Control of freshwater fish and crayfish community structure in Taranaki, New Zealand: dams, diadromy or habitat structure?

Transcription

1 Freshwater Biology (2001) 46, APPLIED ISSUES Control of freshwater fish and crayfish community structure in Taranaki, New Zealand: dams, diadromy or habitat structure? MICHAEL K. JOY and RUSSELL G. DEATH Institute of Natural Resources-Ecology, Massey University, Palmerston North, New Zealand SUMMARY 1. Freshwater fish and crayfish communities were surveyed along elevational gradients in streams radiating from Mount Taranaki, New Zealand. Six of the 38 streams surveyed had dams or weirs and 32 of the 85 sites were above these barriers. 2. Of the 15 fish and one crayfish species captured, 14 were native. The number of species declined with increasing elevation and distance from the sea. Species richness was also lower above dams even when effects of elevation and distance from the sea were accounted for. 3. Linear regression models using sites without dams were constructed to predict the effect of elevation on fish and crayfish community structure and to allow the effect of dams to be evaluated. The number of species predicted by these models was consistently higher than the number of species found at the above-dam sites. 4. Four fish community groups were classified using two-way indicator species analysis (TWINSPAN). One high elevation group of sites consisted of short steep streams on the west side of the mountain and a second contained longer, lower gradient streams on the east side. The other two groups (3 and 4) consisted of mid-elevation sites and low-elevation sites, respectively. 5. Discriminant analysis was used to predict biotic group membership using the environmental data. Overall, 80% of sites were classified correctly. Correlation of environmental variables with axis scores in the canonical variate analysis revealed that distance from the sea, site elevation and presence of dams were most strongly associated with fish distribution patterns. Keywords: dams, diadromy, freshwater fishes, multivariate analysis, Taranaki Introduction Correspondence: Mike Joy, Institute of Natural Resources-Ecology, Massey University, Private Bag , Palmerston North, New Zealand. mikejoy@clear.net.nz A number of factors are implicated in controlling the distribution of stream fish species and the regulation of fish community structure worldwide. Biotic interactions have been proposed by some (e.g. Werner & Gilliam, 1984; Ross, 1986; Gilliam, Fraser & Alkins- Koo, 1993) to be major influences on fish community organization, while others (e.g. Gorman & Karr, 1978; Angermeier & Schlosser, 1989; Rahel & Hubert, 1991) have highlighted abiotic factors. Biotic and abiotic factors have also been combined along upstream downstream gradients within the river continuum concept (RCC) (Cummins, 1979; Vannote et al., 1980; Cushing et al., 1983; Minshall et al., 1985; Schlosser, 2001 Blackwell Science Ltd 417

2 418 M.K. Joy and R.G. Death 1990) and in integrated synthetic models (Zalewsky & Nayman, 1985; Schlosser, 1987; Zalewsky et al., 1990). In contrast to deterministic views of community organization, stochastic events principally related to flow variability may also influence fish communities (Moyle & Li, 1979; Grossman, Moyle & Whitaker, 1982; Grossman et al., 1985; Grossman, Dowd & Crawford, 1990; Ross, Mathews & Echelle, 1985), although assemblages may be stable even under extreme flow variation (Matthews, 1986; Meffe & Berra, 1988; Meador & Matthews, 1992). Regardless of the processes that influence stream fish communities, an increase in diversity is almost universally found when moving in a downstream direction from the headwaters. In North America and Europe, this pattern has generally been attributed to increased habitat diversity, habitat volume and more stable conditions in the lower reaches of a waterway (Sheldon, 1968; Lotrich, 1973; Rahel & Hubert, 1991). In New Zealand, however, a downstream increase in species richness has been attributed to the fact that the fauna is dominated by diadromous species (Hayes, Leathwick & Hatchet, 1989; McDowall, 1993, 1996, 1998; Jowett & Richardson, 1996). No other fish fauna with the same or higher number of species has such a high proportion of diadromous species (Mc- Dowall, 1988, 1995, 1997). Consequently, establishing links between species diversity and such factors as land use, habitat and/or biotic relationships in New Zealand may be frustrated by the overwhelming and variable influence of diadromy (Hayes et al., 1989; Hanchett, 1990; Jowett, Richardson & McDowall, 1996; Jowett & Richardson, 1996). Diadromous fish have distributions that reflect diffusion from a source (the ocean) to the limit of their upstream penetration, governed by the interaction of variables such as swimming ability, barrier effects, distance upstream and instinctive migratory drive (Smogor, Angermeier & Gaylord, 1995). Caution is required when associating habitat characteristics with fish distribution in New Zealand, because the observed distribution