Quantitative Sampling of Stream Fish Assemblages: Single- Vs Multiple-Pass Electrofishing

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/229895109

Quantitative sampling of stream fish assemblages: Single- Vs multiple-pass electrofishing

Article in Austral Ecology · July 1998 DOI: 10.1111/j.1442-9993.1998.tb00741.x

CITATIONS READS 61 119

4 authors:

Bradley James Pusey Mark J Kennard Charles Darwin University Griffith University

131 PUBLICATIONS 4,320 CITATIONS 127 PUBLICATIONS 3,296 CITATIONS

SEE PROFILE SEE PROFILE

James Michael Arthur Angela H Arthington Griffith University Griffith University

24 PUBLICATIONS 518 CITATIONS 236 PUBLICATIONS 12,271 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

PhD Research Project View project

The life history and ecology of Macrobrachium spinipes in northern Australia: Exploring the role of hydrological connectivity through a model species View project

All content following this page was uploaded by Mark J Kennard on 23 August 2017.

The user has requested enhancement of the downloaded file. Australian Journal of &ology (1998) 23, 365-374

BRADLEY J. PUSEyl,2, MARK J. KE]~ARD1, JAMES M. ARTHUR3 AND ANGELA H. ARTHINGTON1,2 1Centre for Catchment and In-Stream Res/!arch..Griffith University, Nathan 4111.. Qld.. ZCRC for Tropical Rainforest Ecology and Management.. and 3Paculty of Environmental Sciences..Griffith University, Nathan 4111 Qld.. Australia

Abstract Stream fish assemblageswere sampled by multiple-pass electrofishing and supplementary seine netting in 31 sites in the Johnstone River, north C:ueensland and 28 sites in the Mary River, southeastern Queensland to determine the sampling effort required to adequately describe the assemblages in terms of fish abundances, species composition and assemblagestructure. A significantly greater proportion of the total number of fishes present at each site was collected by the first electrofishing pass in the Mary River (46%) than in the Johnstone River (37%) and this difference was suggested to be due to higher water conductivity in the former river. The mean proportion of the total species richness detected by the first pass was also significantly higher in the Mary River than in the Johnstone River (89% and 82%, respectively). Multivariate comparisons offish assemblagestructure revealed that data collected by the first electrofishing pass poorly estimated the actual assemblage structure within a site and that up to three passes were required for estimates of assemblage structure to stabilize. This effect was evident for comparisons based on both absolute abundance and relative abundance data and was particularly marked for comparisons based on presence/absence,jata. This latter result suggests that, even though most species were detected on the first pass, the addition of rsre species by subsequent passeshad an important effect on the resultant description of assemblage structure. Supplementary seine netting had a greater effect on the determination of assemblagestructure in the Mary Riv,:r than in the Johnstone River. The results are discussed with reference to sampling design in studies of stream fish assemblages and a sampling protocol is recommended that enables the accurate determination of abundance, ri:hness and assemblage structure in small- to medium-sized streams.

Key words: electrofishing, multivariate analysis, Queensland, sampling effort, seine netting.

INTRODUCTION shocking) is the use of an electrical field applied to the aquatic environment to attract and stun fish, thus The quantitative collection of freshwater fishes is not enabling their capture. The earliest documented use an easytask and the accurate determination cf fish den- of electric fields to capture fishes was in 1863, but sity has been a '...weak point in much SOI,histicated its routine use in fisheries research did not occur until fish research' (Zalewski 1983 p. 177). MallY collect- the 1930s (Hartley 1989). Electrofishing has been ing techniques have been devised, ranging from netting using a variety of different forms (fyke, gill, seine, routinely used in Australia since the mid-1960s (Koehn etc.), to poisons such as rotenone and to the Ilse of elec- & McKenzie 1985). One of the advantages of electric currents. Each collecting method is likc:ly to have trofishing is that capture does not usually result in death and fishes are able to be released back into the some bias or selectivity for different taxa or sizesranges. environment. Injury does occur occasionally, particu- The poisoning of entire stream reaches a)pears the larly to the spinal column, but often these injuries are most effective means of acquiring accurat<: estimates not fatal (Spencer 1967). Our own experience suggests of fish density and assemblage structure (Larimore that electrofishing-related mortality rates for a range 1961; Boccardy & Cooper 1963); however: it is rarely of species, in a range of rivers, are generally less than desirable in view of its unpredictable natllre and its 5% (B. J. Pusey et al. unpubl. data). negative environmental effect, particularly if studies are The method is not without bias however. Galvano- intended to be long term. taxic and galvanonarcotic responsesvary among species Wiley and Tsai (1983) found that ei<:ctrofishing and among size classeswithin species (Larimore 1961; provided better and more consistent estirrates of fish Boccardy & Cooper 1963; Mahon 1980; Mahon & densities than did seine netting. Electrofishing (electro- Balon 1980; Balayev 1981; Wiley & Tsai 1983; Koehn & McKenzie 1985). Catchability reportedly decreases Accepted for publication September 1997. with increases in the number of times an individual has 366 B. J. PUSEY ET AL,

