thenearest centimeter, and then all emergent plantswithin the throw trap were identified and enumer-ated. Floating vegetation (periphyton, Bacopa,Utricularia) present in the throw trap was removedand measured volumetrically. After habitat struc-ture was measured, a combination of bar seines(1.5-mm mesh) and dip nets (0.5- and 1.5-rommesh) were used to remove fishes from within theemergent vegetation. The bar seine was used firstto remove the bulk of fishes, and then dip netswere swept through the throw trap until ten con-secutive empty sweeps were obtained. Additionalinformation on construction and use of throw trapscan be found in Kushlan (1974, 1981), Freemanet at. (1984), Chick et at. (1992), and Rozas andMinello (1997).

Field evaluation and analyses.-Clearing effi-ciency and sampling accuracy were determined bycomparing the numbers of fishes collected in throw

1014 JORDAN ET AL.

TABLE I.-Environmental characteristics and relative abundance (%) of plant species within eight block-netted studyareas at two Everglades wet prairie sites used in this study. Asterisks denote habitat features that differed between thetwo sites (Mann-Whitney test, P < 0.05).

66,3006776020

668.833

914038

6S

6.33371

.so27

Water depth (cm)*Peripbyton (mUm2)*Stem8/m2Plant coverage (%)Canopy height (cm)*Relative plant abundance (%)

Cladiwn jamakenseEleocharis cellJ4lo.roEleocharis elongataEriocauloll compresswnHymenocallis sp.JlUticia 0)IafQNymp/raea 0d01Ylt4ZNympitoides aquaticaPanicwn /IemUonJOnPaspa/adiwn gemil!QlumRhynchospora tracyi*Sagittaria loncifolia

0t

"970I0001000

077

00.0

80

19

I0.

000020

SIS

20S

160

6.18

0000

100

2700

0

042

002000

180

354

0SO0020r0

..,G

a.0

07500

100001t10

0.~I,II009260

traps with those inside four enclosed areas of wetprairie habitat at each of the two sampling sites.Block nets (25 X I m, 1.5-mm mesh) were usedto enclose 25-m2 areas of marsh. Block nets withina site were about 150 m apart, whereas the twosites were about 12 km apart. To prevent the escapeof fishes, the heavily weighted bottoms of theblock nets were pressed into the substrate, and thetops were supported by float lines. We allowedfishes to acclimate for 1 h and then simultaneouslydeployed three throw traps into the blocked-offarea. Fishes were cleared from the first throw trapas described above, identified, and measured to thenearest millimeter for standard length (SL). A sub-set of fishes collected from the first throw trap weremarked by fin-clipping and then released into thesecond throw trap. Similarly, a subset of the fishesremoved from the second throw trap were fin-clipped and released into the third throw trap. Allmarked and nonmarked fishes were released backinto the block net after the three throw traps wereemptied.

After all three throw traps were emptied andremoved, we removed as much vegetation as pos-sible from within the block-net areas to minimizeentrapment of fishes, which would have biased ourestimates of fish population density. We then ap-plied a fish toxicant (rotenone) to the area withinthe block net while vigorously agitating the watercolumn to ensure adequate mixing and dispersal.Dead and dying fishes were collected periodicallyduring the next 36 h to obtain estimates of fish

population density. Fishes recovered from enclo-sures were identified and checked for fin clips, andtheir SL was measured to the nearest millimeter.

Three measures of throw trap sampling accuracywere calculated to facilitate comparison with ear-lier studies. First, we determined accuracy by di-viding the meaD density of fishes in throw trapsby the density of fishes in block nets (accuracy 1;compare KushI.an 1981). Second, we adjusted pop-ulation density estimates by dividing the meandensity of fishes in block nets by mean recapturerates of marked fishes in block nets (i.e., block-net clearing efficiency). We pooled data for allspecies of marked fishes to calculate mean recap-ture rates because of similarities in size and be-havior and because of unequal sample sizes amongblock nets. The mean density of fishes in throwtraps was divided by this adjusted population den-sity estimate to obtain a second measure of sam-pling accuracy (accuracy 2; compare p~ and Ro-senberg 1982). Third, we adjusted throw trap den-sity estimates by dividing the mean density of fish-es in throw traps by mean recapture rates ofmarked fishes in throw traps (i.e., throw trap clear-ing efficiency). Adjusted throw trap density esti-mates were then divided by adjusted populationdensity estimates to obtain a third measure of sam-pling accuracy (accuracy 3). To minimize the num-ber of estimates in our calculations, we used meanrecapture rates to calculate measures of samplingaccuracy.