may not be the result of the proximal habitat but rather of access to that habitat. A fauna with such a high proportion of diadromous species has an enormous potential to be affected by barriers to migration. Dams, regardless of their purpose, are likely to impact on fish habitat in a number of ways (Sale, 1985; Bain, Finn & Booke, 1988; Kanehl, Lyons & Nelson, 1997). However, the most obvious negative effect is likely to be impediment to migration (Davis & Teirney, 1987; Jowett, 1987; Lusk, 1995; Reyes-Gavilan et al., 1996). Our initial objective in this study was to characterize fish and crayfish community structure over elevational gradients in Taranaki Ring Plain streams. To accomplish this, the sites were organized into groups based on their faunal composition and then the habitat features associated with these groupings were investigated. While many studies have illustrated the importance of distance/elevation in controlling New Zealand fish distribution, the important effects of barriers have generally been neglected. Our second objective was, therefore, to evaluate the influence of migratory barriers on fish and crayfish distribution. To accomplish this, we modelled the influence of elevation/distance on the communities using streams that had free access and then used this model to evaluate the impact of dams with and without fish passes using data from sites above those barriers. Methods Study area and sampling methods Mount Taranaki is an andesitic volcano (2158 m above sea level (a.s.l.)), which is symmetrical in shape with intact native forest down to an altitude of about 400 m a.s.l. These factors result in streams having similar gradients and all emerge at similar altitudes from native forest into intensively farmed land where the predominant land use is dairy farming. There are only two non-migratory species found on the mountain Ring Plain, possibly because of its relatively recent volcanic origin. Fish and crayfish were collected from 85 sites on 38 streams draining Mount Taranaki from December 1997 to March Thirtytwo sites are above man-made barriers, nine of which are above dams with fish passes, although not all passes are equally effective (Joy & Death, 2000). Fish sampling Two sampling methods, electrofishing (44 sites) and night spotlighting (41 sites), were used. Spotlighting was used in streams over a range of altitudes (mean 283 m a.s.l., standard error (SE) 20 m a.s.l.) where water clarity enabled fish to be seen in all habitats. The majority of New Zealand freshwater fish are

3 Determinants of freshwater fish community structure 419 nocturnal and benthic; thus, spotlighting is very effective where water clarity enables all habitats to be observed. Spotlighting is more efficient than electrofishing in upper elevation streams where large substrate provides daytime refuge in gaps between boulders in which fish can escape electrofishing. Electrofishing was used at sites over a similar range of elevations (mean 269 m a.s.l., SE 25 m a.s.l.), where the size, depth or lack of water clarity restricted the use of spotlighting. Fish were stunned using a battery-powered pulsed-dc backpack electrofisher (EFM300; NIWA Instrument Systems, Christchurch, New Zealand), and collected in handheld stop or dip nets. All habitats (i.e. pool, run and riffle) were sampled at each site. Fish stunned by the electrofisher and those caught in nets during spotlighting were identified to species, counted and returned to the water. The average reach area surveyed (width length) was 175 m 2 (SE 7.34 m 2 ), with a minimum of 50 m 2 (four sites) and a maximum of 400 m 2 (one site). Fish densities at each site were calculated by dividing the number of fish by the area fished and standardized to the number of fish 100 m 2. Three measures of fish diversity were calculated for each site. They were: 1. Species richness (S); 2. H = pilnpishannon Weaver index (H ): H = p i ln p i, where p i is the proportion of individuals in the ith species (Shannon & Weaver, 1949); and 3. evenness (E): H /ln S (Pielou, 1969). Environmental measures At each site, the percentage of riparian vegetation (native forest, exotic forest, grass/tussock and scrub) was visually assessed. The percentage of catchment land use (Native forest, exotic forest, farming and scrub), altitude and distance from the sea were estimated from NZMS 1:50000 topographic maps. Over the length of stream fished, the mean and maximum water depth, mean width, percentage overhead shade, substrate composition and fish cover were subjectively assessed. Fish cover was assessed as the percentage of undercut banks, weed, bank vegetation and instream debris. Substrate composition was visually assessed as the percentage of the substrate composed of mud ( 1 mm), sand (1 2 mm) fine gravel (2 20 mm), coarse gravel (20 60 mm), cobble ( mm), boulder ( 260 mm) or bedrock. The percentage of backwater, pool, run, riffle or rapid was estimated at each reach surveyed. Riffles were areas of fast, shallow water with a broken surface appearance; pools were areas of slow, deep