been shocked, with this refractory period 1:lsting generally similar in structure except for slightly lower between three and 24 h (Cross & Stott 1975). mean water velocities and a lower proportion of coarse Electrofishing efficiency can also vary with water con- substrata (rock and bedrock). Various types of in- ductivity (Hill & Willis 1994), voltage, direction of stream cover (woody debris, macrophytes, etc.) were movement of target fish within the electric fielj and present in most sites but constituted such low propor- water temperature (Regis et al. 1981), stream width tional coverageof a site area that they were not included (Kennedy & Strange 1981) and a range of other bio- in Table I. Water temperatures in the Mary River were logical, environmental and technical factors (Za.ewski similar to those recorded in the Johnstone River & Cowx 1989). None the less, electrofishing is s1jll the (19.8 :t 0.6°C); however, water conductivity differed most effective, non-destructive sampling procedllre for greatly between the two rivers. Conductivity ranged fishes in small- to medium-sized streams (Zale\{ski & from 161.5-1889.9jl.S.cm-1 (mean conductivity of Cowx 1989; Schill & Beland 1995). Unfortunately, the 621.7 :t78. 7 jl.S.cm-l) in the Mary River (n = 28). variability in efficiency appears to be site speciflc and Importantly, water clarity was always high in both rivers no general rules are possible (Larimore 1961; Cross & (mean turbidity 1.6:t 0.2 NTU and 3.1 :t 0.5 NTU for Stott 1975; Koehn & McKenzie 1985). the Johnstone and Mary Rivers, respectively) and is The increasing focus within Australia on the sus- unlikely to have been a significant impairment to tainable management of water resources has resulted electrofishing efficiency in this study. The mean area in a greater emphasis on the determination ofthc: habi- and length of stream electrofished was 320:t 48.5 m2 tat and flow requirements of native fish species (Harris and 39.6 :t 4.0 m, respectively, for the Johnstone River 1995). Our current research program is focu~,ed on and 272:t 39.7 m2 and 39.4:t 2.8 m, respectively, for determining the effects of stream discharge on spatial the Mary River. and temporal variation in fish assemblagestructlre and one applied outcome of this program is the pn>vision of simple methods for use by goverrlment agenci,~scon- Sampling procedures cerned with the management of Queensla~d'~ water Electrofishing was performed in both rivers using a resources.The aim of this contribution is therefore two- portable back-pack electro fisher. In the Johnstone fold. First, we want to detail the method emplcyed by River, a Smith-Root Mk 12 POW electrofisher was used us in current researchand to allow comparison between whereas a Smith Root Mk 7 electrofisher was used in the current and prior research (Pusey et al. 1993,1995; the Mary River. Various output wave forms, voltages Pusey & Kennard 1996). Second, we want to as~.essthe and pulse frequencies can be selected on the Mk 12 effort required to accurately determine the density, electro fisher but the choice is more limited in the Mk species composition and assemblagestructure offresh- 7 model. In general, we chose to restrict the use of alter- water fishes in well-defined hydraulic units (i.e. riffle, native settings in the Johnstone River (200-400 V and run or pool) of streams of eastern Queensland. setting J4; frequency: 70 Hz, pulse width: 4 ms) so as to approximate the output generated by the Mk 7 model. Prior experience suggests that this output was METHODS the most effective for collecting a wide range of species