Throw trap clearing efficiency, block-net clearing

FISH SAMPLING IN VIlGBTATBD HABITATS 1015

TABU! 2.-Fish densities (individuals/m2) and species richness (number of species collected for each med1od) fordtrow traps and block nets and three measures of dtrow trap sampling ~uracy. Asterisks denote sampling characteristicsthat differed between the two wet prairie sites (Mann-Whitney test, P < 0.05).

90903846.11677.1.1769.19

8671364356736449S96

10

831002631293890698267

1006715182735.S643oW6

11

(A) Throw trap clearing efficiency (%)(8) Block-net clearing efficiency (%)(C) Meandlrow b'ap fish density.(0) Estimated dIrow b'ap fish density (ClA)*(E) Block-net fish density.(F) Estimated block-net fish density (E/B).Accuracy 1 (C/E)Accuracy 2 (C/F)Accuracy 3 (D/F)Throw trap species richnessBlock-net species richness

Therefore, we used the Kolmogorov-Smirnov(KS) test to compare size distributions obtainedfor pairs of throw trap and block-net collections(N = 8). Finally, Kendall's test was used to mea-sure concordance of species ranks between throwtrap and block-net collections (N = 8). Statisticalprocedures generally follow the methods of Sokaland Rohlf (1995).

efficiency, and measures of accuracy were corre-lated with mean habitat structure data (water depth,

periphyton volume, stem density, plant cover, andcanopy height) by using Pearson's correlation co-efficient to test for independence of sampling andenvironmental characteristics. Given our samplesize (N = 8) and an alpha level of 0.05, correlation

coefficients (r) would have to be larger than 0.795to reject the null hypothesis of no relationship be-tween sampling and environmental characteristics.

Finally, correlation coefficients for each measureof accuracy are equivalent since we used mean

recapture rates (i.e., constants) to produce theseestimates.

A central goal of this study was to determinewhether the fish assemblage sampled by throwtraps represented the actual fish assemblage pres-ent in our study area (Loftus and Kushlan 1987;Jordan 1996a; E Jordan and J. C. Trexler, unpub-lished data). Accordingly, we examined how sizestructure and species composition differed be-tween throw traps and block nets. First, we useda two-way analysis of variance (ANOV A) to testfor the effects of collection method (throw trapversus block net) and sampling location (locationof the eight block nets) on the mean SL of fishes.Each block-net location was considered indepen-dent for this and other analyses because prelimi-nary ANOV As indicated that SL did not differconsistently between our two study sites. Standardlengths obtained with each collection method werethen compared via least-squares (LS) means foreach sampling location because of a significantinteraction between collection method and sam-pling location. The relative importance of sparselydistributed, large fishes to overall size structuremay be overestimated by comparing mean lengths.

ResultsThe two wet prairie sites differed with respect

to water depth, periphyton volume, and canopyheight but were similar with respect to overallplant species composition and plant stem densities(Table 1). Sites were difficult to distinguish fromone another because there was considerable vari-ation among block nets within each wet prairiesite. At our first wet prairie site, for example, plantstem densities ranged from 18 to 677/m2, and therelative abundance of floating water lily Nymphaeaodorata ranged from 0% to 51 %. Overall, it ap-pears that our eight block-net samples varied con-siderably with respect to habitat structure.