water with a smooth surface appearance, whereas runs were intermediate in character. Rapids were areas of fast cascading deep water. Data analysis Data were analysed using the regression and Spearman s rank correlation procedures of SAS (1996). Spearman s rank correlation was used to identify relationships between fish density or diversity and environmental measures. To explore differences in species richness at sites above and below dams (with and without fish passes), analysis of covariance (AN- COVA) was employed using elevation as a covariate (GLM procedure of SAS, 1996). Habitat data from sites without migratory barriers were included in a linear regression analysis to yield a predictive equation for species richness. Using this equation, species richness was predicted for sites above barriers. The residuals from the regression analysis were compared for sites with and without barriers using analysis of variance (ANOVA), with the null hypothesis of barriers having no effect on species richness. Comparisons of mean density and diversity measures for sites above and below dams were made with the nonparametric Kruskal Wallis test (NPAR1WAY procedure of SAS, 1996). The relationships between fish communities and the environmental characteristics found at sites were examined by identifying groups of sites that contained similar fish assemblages using TWINSPAN (PC-ORD; McCune & Mefford, 1997). The TWINSPAN analysis (to two levels) was achieved using five pseudospecies cut levels: 0, 2, 5, 10 and 20 fish 100 m 2. Relationships between environmental measures and groups classified by TWINSPAN were examined using linear discriminant analysis. The models were then used to test the performance of the environmental variables by separating sites into their respective TWINSPAN groups (classification success) and by considering the posterior probability error

4 420 M.K. Joy and R.G. Death rate estimates of the classification results (DISCRIM CROSSVALIDATE procedure of SAS, 1996). To visualize and evaluate the importance of each of the environmental variables in discriminating between the TWINSPAN groups, canonical discriminant analysis was then performed (CANDISC procedure of SAS, 1996). The coefficients of canonical variables can be biased by correlated variables (Williams, 1983). To mitigate this effect, the canonical variables were also interpreted by examining their correlation with each of the environmental variables (CORR procedure of SAS, 1996). Differences between mean diversity and environmental measures between TWINSPAN groups were examined with the Kruskal Wallis test (NPAR1WAY procedure of SAS, 1996) and medians were compared using Tukey s multiple range tests, with the significance level set at 5%. To test for differences between fish communities with and without migratory barriers and between TWINSPAN groups, the multiresponse permutation procedure (MRPP) was employed (PC-ORD; McCune & Mefford, 1997) using Euclidean distance measures. MRPP is a non-parametric procedure for testing the hypothesis of no differences between two or more groups and has the advantage over ANOVA that it is not reliant on multivariate normality and/or homogeneity of variances (Berry, Kvamme & Mielke, 1983). Results Species composition Fifteen fish and one crustacean species were captured from 85 sites around Mount Taranaki; 13 of the fish species were native (common names are used throughout; for scientific names and familial associations, see Table 1). The native crustacean was the large decapod Paranephrops planifrons (koura). The density of fish captured ranged between 1 and 85 fish 100 m 2, with a mean of 24 fish 100 m 2. The longfin eel and redfin bully were the most widely distributed, occurring at 93 and 40% of the sites, respectively (Table 1). The longfin eel made up 32% of the total catch, followed by the redfin bully (19.8%). The next most widespread and abundant species were Table 1 Species composition of fish and crayfish (latin and common names shown) captured from 85 sites sampled over summer in Taranaki Ring Plain streams Occurrence (sites) Species composition (fish) Family Scientific name Common name No. of sites % of total No. of fish % of total Anguillidae Anguilla dieffenbachii Gray Longfin eel Parastacidae Paranephrops planifrons Koura Eleotridae Gobiomorphus huttoni Ogilby Redfin bully Salmonidae Salmo trutta Linnaeus Brown trout* Galaxiidae Galaxias brevipinnis Günther Koaro Anguillidae Anguilla australis Richardson Shortfin eel Galaxiidae Galaxias postvectis Clarke Shortjaw kokopu Eleotridae Gobiomorphus basalis Gray Crans bully Galaxiidae Galaxias fasciatus Gray Banded kokopu Galaxiidae Galaxias maculatus Jenyns Inanga Eleotridae Gobiomorphus cotidianus Common bully McDowall Mugiloididae Cheimarrichthys fosteri Haast Torrentfish Mugilidae Aldrichetta forsteri Valenciennes Yellow eyed Retropinnidae Retropinna retropinna Richardson mullet Common smelt Percidae Perca fluviatilis Linnaeus Perch* Geotriidae Geotria australis Gray Lamprey Total * Denotes exotic species. Denotes non migratory species.