Study area Table I. Average structure of the habitat at each study site Electrofishing was undertaken in two rivers; the in the Johnstone River and Mary River Johnstone River (146°0'E, 17°30'S) of nl)rthern Queensland and the Mary River (152°35'E, 25°30'S) JohnstoneRiver Mary River of southeastern Queensland. Thirty-one sitl:s were sampled by electrofishing within the Jolmstone Habitat variable (n=31) (n=28) River drainage between July and Septembe: 1994. Width (m) 9.56 (7.07) 8.59 (7.92) 0.37 (0.16) 0.28 (0.20) Mean water temperature during this periJd was Depth (m) Water velocity (m.sec.l) 0.14 (0.10) 0.06 (0.11) 20.0 :!: 0.5 (S.E.)OC and conductivity ranged from % Mud 11.3 (14.9) 5.5 (10.7) 10.5-54. 7 ~S.cm-1 with a mean of36.8:!: 1.81J.S.cm-1 % Sand 17.1 (20.0) 15.5 (19.1) (n = 31). The streams investigated were of :;mall to % Fine gravel 16.1 (12.7) 21.3 (9.6) medium width, rarely deeper than 1.5 m and of vary- % Gravel 12.2 (13.0) 27.4 (14.7) ing mean water velocity and contained a wide range of % Cobbles 11.0 (10.2) 23.9 (17.4) % Rock 16.5 (17.8) 6.3 (12.3) types of substratum (Table 1). Areas sampled covered % Bedrock 16.3 (24.3) 0.3 (0.9) a wide range of habitat types (e.g. cascades: rapids, riffles, runs and pools). Sites electrofished in tIe Mary The values given are the means (.:!:.1 SD). Substrate val- River (n = 28) during September and Octo1:er 1994 ues are given as the mean proportion (%) of the substrate covered a similar range of habitat types al.1d were for each site. QUANTITATIVE SAMPLING OF STREAM FISH 367 within a range of different rivers and habitat type!l. Both the number of fish present and the number of elec- electrofishers used a standard Smith-Root anode trofishing passesconducted. (25 cm diameter ring attached to a 2 m poll:) and cathode (3.2 m wire cable). Electrofishing was conducted within discretl~ hyd- Data analysis raulic habitat units (i.e. riffle, run or pool) within a Univariate analysis ofabundance and speciesrichness stream reach. Prior to electrofishing, block seines (9 mm stretched mesh size) were placed at the top and The total number of fishes collected from each study bottom of the study site. The bottom of the net was site after all electro fishing passeshad been completed, weighed down with substratum to prevent the move- and after any additional seine netting, was taken to rep- ment of fishes either into or out of the study sitl~. The resent the total number of fishes present within each bottom seine contained a central bag or purse which site (NT). The cumulative total abundance collected up was always positioned in the thalweg. Nets were to pass i was represented by Ni. The proportional con- deployed simultaneously whenever possible, otherwise tribution of each cumulative pass was estimated the bottom net was positioned first. Approximately (Ni/NT). Therefore, at each site, the cumulative abun- 20 min were allowed to elapse prior to comml~ncing dance was standardized by the total site abundance. electrofishing in order for fish to resume IJormal The proportional contribution of seine netting to NT behaviour. was similarly standardized. Six electrofishing passes,as The operator and one assistant then commenced opposed to five, were necessary to achieve full deple- electrofishing, using short, intermittent pulses, ensur- tion in only two of the 28 sites in the Mary River. In ing that all of the enclosed area was electrofished once. order to simplify analyses, the sixth pass was added to This procedure was considered to represent a single the fifth pass in these two sites. This was not considered electrofishing pass. In the casewhere stream width was problematic because the fish collected on the sixth pass less than 4 m, electrofishing commenced at the constituted less than 5% of the total collected at each upstream seine and proceeded to the bottom seine. In site and consisted of only one species. Initial analysis wider streams, the operator moved through the study of the relationship between variances and means indi- site in a zig-zag fashion. All stunned fish were 1etted cated heterogeneous variances. To stabilize the vari- and placed immediately in stream water in ~ 70 L ances, the standardized cumulative abundances were plastic container towed behind the assistant or, in the log (x + 1) transformed prior to further analyses case of large fish (e.g. therapontids, plotosids and (Underwood 1997). anguillids), immediately transferred to an add:,tional Preliminary analysis indicated that the partial corre- 70 L container on the stream bank. Fish present in the lations obtained from the sums of squares and bag of the bottom net were removed at the completion crossproducts error matrix betWeencumulative passes, of each pass and were counted as part of each rl~spec- were significant (P< 0.0001, d.f= 57). A test of tive electrofishing pass. sphericity on this matrix was also significant Approximately 15 min were allowed to lapse IJefore (P< 0.0001, d.f. = 14). These analysesindicated inter- the second pass commenced. All subsequent passes dependency of the within-river effects (standardized proceeded in the same manner until few or no more cumulative passes), therefore a repeated-measures fish were collected by electrofishing. A maximum of five Analysis of Variance (SAS Institute Inc. 1989) was per- passes was required for complete depletion ir both formed. For all significant standardized cumulative rivers except for two sites in the Mary River. llle rate pass-by-river interactions, orthogonal contrasts of depletion of some small agile species (RelTllpinna betWeen the nth level cumulative pass and the total col- semoni, Melanotaenia spp. and Craterocephalusspp.) lected by electro fishing were performed for each river. from the study site with each sequential pass wa! often A single contrast betWeen the first pass of each river observed to be lower than the remaining assemblage relative to the total cumulative abundance for each river members. Whenever possible, seine netting (melih size was also performed. 9 mm) was used to collect all of these remaining s:hool- Total species richness was taken to equal the num- ing fishes. Snorkelling was occasionally underta]~en to ber of species collected at an individual site once all assessthe success of the sampling, and in most cases, fishes had been removed by both electrofishing and few fish were observed. If many fish were obser..ed to seine netting. These data were treated similarly to abun- be present, then electrofishing and seining continued. dance data in that we estimated the cumulative pro- Most fish collected were returned to the study site alive portion of the total richness detected by each after identification and measurement. The dural ion of electro fishing pass and all preceding passes. A similar the sampling procedure described above was commonly repeated-measures analysis was used to analyse vari- between two and four hours in the Johnstone Ri¥er and ation using log (x + 1) transformed cumulative species one and three hours in the Mary River with duration richness. The effect of habitat structure on the effi- being determined mainly by the size of the study site, ciency of the first pass was assessed by correlation 368 B PUSEY ET AL analysis of N1/Nr (arcsine transformed) again!;t the tions were based on two dimensions with stress levels mean of a range of variables describing habitat !itruc- consistently less than 0.05. ture including stream width, depth, water velocity, the The second analysis was based on comparisons of proportional contribution of substrate types lisled in ordinations for each data set within each river. In this Table 1 and the proportional areal cover of cover case, the underlying matrix of Bray-Curtis dissimilar- elements including macrophyte beds, small and large ities was derived from the original site by species matrix woody debris, leaf litter and root masses. and ordination (SSHMDS) and was performed using the SSH procedure in PATN (Belbin 1995). Fifty random starts were performed and the randomization Multivariate analysis of assemblagestructure which showed the lowest stresslevel was chosen for fur- In this analysis we were interested in assessingwl1ether ther examination. Each ordination was performed for our determinations of the fish assemblage struCtilIe at three dimensions because any further increase in the each site changedsignificanrly with increasing SaIrlpling number of dimensions did not greatly decrease the effort. To this end we examined data from the firS1 eiec- stress levels, all of which were below 0.15. Resultant trofishing pass, first plus second electrofishing pal;sand ordinations were compared using a Procrustes analy- so on and compared the assemblage structures deter- sis (Gower 1971; Sibson 1978), in which the ordina- mined from each such combination against the al;sem- tion plot obtained for each successivecumulative pass blage structure determined from all electrofishing (the test plot) is rotated, scaled or reflected to fit the passes and any supplementary seine netting (= total ordination plot generated for the total assemblage (the assemblage). Differences in fish assemblage str1Jcture target plot) in order to minimize the sum of squared were examined by tWo different methods for each of distances between the samples. This yields a measure three separate data sets. These data sets were ~ener- of overall fit, the Root Mean Squared Residual (RMS), ated for each river and consisted of: (i) species abun- which is the square root of the mean of the squared dance data (no standardization), (ii) relative distances between corresponding samples in the fitted abundances (standardized by site total), and (iii) and target ordinations (Sibson 1978). This method species presence/absencedata. The first tWo genc:rated quantified the match between the multidimensional data sets were log (x + 1) transformed to downvreight scaling ordinations of each cumulative pass and the the abundance of a few numerically dominant species. total assemblage. The RMS obtained from this pro- All analyseswere based on the Bray-Curtis dissimilar- cedure is a composite of the differences in the under- ity measure (Bray & Curtis 1957). Simulation studies lying dissimilarity matrices and the ability of the (Faith et aI. 1987) have indicated that this dissimilar- ordination process to fit these distances in a dimen- ity measure is an effective measure of ecological dis- sionally reduced space. tance. We suggest that both the Mantel analysis and the The first analysis used a Mantel test (Mantel 1967; Procrustes analysis are informative. Mantel's analysis Manly 1991) to compare the site by species di!;simi- directly tests the 'distance' between two dissimilarity larity matrix for each cumulative pass with the I1.1atrix matrices. Ordination, which most researchersuse as an derived for the total assemblage, for each of the three appropriate means to examine variation in species generated data sets within each river. The Mantl:l test abundances, is based on interpretation of the under- (a randomization procedure) uses a standardized dis- lying dissimilarity matrix, except that the interpretation similarity matrix obtained by subtracting the I1.1atrix occurs in a dimensionally reduced space with a poten- minimum and dividing by the matrix range; it pro:luces tial consequent loss of information. Therefore, it is use- an averagedistance measure betWeeneach matri,: pair. ful to directly compare the differences in the matrices The test matrix was compared with onethousarul ran- (i.e. Mantel comparisons of the standardized distances) domized test matz'icesusing the same standardized dis- and to indirectly compare differences when the under- tance measure to obtain each Mantel test statistic. lying structure is displayed in a reduced set of dimen- All possible betWeen-passMantel distances (i.e. pass sions (i.e. RMS comparisons from the Procrustes 1 vs pass 1 + 2, pass 1 vs all passesplus sein~ net- analysis). ting, pass 1 + 2 vs pass 1 + 2 + 3, pass 1 + 2 vs all passes plus seine netting, etc.) were calculated l:or aU RESULTS possible betWeen-pass comparisons. These Nlantel distances were then used to form new matrices for each Abundance of the three data sets for each river. These new matrices were then used in subsequent Semi-Strong l1ybrid The number of fish collected from each site varied Multidimensional Scaling (SSHMDS) ordiru!tions betWeen 90 and 200 and 128-190 for the Johnstone (Belbin 1991), as implemented in the SSH procedure and Mary Rivers, respectively. The exact F statistic in PATN (Belbin 1995), to generate plots reprl:sent- (Wilks Lambda) from the repeated measures ANOYA, ing the differences betWeen all passes. These ordina- indicated a significant cumulative pass-by-river inter- QUANTITATIVE SAMPLING OF STREAM FISH 369