Both throw trap and block-net estimates indi-cated that fish densities differed between wet prai-rie sites, whereas estimates of fish species richnessdid not differ between wet prairie sites (Table 2).More importantly, no evidence indicated thatthrow trap clearing efficiency, block-net clearingefficiency. or throw trap accuracy varied betweenwet prairie sites (Table 2). No significant corre-lations were found between habitat structural fea-tures and sampling characteristics (Table 3) at theP < 0.05 level. This lack of significant correlationsis striking considering that we did not control theprobability of making a type I error by adjustingour experimentwise error rate for the 25 multiple

1016 JORDAN ET AL.

TABLE 3.-Pearson correlations between sampling characteristics (Table 2) and environmental characteristics (TableI). Replicate throw traps were averaged for this analysis, and N = 8 for all correlations. None of these correlationswere significant at the P < 0.05 level.

varied considerably among sampling locations (F= 21.3; df = 7, 8,871; P = 0.0001). There wasan interaction between collection method and sam-pling location (F = 3.8; df = 7, 8,871; P =0.0004), which indicated that collection methodsshould be evaluated separately for each samplinglocation (Figure 1). Fishes collected by throw trapswere, on average, 1.4 mm shorter than fishes col-lected in block net 7 (LS means, t = 2.4, df =771, P = 0.0158). In contrast, fishes collected bythrow traps were 2.7 mm longer than fishes col-lected in block net 5 (LS means, t = -4.0, df =807, P = 0.0001). The mean lengths of fishes col-lected with throw traps and block nets did not dif-fer at the other six sampling locations.

Comparison of size distributions revealed sim-ilar patterns to those of mean length data. Thecumulative distribution of fish lengths collected bythrow traps was indistinguishable from the cu-mulative distribution of fish lengths collected byblock nets (KS test, X2 = 5.6, P = 0.1241). Sizedistributions of fishes collected by throw traps dif-fered from size distributions collected by block

comparisons in Table 2 (e.g., the Bonferroni meth-od; Sokal and Rohlf 1995).

Accuracy 1 (82%) estimated how closely throwtrap and block-net estimates matched, assuming100% clearing efficiency of fishes from withinblock nets (after Kushlan 1981). However, recap-ture rates were low for block nets (77%). Accuracy2 (63 % ) included an estimate of population densitythat had been corrected for low clearing efficiencyof fishes from block nets (after Pihl and Rosenberg1982). Therefore, throw trap density estimateswere about 37% lower than actual population den-sities. Approximately 17% of this difference ap-pears to result from low clearing efficiency (83%)of fishes from within throw traps. Including thissource of sampling error, it appears that throwtraps captured about 80% (i.e., 63% + 17%) ofthe marsh fishes present. Density estimates derivedfrom throw trap and block-net data were positivelycorrelated (r = 0.82, P = 0.0130).

Overall, the mean length of fishes did not differbetween throw traps and block nets (F = 0.2; df= 1,8,871; P = 0.6643). In contrast, mean length

- 21EE 18-:5 15C)C.! 12

"C... 9CO

"CC 6CO

-

(I) 3.c(I)

ii: 01 2 3 4 5 8 7 8

SampleFIGURB I.-Mean (+SE) standard length (mm) of fishes collected with throw traps (open bars) and block nets

(solid bars) in eight sampling areas. Asterisks denote means that are significantly different between collectionmethods (least-squares means, P < 0.05).

FISH SAMPLING IN VEGETATED HABITATS 1017

0.790.900.820.840.880.910.940.80

0.00040.00010.00030.00010.00010.00010.00010.0002

a34,678

0.000.00

0.02O.al

0.000.66

0.080,56

EMeactmthus gloriosu.r 0.00 0.01ErimyWn suceff4 0.00 0.00FWldldu.r chrys- 8.34 6.02Gombu.ri4 holbrooki 32.55 42.56HeterandriofonlODsa 37.82 30.16JordQllello jtorlIlae 4.23 4.00Labidest1les ricculu.r 0.00 O.~Lepomi.r spp. 1.25 1.32[.,..,a,.la gaadei 14.12 14.74NotunloS gyrilwoS 0.00 0.02PoeciUa Iotipiltna 1.03 0.35

Everglades pygmysunfish

Bluespotted sunfishLake chu~kerGolden topminnowEastern mosquitofishLeast killifishFlagfishBrook silversideSunfishesBluefin killifishTadpole madromSailfin molly

nets at plots 5 (KS test, X2 = 14.2, P = 0.0016)and 7 (KS test, X2 = 7.9, P = 0.0393). Small-sizedspecies, such as eastern mosquitofish, least killi-fish, and bluefin killifish, dominated the wet prairiefish assemblages sampled (Table 4). Species rankswere highly concordant between pairs of throwtraps and block nets (Table 5), indicating that thesetwo collection methods sampled the same fish as-semblage.