5 Determinants of freshwater fish community structure 421 Table 2 Species composition and diversity of fish and crayfish above and below dams sampled over summer on the Taranaki Ring Plain No dam (n=53) Dam (n=32) Species No. of fish No of sites % of total no. of fish No of fish No of sites % of total no. of fish Longfin eel Redfin bully Brown trout Shortjaw kokopu Koaro Inanga Shortfin eel Crans bully Koura Mullet Banded kokopu Torrentfish Common bully Lamprey Common smelt Perch Diversity and density Richness*** Density Evenness Diversity* Mean Shannon diversity and density measures ( SE) and Kruskal Wallis values indicating differences between sites above dams and those with free access. * P 0.05; *** P brown trout, koaro and shortfin eel. The remaining ten species shortjaw kokopu, Crans bully, inanga, banded kokopu, common bully, torrentfish, yellow eyed mullet, common smelt, perch and lamprey were uncommon and made up 14% of the total catch (Table 1). Species richness was inversely related to reach elevation (r s = 0.33, P=0.002) and distance from the sea (r s = 0.29, P=0.007). Fish density showed a similar pattern (elevation: r s = 0.42, P 0.001; distance from the sea: r s = 0.29, P=0.006). There was a reduction in Shannon diversity with reach elevation (r s = 0.24, P=0.03), but not with distance from the sea (r s = 0.14, P=0.21). Species evenness showed no relationship with elevation or distance inland (r s = 0.09, P=0.39; r s = 0.06, P=0.58, respectively). Migratory barriers Thirty-two sites had potential migratory barriers (dams) and 53 sites had free access (no natural barriers are known). There were differences in the fish assemblages of the two groups (Table 2); inanga, yellow eyed mullet, banded kokopu, common bully, lamprey, common smelt and perch were all absent from the sites above barriers. Of the 32 sites above dams, nine were potentially accessible by fish passes. Species richness was lower at sites above barriers. both when elevation was employed as a covariate (F 3, 81 =3.70, P=0.03) and when distance from the sea was a covariate (F 3, 81 =3.44, P=0.03). However, the sites above dams with fish passes were not significantly different from either sites with free access or those without fish passes (Tukey s test). The mean number of species at sites with free access was 3.7 (n=52); for sites above dams with fish passes, the mean was 3.2 (n=9); and for the sites above dams without fish passes, the mean was 2.6 (n=24). The effect of barriers (with or without fish passes) on diversity was not significant when elevation or distance from the sea was employed as a covariate

6 422 M.K. Joy and R.G. Death (F 3, 83 =2.84, P=0.06; and F 3, 83 =2.62, P=0.08, respectively). MRPPs showed a significant difference in fish community composition at sites with and without migratory barriers (T= 7.43, P 0.001). Elevation and distance from the sea were used to construct linear regression models for species richness using the 53 sites without migratory barriers. The model for inland distance (F 1, 50 =9.63, P 0.003) yielded the following equation: species richness= (log distance inland). The model for altitude (which is highly correlated with distance inland) (F 2, 50 =10.77, P 0.002) yielded the equation: species richness= (log altitude). The 32 sites with Fig. 1 Box plots of species richness residuals (observed species richness minus predicted for elevation) for sites with free access, and those above dams with and without fish passes. (b) Box plots of species richness residuals (observed species richness minus that predicted for distance from the sea) for sites with free access, and those above dams with and without fish passes. Box contains interquartile range; circles denote 5th and 95th percentiles; centreline denotes median; and whiskers extend to 1.5 interquartile range). barriers were examined with the predictive equations and the residuals were compared with those from sites having no barriers. The majority of sites with barriers had fewer species than predicted from the above equations (distance inland F 2, 82 =4.06, P= 0.02; altitude F 2, 82 =4.56, P=0.01; Figs 1a and b), indicating that species richness was reduced by the dams. There was no significant difference between the residuals for sites above dams with or without a fish pass. TWINSPAN classification and fish assemblages Four fish assemblages were identified from the TWINSPAN analysis at level two. The 18 sites in group 1 were dominated by shortjaw kokopu, koaro, koura, redfin bully, banded kokopu and longfin eel; trout were present but rare (Table 3). The 50 sites making up group 2 were dominated by longfin eels, koura, and Crans and redfin bullies. Group 3 contained 12 sites and had a simple fish assemblage dominated by redfin bullies, longfin eels and, to a lesser extent, inanga and trout. Banded kokopu and inanga were more common in this group than any other. The most abundant taxa at the five sites in group 4 were the redfin bully followed by short and longfin eels, and then yellow eyed mullet. The classification of the sites from the species associations revealed that mean elevation declined from group 1 to 4 (Table 4). In parallel with this pattern, the percentage of native riparian vegetation declined, and stream size and percentage of grass riparian vegetation increased. MRPPs revealed a difference between fish communities in the TWINSPAN groups (T= 12.38, P 0.001); using all possible pairwise and three-group combinations, all were different (P ) except groups 3 and 4 (T= 1.39, P=0.093). The environmental variables associated with the four groups were also significantly different (T= 12.50, P 0.001). Group 1 sites were at the highest altitude, and had the most native riparian vegetation and overhead cover. Group 2 sites occurred at lower elevations but mean distance from the sea was greater, the sites were on average wider and deeper, and had a higher percentage of grass riparian vegetation (Table 4). Group 3 sites were on average deeper, wider and had finer substrates, and had less native but more grass riparian vegetation, and less overhead