Table 2. Fvalues and their associatedlevels of significancefor a repeated-measuresAnalysis of Varianceof within river (cumulative passes)and betweenriver variation in log(x+ l) transformedproportional cumulativeabundance and speciesrichness

Cumulativeabundance Cumulative speciesrichness

Source of variance dof. F P F P River I 2.14 NS 1.17 NS Site (River) ji7 Pass 4 136.20 <0.0001 13.12 <0.0001 Pass X River (Error I) 5 10.75 <0.0001 4.79 <0.001 Contrast 1: Johnstone R.(PI) vs Mary R.(pI) 5.38 <0.05 9.91 <0.005 Contrast group 2 (Johnstone R.): PI vs P5 307.02 <0.0001 52.09 <0.0001 P2 vs P5 92.28 <0.0001 21.64 <0.0001 P3 vs P5 88.47 <0.0001 10.57 <0.005 P4 vs P5 78.19 <0.0001 7.47 <0.05 Contrast group 3 (Mary R.): PI vs P5 121.79 <0.0001 11.90 <0.005 P2 vs P5 51.81 <0.0001 0 NS P3vsP5 13.99 <0.001 0 NS P4 vs P5 3.96 NS 0 NS Pass X Site (River) (Error 2) ,)1-3

See text for further explanation of contrasts. Cll1Dulative electrofishing passes as denoted by PI to P5 where the numeric value is the number of passes (i.e. P5 = pass I + 2 + 3 + 4 + 5). NS, not significant.