Discussion

Aquatic systems such as wetlands are composedof diverse habitats that differ with respect to plantspecies composition, plant cover, canopy height,stem density, plant biomass, periphyton volume,and other structural features (Loveless 1959; Gun-derson 1994; Jordan et al. 1994, 1996, 1997).Aquatic macrofauna (e.g., fishes, decapod crus-taceans, insects) probably move about within thesehabitats, searching for foraging opportunities andpotential mates and avoiding potential predatorsand competitors (Wiens 1976). To characterize thespatial ecology and dynamics of aquatic macro-fauna, aquatic ecologists need versatile samplingmethods that are effective (i.e., precise and ac-curate) across a range of environmental conditions(Loftus and Eklund 1994; Rozas and Minel1o1997). Our data, and previous research (Kushlan1981; Freeman et al. 1984; Jacobsen and Kushlan1987; Chick et al. 1992), indicate that throw trapsare a versatile and effective method for samplingfishes in vegetated habitats. Specifically, data ob-tained with throw traps corresponded well with the

actual density, size structure, and relative abun-dance of fish populations sampled. More impor-tant. accuracy of throw traps did not appear to berelated to habitat structural complexity across therange of environmental conditions experienced inthis study. Larger sample sizes might have pro-duced significant correlations between environ-mental conditions and sampling accuracy, but wedo not believe that the strength of observed cor-relations (e.g., r = -0.46 or,.2 = 0.21) is sufficientto mask biologically meaningful differences in fishabundance among habitats. Data from differenthabitats and systems can probably be comparedwithout adjusting for variation in habitat structure;certainly data collected from wet prairies andsloughs throughout the Everglades system (Gun-derson 1994; Jordan et al. 1997) with throw trapsare comparable. The relatively light Kushlan-typethrow trap cannot be used effectively in thick, stiffvegetation such as Cladium, Juncus, Phragmites.Spartina, or Typha; however, throw traps con-structed from heavy aluminum or sheet metal haveproven useful in these habitats (Chick et al. 1992;Jordan 1996a).

Our data indicate that throw traps capture about63% (accuracy 2) of the fishes present in a habitatand that throw trap density estimates are strongly,positively correlated with actual population den-sities. This estimate of accuracy is 10% lower thanearlier estimates provided by Kushlan (1976). Ourclearing efficiency data suggest that Kushlan's es-timates of actual population densities (measuredwithin 190-369-m2 enclosures) were too high. Byusing a clearing efficiency from the throw trap of77%, Kushlan's population density estimate islowered to 60%, which is close to our estimate.This adjustment is probably conservative, giventhat clearing efficiency of small fishes declineswith increasing enclosure size (Shireman et al.1981; Miller et al. 1991). Pilil and Rosenberg(1982) reported throw trap accuracy of 97% forsamples obtained in estuarine sand flats. This value

1018 JORDAN ET At

is higher than ours, possibly because (1) a rela-tively sedentary benthic fish was studied, (2) onlyone comparison was performed, and (3) 63% ofthe enclosure area was sampled by throw trapping(versus 12% in our study and 8% in Kushlan 1976).