7 Determinants of freshwater fish community structure 423 Table 3 Percentage occurrence (number of individuals for each species/total individuals in group) of fish species in each of four TWINSPAN groups for 85 sites on the Taranaki Ring Plain sampled over summer TWINSPAN group Species Group 1 (n=18) Group 2 (n=50) Group 3 (n=12) Group 4 (n=5) Mullet 12.9 Redfin bully Common smelt 0.3 Longfin eel Koura Trout Crans bully 11.2 Shortjaw kokopu Shortfin eel Perch 0.3 Torrentfish Inanga Banded kokopu Common bully Lamprey 0.6 Koaro cover than groups 1 and 2. Group 4 sites were all 6-m a.s.l. with no overhead cover and a high percentage of grass riparian vegetation. Diversity and density of fish also varied between the TWINSPAN groups. Groups 1 and 2 were similar and only group 4 had significantly higher density and diversity than the other three (Table 4). Discriminant analysis Site elevation was the most important environmental factor discriminating between TWINSPAN groups (Table 5). A discriminant model with this factor alone correctly predicted group membership of 66% of the sites. Adding distance inland increased the rate of correct predictions to 73%. Including all 11 variables from the likelihood ratio criterion analysis (P 0.05) increased the predictive ability of the model to 80% (Table 6). Correlations between the discriminant factors and the environmental variables revealed that the first canonical variate, accounting for 85% of the separation between the groups, was positively correlated with altitude, distance inland, native riparian vegetation and overhead cover (Table 7). Canonical variate 1 (Can 1) was negatively correlated with width, depth and grass riparian vegetation. The second canonical variate (Can 2), accounting for 11% of the separation between the groups, was positively correlated with distance inland, percentage pool, percentage sand fine gravel substrate and percentage grass riparian vegetation, and was negatively correlated with percentage of rapid. To illustrate the separation of the groups by the discriminant analysis, the sites are shown in discriminant space (Fig. 2). Can 1 shows the separation of groups 1, 3 and 4. Can 2 separates groups 1 and 2. The group 4 sites, which are negative on Can 1, are low elevation, wide, deep sites with a high proportion of grass riparian vegetation. At the opposite end of Can 1 are the group 1 sites which are the high elevation, shallow, narrow stream sites with a high proportion of native riparian cover and overhead shade. Thus, the influences of the environmental variables on Can 1 moving from negative to positive are increasing elevation, distance from the sea, native riparian vegetation and overhead shade. Moving in a positive direction along Can 2, elevation and percentage rapid are decreasing, and distance inland, percentage pool and percentage grass vegetation are increasing. Discussion As with many other New Zealand West Coast waterways, the Taranaki Ring Plain fish fauna was dominated by the longfin eel and redfin bully (Main,

8 424 M.K. Joy and R.G. Death Nicholl & Eldon, 1985; Taylor & Main, 1987; Hayes et al., 1989; Hanchett, 1990). The 13 native fish species found in this survey exceeded that found in similar surveys of North Island waterways, which ranged between eight and 11 (Rowe, 1981; Hicks & Watson, 1983; Strickland, 1985; Cudby & Strickland, 1986; Hayes et al., 1989; Hanchett, 1990). Species richness and diversity increased along downstream gradients of depth, width, decreasing native vegetation, increasing farmland, and decreasing distance from the sea and elevation. Site elevation, distance from the sea and the presence of a migratory barrier were the variables most highly correlated with species richness and diversity and, although they are intercorrelated, their association with fish communities was at least partially independent. This was illustrated by the different communities found at similar altitudes in the short run streams on the western side of the mountain compared with the longer eastern waterways. Similar conclusions to these have been reached in a number of other studies and diadromy has been proposed by Hanchett (1990) and McDowall (1990, 1993) as the most important factor influencing New Zealand freshwater communities. Table 4 Mean site characteristics and community descriptors ( 1SE) in each of four TWINSPAN groups for 85 sites on the Taranaki Ring Plain over the summer of TWINSPAN groups Variable and Duncan s test Group 1 (n=18) Group 2 (n=50) Group 3 (n=12) Group 4 (n=5) Altitude*** Inland*** Mean width* Mean depth* Reach fished % Pool % Run % Riffle % Rapid Substrate composition % Mud % Sand % Fine gravel % Coarse gravel % Cobble % Boulder % Bedrock Riparian vegetation % Native*** % Exotic % Grass*** % Scrub % Overhead cover*** No. of sites above barrier*** Community descriptor Richness* Density* Evenness Diversity* Significant differences between TWINSPAN groups from species abundance data using the non-parametric Kruskal Wallis test are also given (* P 0.05; ** P 0.01; *** P 0.001). Group numbers italicized are not significantly different at P 0.05 level using Tukey s multiple range test.