(a) action (P< 0.0001, Table 2). Contrasts betWeenthe nth cumulative pass and the total indicated that for the Johnstone River, each electrofishing pass significantly increased the mean cumulative proportion of the total 0.8 abundance (P< 0.0001 for all comparisons, Table 2, Fig. la). In the Mary River the first three passes con- tributed significantly to the total proportion of fish col- 0.6 lected (P < 0.001) but the remaining passes did not significantly alter the total number of fish collected by electrofishing (P> 0.05, Table 2, Fig. la). Electrofish- t: 0 ing efficiency (proportion of fish collected on the first :;:... 0.4 0 pass) differed significantly betWeen the tWo rivers Co e (P< 0.05, Table 2) with a higher proportion offish col- Co lected on the first p.assin the Mary River than in the .~ 0.2 Johnstone River (mean proportion = 0.46:t 0.04 SE iU 1 2 3 4 5 ~ and 0.37:t 0.03 SE, respectively) (Fig. la). Seine net- E (b) ~ 1 ting, after completion of all electrofishing passes,col- (J t: lected a mean of 14% of the total number of fishes !IS Q) present at each site at those sites in which it was used. ~ Electrofishing efficiency was not significantly correlated (P> 0.05) with any parameter describing habitat structure in either river (Table 1). 0.8

Fig. 1. Sequential increase in the mean cumulative proportion CSE)of Ca)total abundance and (b) total species richness collected by each electrofishing pass in the Mary and Johnstone Rivers. C8) Mary River; CO)Johnstone River. Data 0.6 from supplementary seine netting is not included in this 2 3 4 5 figure and hence mean cumulative proportions do not sum Pass number to 1. 370 B. J. PUSEY ET AL

Species richness presence/absence) from both rivers, that the standardized difference between the matrix of site dissimilari- The number of species offish collected from each site ties for each cumulative pass and the total assemblage varied from 3 to 14 and 2 to 14 for the Johnstone and dissimilarity matrix was significantly different from ran- Mary Rivers, .respectively. The first electrofishing pass dom (P< 0.001). This indicated that, for both rivers, detected a significantly higher mean proportion 0;[ the a single pass was sufficient to indicate that the spatial total number of species in the Mary River than iI1lthe pattern of fish assemblagesdeviated significantly from Johnstone River (0.89 :t 0.02 SE and 0.82 :t 0.03 j).E., random. Further sampling effort did not alter this con- respectively) (P< 0.005, Table 2, Fig. Ib). The exact clusion. However, it is evident in Fig. 2a, that further F statistic from the repeated measures ANOVA,indicated sampling effort did reduce the magnitude of the a significant cumulative pass-by-river interaction Mantel statistic and therefore the addition of new infor- (P< 0.001). Contrasts between the nth cumulative pass mation from each sequential pass resulted in a better and the total indicated that for the Johnstone River, representation of the total assemblage structure. eachelectrofishing pass significantly increased the I11ean The Procrustes analyses for both the Mary River and cumulative proportion of the total speciesrichness col- Johnstone River data sets revealed a similar pattern of lected (P< 0.005 for all comparisons, Table 2, Fig. Ib). decreasing RMS values with the addition of data from In the Mary River however, no additional species Iwere the second pass,followed by little change with the addi- collected after the first two passes (Table 2, Fig. tlb). tion of data from the third pass, especially in the Mary Seine netting collected fewer species than did I~lec- River (Fig. 2b). The presence/absence data set was trofishing because the majority of species had been more influenced by the addition of rare species and detected already and removed by electrofisbing. therefore generally maintained higher Mantel distance Moreover, on only three of the 22 occasions in which values and RMS values than those comparisons based seine netting was employed in the Johnstone Rive:: did on abundance or relative abundance (Fig. 2). These it collect a species not already collected by electro fish- data indicate that after three passes,little change in the ing. Six of the 14 occasions in which seine netting was determination of fish assemblagestructure was evident. employed in the Mary River resulted in the dete<:tion The abundance and relative abundance data sets for of any additional species, but in all cases except one, the Mary River maintained higher Mantel distance this was limited to one additional species. values and RMS values after completion of all electrofishing passes than those of the Johnstone River (Fig. 2), suggesting that supplementary seine netting Multivariate analysis of assemblage structure influenced the determination of total assemblage struc- The Mantel randomization tests showed, fo:, all ture more in the Mary River than in the Johnstone data sets (absolute abundance, relative abundance and River.

(a) Mary River Johnstone River 0.12 0.12

Q) 0.1 0.1 c0 cIS 0.08 0.08 u; :c o.~ O.

Q) C 0.04 0.04 cIS ~ 0.02 0.02

0 IvsT 2vsT 3vsT 4vsT 5vsT IvsT 2vsT 3vsT 4vsT 5vsT

~ ro :J 0- In C Fig. 2. Changes in (a) Mantel as Q) scores (cross matrix correlations) E and (b) residual mean square iO :J values with increasing number of "C electrofishing passesin the Mary °in Q) and Johnstone Rivers for com- II: parisons based on abundance, relative abundance and presence/absencedata. QUANTITATIVE SAMPLING OF STREAM FISH ~71