Reductions in sampling accuracy arise for a va-riety of reasons, and this is the first study to dis-tinguish between different sources of error. Wefound that 17% of the fishes enclosed in throwtraps were not recovered (i.e., 83% throw trapclearing efficiency). These fishes possibly bur-rowed (Frederick and Loftus 1993) or were dis-carded with sorted detritus (Freeman et al. 1984).Discarding of fishes seems unlikely because wecarefully sorted through detritus in each sweep ofthe bar seine and dip nets. Similar clearing effi-ciencies (-70-90%) have been reported for throwtraps (reviewed by Rozas and Minello 1997).Clearing efficiency may be improved by (1) usinga solid-walled throw trap; (2) removing all vege-tation before emptying the throw trap; (3) pumpingout and filtering water inside a solid-walled throwtrap; or (4) more rigorous sorting of plants anddetritus in the laboratory (Freeman et al. 1984;Jordan 1996a). Although no earlier studies mea-sured both throw trap clearing efficiency and sam-pling accuracy, it was often implied that highthrow trap clearing efficiency translated into highaccuracy (i.e., clearing efficiency was the onlysource of sampling error). Our results indicatedthat an additional 20% (i.e., 100% population den-sity - 63% sampling accuracy - 17% throw trap

clearing efficiency) of the fishes present in a givenlocation somehow avoided throw traps. Meanlength, size distribution, and species compositiondata obtained by throw trapping were highly con-cordant with actual population parameters. There-fore, avoidance of throw traps did not appear tobe size-based or to reflect avoidance behavior ofindividual species. Fishes greater than 100 mm inSL are rare in the Everglades marshes we studied(Loftus and Kushlan 1987; Jordan 1996a; Jordanand Thexler, unpublished data). A very large num-ber of I-m2 samples would be required to deter-mine if these large specimens are accurately sam-pled by throw traps. However, this study does notindicate that these fishes are poorly sampled bythrow traps. The reduction in accuracy that weobserved is very close to the 19% reported byJacobsen and Kushlan (1987). These authors hy-pothesized that throw traps had a reduced effectivesampling area (0.81 m2) because fishes near theedges of a descending trap actively avoided thetrap or were passively flushed outside. Our data

and field observations also suggest that this is themost parsimonious explanation for reduced sam-pling accuracy of throw traps.

Acknowledgments

Casey Caulkins, Jane Jimeian, Betty Jordan,Kate Moss, Heather Soulen, and Andy Thmer pro-vided assistance in the field. Dale Gawlik, WileyKitchens, Franklin Percival, and the Florida Co-operative Fish and Wildlife Research Unit provid-ed logistical support. Dale Gawlik, Lawrence Ro-zas, and an anonymous reviewer provided helpfulcomments on earlier drafts of this paper. Researchwas supported by contract C-E6636 from the SouthFlorida Water Management District. This paper iscontribution 49 of the Southeast EnvironmentalResearch Program.

References

Chick, J. C., F. Jordan, J. P. Smith, and C. C. McIvor.1992. A comparison of four enclosure traps andmethods used to sample fishes in aquatic macro-phytes. Journal of Freshwater Ecology 7:353-361.

Chick, J. C., and C. C. McIvor. 1994. Patterns in theabundance and composition of fishes among bedsof different macrophytes: viewing a littoral zone asa landscape. Canadian Journal of Fisheries andAquatic Sciences 51 :2873-2882.

Connolly, R. M. 1994. Comparison of fish catches froma buoyant pop net and a beach seine net in a shallowseagrass habitat. Marine Ecology Progress Series109:305-309.

Dewey, M. R., L. E. Holland-Bartels, and S. J. Zigler.1989. Comparison of fish catches with buoyant popnets and seines in vegetated and nonvegetated hab-itats. North American Journal of Fisheries Man-agement 9:249-253.

Dionne, M., and C. L. Folt. 1991. An experimental anal-ysis of macrophyte growth forms as fish foraginghabitat. Canadian Journal of Fisheries and AquaticSciences 48:123-131.

Downing, J. A. 1991. The effect of habitat structure onthe spatial distribution of freshwater invertebratepopulations. Pages 87-106 in S. S. Bell, E. D. Mc-Coy, and H. R. Mushinsky, editors. Habitat struc-ture: the physical arrangement of objects in space.Chapman and Hall, New York.