9 Determinants of freshwater fish community structure 425 Table 5 Ability of the 11 environmental variables to discriminate between the four TWINSPAN groups significant at the P 0.05% level as measured by partial correlation (R 2 ) and Wilks Lambda, and the likelihood ratio criterion (F) Environmental variable R 2 F statistic Altitude Distance from the sea % Pool Mean Width % Sand % Grass riparian vegetation Mean depth % Riffle % Native riparian vegetation % Mud substrate % Bedrock Probabilty F In Taranaki, overall water quality as measured by chemical analysis (Taranaki Catchment Commission 1984) and macroinvertebrate indices (Taranaki Regional Council 1998) improves with increasing elevation, as does the amount of overhead cover and native vegetation. In contrast, the number of native fish species declines with elevation independent of water quality, mainly because of the influence of diadromy (McDowall, 1996, 1998; Joy & Death, 2000). The increase in native vegetation and water quality with elevation reflects the decreasing human impact on catchments, while the reduction in fish species richness illustrates the influence of diadromy in structuring fish communities. Differentiating between the two influences, however, may not be possible. Potentially, the best quality water is unavailable to the majority of the species. At lower elevations, where Table 7 Coefficients for correlation between environmental variables with the first two axes of a canonical discriminant analysis Correlations Environmental variable Can 1 Can 2 Altitude 0.858*** Inland 0.599*** 0.488*** Mean width 0.309** Mean depth 0.231* Reach fished % Pool ** % Run % Riffle % Rapid * Substrate composition % Mud % Sand ** % Fine gravel * % Coarse gravel % Cobble % Boulder % Bedrock Riparian vegetation % Native 0.412*** % Exotic % Grass 0.413*** 0.225* % Scrub % Overhead cover 0.497*** Migratory barrier Environmental measures made at 85 sites surveyed over the summer of on the Taranaki Ring Plain. * P ** P *** P the largest potential species pool is available, water quality is generally at its lowest. Access may be the most important habitat attribute for many New Table 6 The number and percentage of sites classified into each of the four TWINSPAN groups by linear discriminant analysis using the concurrently recorded environmental measures listed in Table 5 Predicted group membership (to group) Group (from group) No. of sites % of sites correctly predicted Total

10 426 M.K. Joy and R.G. Death Zealand freshwater fish by allowing passage to areas with suitable proximal habitat characteristics. The downstream increase in species richness has in North American and European studies been attributed to increased habitat diversity, volume and more stable conditions (Gorman & Karr, 1978; Schlosser, 1982; Rahel & Hubert, 1991). The RCC (Cummins, 1979; Vannote et al., 1980; Cushing et al., 1983; Minshall et al., 1985; Schlosser, 1990) also predicts increasing species richness in an upstream downstream gradient, driven by changes in resource availability and habitat characteristics. The patterns observed in this study could reflect those predictions. However, they may equally reflect the variability in penetrative ability of the predominantly diadromous species. If the latter is the case, the relationship between fish community structure and habitat characteristics is confounded because of the open nature of these communities. McDowall (1995) pointed out that open fish communities (i.e. communities dominated by diadromous species) are structured by the influx of species from outside the community (from the sea). In contrast, closed communities are structured by proximate biotic and abiotic influences. Thus, when considering habitat quality for New Zealand freshwater fish, migratory access must be considered above all other factors. Migratory barriers Given the importance of upstream access for the New Zealand fish fauna, dams are one of the most prominent devices that act as potential barriers to upstream Fig. 2 Position of the 85 Taranaki Ring Plain sites in discriminant space obtained by canonical variate analysis on the 11 environmental variables listed in Table 5. and downstream passage. Species richness and diversity were lower at sites above dams and the difference remained significant even when the effect of elevation was removed. However, because the communities are dominated by diadromy, the relative impact of a given dam on the fish fauna will depend on the altitude and distance from the sea of the dam. The native fish species of New Zealand vary widely in their migratory abilities, migratory drive and ability to negotiate barriers (McDowall, 1990, 1993; Joy & Death, 2000) and, consequently, the dams vary in their barrier function. Many species are limited to low elevations; for example, torrentfish, yellow eyed mullet, common smelt, common bully and lamprey did not occur above 100 m elevation in this study. As most of the dams are above this elevation, their effect on these species would