Ordination plots of all pairwise Mantel comparisons suggeststhat supplementary seine netting added impor- between sequential passes,summarize the changes in tant new information in the Mary River. assemblagestructure for both rivers (Fig. 3) Imd support the conclusjons drawn from the previous two analyses. In the Johnstone River, each succesiiive pass had a decreasing effect on the determination of assem- DISCUSSION blage structure. The addition of data from thc~second Meaningful environmental monitoring programs must and third passes most strongly affected the deter- be based on both accurate and precise sampling mination of total assemblage structure, and sc:inenet- methods (Maher et al. 1994). Unfortunately, this ting added very little new information. Data from the increases the costs of such programs due to the Mary River catchment shows similar trends 'mth the increased personnel and time required to quantify greatest amount of new information being added with abundances, assemblage composition and the re- the second pass and subsequent passes addiJlg little. lationships of both to the habitat and environment Although the Mantel tests indicated that the difference (Sheldon 1984). In the past, attention has been between the final electrofishing pass and tfle total focused on determining whether standard collection assemblageare closely 'correlated' in both rivers, Fig. 3 and analytical methods may be adapted to allow a more

(SI) Mary River

0.5

-0.5

1.5 -0.5 0 0.5

0.5

0 Fig. 3. Ordination plots of changes in the estimation of fish -05 assemblage structure with increasing number of electrofishing passes. (a) Abundance, (b) relative proportion, and (c) presence/absence. The matrix .1.5 upon which each ordination is -1-5 -0.5 0 0.5 based consisted of Mantel distance scores for each possible comparison (i.e. pass 1 vs pass 1 + 2 ...pass 1 vs all passesplus seine netring, pass 1 + 2 vs pass 1 + 2 + 3 ...pass 1 + 2 vs all passes plus seine netting, etc.) Each data point represents the number of passes (i.e. 2 = first plus second pass, 3 = first plus second plus third pass) and T denotes the total assemblage (all electrofishing passes plus sup- plentary seine netting). Stress levels for all ordinations were below 0.05. 372 B. J. PUSEY ET AL.

rapid but still meaningful description of biological com- passes. Despite this apparent improvement over that munities (Storey et at. 1991; Armitage & Pardo 1 S'95; observed in the Johnstone River, relatively higher Marchant et at. 1995; Ruse & Wilson 1995; Resh et al. Mantel scores and RMS values on the last pass 1995). With the exception of Harris (1995) and suggest that supplementary seining after electrofishing Growns et al. (1996), Australian studies that address provided more new information than it did in the John- such issues in freshwaters have focused on mac:ro- stone River. invertebrates. Elsewhere, where electrofishing has l1ad Angermeier and Smogor (1995) reported that in a longer history and freshwater fisheries are tradition- order for estimates of species richness and relative ally a more prominent issue, this is not the case. abundances to stabilize, then single-pass electrofishing For example, comparisons have been made of the rel- must be conducted over stream lengths of between 22 ative efficiencies of electrofishing and other census tc~ch- and 67 times the stream width. This high sampling niques in North America (see Introduction). Vadas and effort was in part related to different habitat usage arth (1993) compared electrofishing and seine netl:ing by different species (and therefore as many habitat in an open sampling design (only the bottom of the elements as possible needed to be sampled), but was study area blocked by a seine) and found that both sam- most strongly related to overall low population densi- pling techniques provided concordant fish assemb]age ties of fishes. Rare species required greater effort to patterns for riffles and runs, but pool samples differed detect. Estimates of relative abundance required less substantially in species diversity and dominance pat- effort than determination of species richness (Anger- terns. They further suggested that both methods meier & Smogor 1995). Previous research on fish should be used together wherever possible. In some cir- assemblagesin eastern Queensland (Pusey et al. 1993, cumstances (i.e. large expanses of moderately deep 1995; Pusey & Kennard 1996) used a single-pass elec- water over a fine substrate with little in-stream cover), trofishing approach which differed from the approach seine netting is effective (B. J. Pusey et al. unpubl. data). described here in that the single pass occurred over However, in more heterogeneous habitats, electrofish- 200-300 m of stream. In most cases, the length of ing is able to detect more species than does seine net- stream sampled approximated about 20-30 stream ting (Pusey & Kennard 1996). The relative efficien::ies widths. In this regard, estimates of speciesrichness and of each method appear to be related to the prefexred relative abundances were probably adequate given that habitat of individual species (e.g. benthic vs pelal~ic) described patterns of spatial and temporal variability (Vadas & arth 1993). In a study of habitat use by YOllOg were often strongly expressed. However, when a study brown trout and Atlantic salmon, surface observati