Durocher, P. P., W. C. Provine, and J. E. Kraai. 1984.Relationship between abundance of largemouthbass and submerged vegetation in Texas reservoirs.North American Journal of Fisheries Management4:84-88.

Frederick, P. C., and W. F. Loftus. 1993. Cold-weatherbehavior of estuarine and freshwater marsh fishesas a determinant of nesting success in wading birdsin the Everglades of Florida. Estuaries 16:216-222.

Freeman, B. J., H. S. Greening, and J. D. Oliver. 1984.Comparison of three methods for sampling fishes

FISH SAMPLING IN VBGBT A TED HABITATS 1019

nekton abundance in shallow waters. Thansactionsof the American Fisheries Society 104:46-49.

Kushlan, J. A. 1974. Quantitative sampling of fish pop-ulations in shallow, freshwater environments.Thansactions of the American Fisheries Society 103:348-352.

Kushlan, J. A. 1976. Environmental stability and fishcommunity diversity. Ecology 57:821-825.

Kushlan, J. A. 1981. Sampling characteristics of enclo-sure fish traps. Transactions of the American Fish-eries Society 110:557-662.

Larson, E. W., D. L. Johnson, and W. E. Lynch, Jr. 1986.A buoyant pop net for accurately sampling fish atartificial habitat structures. Thansactions of theAmerican Fisheries Society 115:351-355.

Loftus, W. E, and A. M. Eklund. 1994. Long-term dy-namics of an Everglades small-fish assemblage.Pages 461-483 in S. M. Davis and J. C. Ogden,editors. Everglades: the ecosystem and its restora-tion. St. Lucie Press, Delray Beach, Florida.

Loftus, W. E, and J. A. Kushlan. 1987. Freshwater fishesof southern Florida. Bulletin of the Florida StateMuseum, Biological Sciences 31:147-344.

Loveless, C. M. 1959. A study of the vegetation of theFlorida Everglades. Ecology 40:1-9.

Maceina, M. J. 1996. Largemouth bass abundanco. inaquatic vegetation in Florida lakes: an alternativeinterpretation. Journal of Aquatic Plant Manage-ment 34:43.

Miller, S. J., and six coauthors. 1991. Comparison ofWegener ring and 0.08-hectare block net samplesof fishes in vegetated habitats. Proceedings of theAnnual Conference Southeastern Association ofFish and Wildlife Agencies 44(1990):67-75.

Peterson, C. H., and R. Black. 1994. An experimental-ist's ~hallenge: when artifacts of intervention in-teract with treatments. Marine Ecology Progress Se-ries 111:289-297.

Pfennig, D. W., M. L. G. Loeb, and J.-P. Collins. 1991.Pathogens as a factor limiting the spread of can-nibalism in tiger salamanders. Oecologia 88:161-166.

Pihl, L., and R. Rosenberg. 1982. Production, abun-dance, and biomass of mobile epibenthic marinefauna in shallow waters, western Sweden. Journalof Experimental Marine Biology and Ecology 57:273-301.

Rozas, L. P., and T. J. Minello. 1997. Estin1ating den-sities of small fishes and decapod crustaceans inshallow estuarine habitats: a review of sampling de-sign with focus on gear selection. Estuaries 20: 199-213.

Rozas, L. P., and W. E. Odum. 1987. Fish and macro-crustacean use of submerged plant beds in tidalfreshwater marsh creeks. Marine Ecology ProgressSeries 38:101-108.

Serafy, J. E., R. M. Harrell, and J. C. Stevenson. 1988.Quantitative sampling of small fishes in dense veg-etation: design and field testing of portable "popnets." Journal of Applied Ichthyology 4:149-157.

Shireman, J. V., D. E. Colle, and D. E DuRant. 1981.Efficiency of rotenone sampling with large and

and macroinvertebrates in a vegetated freshwaterwetland. Journal of Freshwater Ecology 2:603-610.

Gilmore, R. G., and five coauthors. 1978. Portable tripoddrop net for estuarine fish studies. U.S. NationalMarine Fisheries Service Fishery Bulletin 76:285-289.