be minimal. However, dams at low elevation or on low gradient waterways have the potential to have impacts that are more serious on fish communities than if they were at higher altitudes. Dams act to not only truncate distribution but also to reduce the density of species that are able to negotiate passage. Biotic interactions are also considered important in structuring fish communities (Werner & Gilliam, 1984; Ross, 1986; Gilliam et al., 1993). In New Zealand, predation has been recorded by eels on banded kokopu (Mitchell & Penlington, 1982), bullies, smelt and small trout (McDowall, 1990). Large eels were found at most sites in this study, and at upland sites, where large galaxiids were found, juvenile galaxiids were rare or absent. Predation may, therefore, be affecting species distributions. Trout were similarly absent from many upland sites where koaro, shortjaw and banded kokopu were found at high densities. Limitation of native fish abundance by predation from trout (Townsend & Crowl, 1991) or through competition for food (Cadwallader, 1975; Sagar & Eldon, 1983) may be having an effect in Taranaki. In conclusion, while freshwater fish species richness on the Taranaki Ring Plain is higher than that found in other North Island rivers, much of the fauna has a relatively restricted distribution. The region has a large number of dams and they have a negative effect on fish passage. Elevation, distance from the sea and whether or not the site is above a dam were the variables that had the biggest impact on fish community organization. The overriding influence on the distribution of fish was, however, the prevalence

11 Determinants of freshwater fish community structure 427 of diadromy, with species varying in their ability to penetrate inland both naturally and because of the impediments to migration. As a number of variables covary with elevation and distance from the sea, it is difficult to isolate influential environmental effects on community structure over and above the effect of diadromy. The prevalence of diadromy, however, means that where dams are present, they become the most important influence on community structure. Acknowledgments We thank Dean Caskey (Department of Conservation), Hamish McWilliam (Taranaki Regional Council) and Allison Hewitt for assistance in the field. Thanks to Rosemary Miller for her persistence in setting up the original project. We also thank Ian Henderson, Gerry Gloss and an anonymous referee for their comments, which greatly improved this paper. The research was funded by Department of Conservation grant 2355 to RGD. References Angermeier P.L. & Schlosser I.J. (1989) Species area relationships for stream fishes. Ecology, 70, Bain M.B., Finn J.T. & Booke H.E. (1988) Streamflow regulation and fish community structure. Ecology, 69, Berry K.J., Kvamme K.L. & Mielke P.W.J. (1983) Improvements in the permutation test for the spatial analysis of the distribution of artifacts into classes. American Antiquity, 48, Cadwallader P.L. (1975) Feeding relationships of galaxiids, bully, eels and trout in a New Zealand river. Australian Journal of Marine and Freshwater Research, 26, Cudby E.J. & Strickland R.R. (1986) The Manganui o te Ao River Fishery.. Fisheries Environmental Report No. 14. Ministry of Agriculture and Fisheries. Cummins K.W. (1979) The Natural Stream Ecosystem. Plenum Press, New York. Cushing C.E., McItire C.D., Cummins K.W., Minshall G.W., Petersen R.C., Sedell R.C. & Vannote R.L. (1983) Relationships among chemical, physical and biologic indices along a river continua based on multivariate analyses. Archiv für Hydrobiologie, 98, Davis S.F. & Teirney L.D. (1987) Fish and fisheries. In: Aquatic Biology and Hydroelectric Power Development in New Zealand (ed. P.R. Henriques), pp Oxford University Press, Auckland. Gilliam J.F., Fraser D.F. & Alkins-Koo M. (1993) Structure of a tropical stream fish community: a role for biotic interactions. Ecology, 74, Gorman O.T. & Karr J.R. (1978) Habitat structure and stream fish communities. Ecology, 59, Grossman G.D., Moyle P.B. & Whitaker J.O. (1982) Stochastisity in structured and functional characteristics of an Indiana stream fish assemblage: a test of community theory. American Naturalist, 120, Grossman G.D., Freeman M.C., Moyle P.B. & Whitaker J.O. (1985) Stochasticity and assemblage organisation in an Indiana stream fish assemblage. American Naturalist, 126, Grossman G.D., Dowd J.F. & Crawford M. (1990) Assemblage stability in stream fishes: a review. Environmental Management, 5, Hanchett S.M. (1990) Effect of land use on the distribution and abundance of native fish in tributaries of the Waikato River in the Hakarimata Range, North Island, New Zealand. New Zealand Journal of Marine and Freshwater Research, 24, Hayes J.W., Leathwick J.R. & Hatchet S.M. (1989) Fish distribution patterns and their association with environmental factors in the Mokau River catchment, New Zealand. New Zealand Journal of Marine and Freshwater Research, 23, Hicks B.J. & Watson N.R.N. (1983) Quinnant salmon (Oncorhynchs tshawytscha) spawning in the Rangitikei River. New Zealand Journal of Marine and Freshwater Research, 17, Jowett I.G. (1987) Fish passage, control devices and spawning channels. In: Aquatic Biology and Hydroelectric Power Development in New Zealand (ed. P.R. Henriques), 280. Oxford University Press, Auckland. Jowett I.G. & Richardson J. (1996) Distribution and abundance of freshwater fish in New Zealand rivers. New Zealand Journal of Marine and Freshwater Research, 30, Jowett I.G., Richardson J. & McDowall R.M. (1996) Relative effects of in stream habitat and land use on fish distribution and abundance in tributaries of the Grey River, New Zealand. New Zealand Journal of Marine and Freshwater Research, 30,