extent and nature of this bias would be nec:ded if etc.) and an increased number of mesohabitat units (i.e. single-pass electrofishing was the sole means of sam- pool, run and riffle) within a stream reach. Additional pling stream fishes. We were initially surprised not to sampling methods (seining, gill netting, angling and detect any relationship between habitat strucnlre and underwater snorkelling) rarely detected additional electrofishing efficiency; however, it is of little concern species in this study (Pusey & Kennard 1996). Some to the outcomes of this study as we sampled a very sim- assessmentof the importance of rare species is needed ilar array of habitat types in both rivers (B. J. Pusey et under a single-pass protocol, particularly if rare species al. unpubl. data). Specific differences in susceptibility are an important component of a region's fish fauna. to capture by electro fishing as indicated by the results For example, 25 of the 66 species recorded by Pusey of this study would also necessitatequantificatioI1. of this and Kennard (1996) collectively comprised less than effect. The estimation of species richness per site by a 1 % of the total number of fishes collected. Rare species single pass initially appears adequate (collecting 82 and may be downweighted or omitted from subsequent 89% of species in the Johnstone River and Mary River, analysesdepending on the focus of the research but this respectively) and if we were only interested in spatial may be problematic if such species are of high conser- or temporal variation in the number of species present, vation or ecological significance. this may be acceptable. To assume a constant or ran- In conclusion, this study has shown that estimates domly fluctuating catchability may however, be a of fish species richness and abundance, and hence major problem (Mahon 1980). Semiquantitative sam- assemblage structure, varied according to the amount pling in species-rich assemblagesand in heterogj:neous of effort expended in their determination. Further- habitats has been found to be ineffective (Mahol:l 1980; more, the amount of effort required to achieve an accu- Zalewskii 1983) and Mahon (1980 p. 357) su~:gested rate determination of these parameters also varied that the most appropriate way of decreasing error was among study rivers. Accordingly, we recommend that to increase the total effort and that this was best done the minimum protocol for sampling stream fishes by'. ..increasing the number of fishings used in the should include the use of top and bottom block seines estimate'. and at least three electrofishing passesfollowed by seine The desired outcome of a sampling design is the most netting where practicable. If analysesare limited to an important point that needs to be addressed at the examination of the distribution of fishes within water- beginning of any research or monitoring program. For sheds only (i.e. presence/absence), the area sampled is example, data collected by single-pass electro fishing relatively small (i.e. individual riffles or pools) and the will only allow the detection of significant aI1ldreal detection of rare species is considered important, then spatial and temporal changes in abundance of most four electrofishing passesmay be necessary.Single-pass species within an assemblageif the probability of cap- electrofishing over an extended length of stream may ture on the first pass is constant for all passes cr vari- improve the ability to detect rare species but may com- ation in abundances is pronounced. This would apply promise the ability of the investigator to clearly exam- also to examination of speciesrichness and asserllblage ine those factors responsible for determining spatial structure. For example Jowett & Richardson ~:1996) patterns of distribution. were able to use data collected by single-pas!! electrofishing in an analysis of the distribution and abun- ACKNOWLEDGEMENTS dance of freshwater fish in New Zealand rivers b'~cause all species were collected with equal probability on the This research was conducted with financial support first pass. The results of the present study suggests that by the Land and Water Resources Research and this is not the case in the two rivers studied her(: given Development Corporation and the Queensland that significant changes in species richness arui sub- Departments of Primary Industries and Natural stantial changes in the estimation of assemblage struc- Resources. We have been aided in the field by the assis- ture occurred with the addition of data from succ:essive tance of Stephen Mackay, Richard Ward and Jason passes. If however, the desired outcome was a deter- Bird. JasonBird aided in the production of figures. The mination of species richness and relative abundances, comments of two anonymous referees greatly improved then single-pass electrofishing over a long length of the manuscript, for which we are grateful. stream may be adequate (Angermeier & Smogor 1995). The main aim of a survey of the freshwaler fish REFERENCES fauna of rivers of the wet tropics region of Queensland (Pusey & Kennard 1996) was to estimate species rich- Angermeier P. L. & Smogor R. A. (1995) Estimating number of ness and detect geographical differences in asseI11blage species and relative abundances in stream-fish communities: effects of sampling effort and discontinuous spatial composition. Single-pass electrofishing over about distributions. Can. J. Fish. & Aqu. Sci. 52, 936-49. 200 m of stream at each site was employed to achieve Armitage P. D. & Pardo I. (1995) Impact assessmentof regula- these aims. This allowed the sampling of many micro- tion at the reach level using macroinvertebrate information habitats (i.e. log jams, macrophyte and leaf lineI' beds, from mesohabitats.Reg. Riv. Res.Manag. 10, 147-58.

1 374 B. J. PUSEY ET AL.