Gunderson, L. H. 1994. Vegetation of the Everglades:determinants of community composition. Pages323-340 in S. M. Davis and J. C. Ogden, editors.Everglades: the ecosystem and its restoration. St.Lucie Press, Delray Beach, Florida.

Heck, K. L., Jr., and L. B. Crowder. 1991. Habitat struc-ture and predator-prey interactions in vegetatedaquatic systems. Pages 281-299 in S. S. Bell, E. D.McCoy, and H. R. Mushinsky, editors. Habitatstructure: the physical arrangement of objects inspace. Chapman and Hall, New York.

Hellier, T. R. 1958. The drop-net quadrat, a new popu-lation sampling device. Publications of the Instituteof Marine Science, University of Texas 5:165-168.

Higer, A. L., and M. C. Kolipinski. 1967. Pull-up trap:a quantitative device for sampling shallow-wateranimals. Ecology 48:1008-1009.

Hoyer, M. V., and D. E. Canfield, Jr. 1996a. Largemouthbass abundance in aquatic vegetation in Floridalakes: an empirical analysis. Journal of AquaticPlant Management 34:23-32.

Hoyer, M. V., and D. E. Canfield, Jr. 1996b. Lake size,aquatic macrophytes, and largemouth bass abun-dance in Florida lakes: a reply. Journal of AquaticPlant Management 34:48.

Hoyer, M. V., D. E. Canfield, Jr., J. V. Shireman, and D.E. Colle. 1985. Relationships between abundanceof largemouth bass and submersed vegetation inTexas reservoirs: a critique. North American Jour-nal of Fisheries Management 5:613-616.

Jacobsen, T., and J. A. Kushtan. 1987. Sources of sam-piing bias in enclosure fish trapping: effects on es-timates of density and diversity. Fisheries Research5:401-412.

Jordan, E 1996a. Spatial ecology of decapods and fishesin a northern Everglades wetland mosaic. Doctoraldissertation. University of Florida, Gainesville.

Jordan, F. I 996b. An assessment of fish and decapodcrustacean responses to hydrology, habitat com-plexity, and marsh management history. Final Re-port to the St. Johns River Water Management Dis-trict. Palalka, Florida.

Jordan, E, K. J. Babbitt, C. C. McIvor, and S. J. Miller.1996. Spatial ecology of the crayfish Procambarusalieni in a Florida wetland mosaic. Wetlands 16:134-142.

Jordan, F., H. L. Jelks, and W. M. Kitchens. 1994. Habitatuse by the fishing spider Dolomedes triton in a north-ern Everglades wetland. Wetlands 14:239-242.

Jordan, F., H. L. Jelks, and W. M. Kitchens. 1997. Habitatstructure and plant community composition in anorthern Everglades wetland landscape. Wetlands17:275-283.

Kjelson, M. A., W. R. Turner, and G. N. Nelson. 1975.Description of a stationary drop-net for estimating

1020 JORDAN ET AL,

small block nets in vegetated and open-water hab-itats. Transactions of the American Fisheries So-ciety 110:77-80.

Sogard, S. M., and K. W. Able. 1991. A comparison ofeelgrass, sea lettuce macroalgae, and marsh creeksas habitats for epibenthic fishes and decapods. Es-tuarine, Coastal, and Shelf Science 33:501-519.

Sogard, S. M., G. V. N. Powell, and J. G. Holmquist.1987. Epibenthic fish communities on Florida Baymudbanks: relations with physical parameters andseagrass cover. Marine Ecology Progress Series 40:25-39.

Sokal, R. R., and E J. Rohlf. 1995. Biometry. 3rd edi-tion. Freeman, New York.

Wegener, W., D. Holcomb, and V. Williams. 1974. Sam-pling shallow water fish populations using the We-gener ring. Proceedings of the Annual ConferenceSoutheast Association of Game and Fish Commis-sioners 27(1973):663-674.

Wiens, J. A. 1976. Population responses to patchy en-vironments. Annual Review of Ecology and Sys-tematics 7:81-120.

Received December 1, 1996Accepted AprIl 16, J 997