BaleyevL. A. (1981) The behaviour of ecologically diffen:nt fish Manly B. F. J. (1991) Randomizationand Monte Carlo Methodsin in electric fields II. Threshold of anode reaction and Biology. Chapman and Hall, London. tetanus. J. Ichlh. 21, 134---43. Mantel N. (1967) The detection of disease clustering and a Belbin L. (1991) Semi-strong Hybrid Multidimensional generalised regression approach. Cancer Research 27, Scaling: a new ordination algorithm. J. veg.Sci. 2, 4~1-6. 209-20. Belbin L. (1995) PAIN: Pattern Analysis Package: 1echnical Marchant R., Barmuta L. A. & Chessman B. C. (1995) Reference. CSIRO Division of Wildlife and E(:ology, Influence of sample quantification and taxonomic resolu- Canberra. tion on the ordination of macroinvertebrate communities BoccardyJ. A. & Cooper E. L. (1963) The use ofrotenoJ:le and from running waters in Victoria, Australia. Mar. Freshw. electrofishing in surveying smaIl streams. Trans.Ame'. Fish. Res.46,501-6. Soc. 92,307-10. PuseyB. J., Arthington A. H. & Read M. G. (1993) Spatial and Bray J. R. & Curtis J. T. (1957) An ordination of the upland temporal variation in fish assemblagestructUre in the Mary forest communities of southern Wisconsin. Ecol. Mo7log.27, River, south-eastern Queensland: the influence of habitat 325-49. structUre. Env. Bioi. Fishes37,355-80. Cross D. G. & Stott B. (1975) The effect of electric fishing on Pusey B. J., Arthington A. H. & Read M. G. (1995) Species the subsequent capture of fish. J. Fish BioI. 7, 349-57. richness and spatial variation in fish assemblagestructUre in Faith D. P., Minchin P. R. & Belbin L. (1987) Compositional two rivers of the Wet Tropics of north Queensland. Env. dissimilarity as a robust measure of ecological diswnce: a Bioi. Fishes42, 181-99. theoretical model and computer simulations. vegelalio69, Pusey B. J. & Kennard M. J. (1996) Species richness and 57-68. geographicalvariation in assemblagestructUre of the fresh- Gower J. C. (1971) Statistical method!. of comparing different water fish fauna of the Wet Tropics region of northern multivariate analysis of the same data. In: Malhemllcicsin Queensland. Mar. Freshw.Res. 47,563-73. Archaeological and Historical Science (eds F. R. Hodson, Regis J., Panee E. & Lebreton J. D. (1981) A new method for D. G. Kendall & P. Tautu) pp 138-49. University Press, evaluating the efficiency of electric fishing. Arch. fur Edinburgh. Hydrobiol. 93, 68-82. Growns I. 0., Pollard D. A. & Harris J. H. (1996) A compari- Resh V. H., Norris R. H. & Barbour M. T. (1995) Design and son of electric fishing and gillnetting to examine the effects implementation of rapid assessmentapproaches for water of anthropogenic disturbance on riverine fish comm\Jnities. resource monitoring using benthic macroinvertebrates. Fish. Manag. & Ecol. 3, 13-24. Aust. J. Ecol. 20, 108-121. Harris J. H. (1995) The use of fish in ecological assessments. Ruse L. P. & Wilson R. S. (1995) Long term assessmentof AUSI.J. Ecol. 20, 65-80. water and sediment quality of the River Thames using Hartley W. G. (1989) The history of electrofishing. In: 1.ishing chironomid pupal skins. In: Chironomids: from Genes to Wilh Electricity-Application in FreshwalerFisheries M 2nage- Ecosystems(ed. P. S. Cranston) pp. 113-24. CSIRO menl(ed. I. G. Cowx & P. Lamarque) pp. 1-3. Fishin:~New Publishing, Melbourne. Books, Oxford. SAS Institute Inc. (1989) sAS/STATUser's Guide, Version 6, 4th HeggenesJ., Brabrand A. & SaltveitS. J. (1990) Comparison of edn. SAS Institute ~c., Cary. three methods for studies of stream habitat use by young Schill D. J. & Beland K. F. (1995) Electrofishing injury studies: Brown Trout and Atlantic Salmon. Trans.Amer. Fish. Soc. a call for population perspective. Fisheries20, 28-':9. 119,101-11. Sheldon A. L. (1984) Cost and precision in a stream sampling Hill T. D. & Willis D. W. (1994) Influence of water conductivity program. Hydrobiol. 111, 147-52. on pulsed AC and pulsed DC electrofishing catches for Sibson R. (1978) Studies in the robustness of multidimensional largemouthed bass. N. Amer. J Fisl!. Manag. 14,202-7. scaling: Procrustes statistics. J. Roy. Stat. Soc. 40, JowettI. G. & RichardsonJ. (1996) Distribution and abundance 234-8. of freshwater fish in New Zealand rivers. N.Z. J Mar. SpencerS. L. (1967) Internal injuries of largemouth bass and Freshw.Res. 30, 239-55. bluegills by electricity. Prog. Fish Cult. 29, 168-9. Keller E. A. & Melhorn w: N. (1978) Rhythmic spacing and Storey A. W., Edward D. H. D. & Gazey P. (1991) Surber and origin of pools and riffles. Geol.Soc. Amer. Bull. 89, 7~:3-30. kick sampling: A comparison for the assessmentof macro- Kennedy G. J. A. & StrangeC. D. (1981) Efficiency of electric invertebrate community structure in streams of south- fishing for salmonids in relation to river width. Fish../oranag. western Ausrralia. Hydrobiol. 211, 111-21. U, 55-60. Underwood A. J. (1997) Experimentsin Ecology: their logical Koehn J. D. & McKenzie J. A. (1985) C:omparisonof e/eccrofisher description and interpretation using analysis of variance. efficiencies.Technical Report Series no. 27, Arthur Rylah Cambridge University Press, Cambridge, U.K. Institute for Environmental Research,Heidelberg. VadasR. L. & Orth D. J. (1993) A new technique for estimating Larimore L. M. (1961) Fish population and electrolishing the abundance and habitat use of stream fishes. J Freshw. success in a warm water stream. J. Wild!. Mana~. 25, Ecol. 8,305-17. 1-12. Wiley M. L. & Tsai C. F. (1983) The relative efficiencies of Maher W. A., Cullen P. W. & Norris R. H. (1994) Framework electrofishing vs seinesin Piedmont Streams of Maryland. for designing sampling programs. Env. Mon. & Assln. 30, N Amer. J Fish. Manag. 3, 243-:-53. 139-62. Zalewski M. (1983) The influence of fish community structure Mahon R. (1980) Accuracy of catch-effort methods for estimat- on the efficiency of electrofishing. Fish. Manag. 14, 177-86. ing fish density and biomass in streams. Env. Biol. Fi~hes5, Zalewski M. & Cowx I. G. (1989) Factors affecting the 343-60. efficiency of electrofishing. In: Fishing with Electricity- Mahon R. & Balon E. K. (1980) Quantitative collection of Application in FreshwaterFisheries Management(ed. I. G. fishes from small fast-flowing streams.Arch. fUr Hy~'robiol. Cowx & P. Lamarque) p 89-110. Fishing New Books, 90,324-9. Oxford.

View publication stats