12 428 M.K. Joy and R.G. Death Kanehl P.D., Lyons J. & Nelson J.E. (1997) Changes in the habitat and fish communities of the Milwaukee River, following removal of the Woolen Mills Dam. North American Journal of Fisheries Management, 17, Joy M.K. & Death R.G. (2000) Development and application of a predictive model of riverine fish community assemblage in the Taranaki region. New Zealand Journal of Marine and Freshwater Research, 34, Lotrich V.A. (1973) Growth, production and community composition of fishes inhabiting a first, second and third order stream of Eastern Kentucky. Ecological Monographs, 43, Lusk S. (1995) Influence if valley dams on the changes in fish communities inhabiting streams in the Dyje River drainage area. Folia Zoologica, 44, Main M.R., Nicholl G.J. & Eldon G.A. (1985) Distribution of freshwater fishes in the Cook River to Paringa River area, South Westland, New Zealand. Fisheries Environmental Report No. 60. Ministry of Agriculture and Fisheries. Matthews W.J. (1986) Fish faunal structure in an Ozark stream: stability, persistence and a catastrophic flood. Copeia, 1986, McCune B. & Mefford M.J. (1997) Multivariate Analysis of Ecological Data. MJM Software, Gleneden Beach, OR. McDowall R.M. (1988) Diadromy in Fishes: Migrations between Marine and Freshwater Environments. Croom Helm, London. McDowall R.M. (1990) New Zealand Freshwater Fishes: A Natural History and Guide. Heinemann Reed, Auckland. McDowall R.M. (1993) Implications of diadromy for the structuring and modeling of riverine fish communities in New Zealand. New Zealand Journal of Marine and Freshwater Research, 27, McDowall R.M. (1995) Diadromy and the assembly and restoration of riverine fish communities: a downstream view. Canadian Journal of Fisheries and Aquatic Sciences, 53, McDowall R.M. (1996) Biodiversity in New Zealand fishes, and the role of freshwater fishes as indicators of environmental health in New Zealand fresh waters.. Consultancy Report No. MFE Ministry for the Environment & the Department of Conservation. McDowall R.M. (1997) Is there such a thing as amphidromy? Micronesia, 30, McDowall R.M. (1998) Fighting the flow: downstream upstream linkages in the ecology of diadromous fish faunas in West Coast New Zealand rivers. Freshwater Biology, 40, Meador M.R. & Matthews W.J. (1992) Spatial and temporal patterns in fish assemblage structure of an intermittent Texas stream. American Midland Naturalist, 127, Meffe G.K. & Berra T.M. (1988) Temporal characteristics of fish assemblage structure in an Ohio stream. Copeia, 1988, Minshall G.W., Cummins K.W., Petersen R.C., Cushing C.E., Bruns D.A., Sedell J.R. & Vannote R.L. (1985) Developments of stream ecosystem theory. Canadian Journal of Fisheries and Aquatic Sciences, 42, Mitchell C.P. & Penlington B.P. (1982) Spawning of Galaxias fasciatus Gray (Salmoniformes: Galaxiidae). New Zealand Journal of Marine and Freshwater Research, 16, Moyle P.B. & Li H.W. (1979) Community ecology and predator prey relationships in warmwater streams. In: Predator Prey Systems in Fisheries Management (ed. H.W. Clepper), pp Sport Fishing Institute, Washington. Pielou E.C. (1969) An Introduction to Mathematical Ecology. John Wiley and Sons, New York. Rahel F.J. & Hubert W.A. (1991) Fish assemblages and habitat gradients in Rocky Mountain Great Plains stream: biotic zonation and additive patterns of community change. Transactions of the American Fisheries Society, 120, Reyes-Gavilan F.G., Garrido R., Nicieza A.G., Toledo M.M. & Brana F. (1996) Fish community variation along physical gradients in short streams of Northern Spain and the disruptive effect of dams. Hydrobiologia, 321, Ross S.T. (1986) Resource partitioning in fish assemblages: a review of field studies. Copeia, 1986, Ross S.T., Mathews W.J. & Echelle A.A. (1985) Persistence of stream fish assemblages: effects of environmental change. American Naturalist, 126, Rowe D.K. (1981) Fisheries investigations of the Motu River. Fisheries Environmental Report, No. 11. New Zealand Ministry of Agriculture and Fisheries, 48 pp.

13 Determinants of freshwater fish community structure 429 Sagar P.M. & Eldon G.A. (1983) Food and feeding of small fish in the Rakaia River, New Zealand. New Zealand Journal of Marine and Freshwater Research, 17, Sale M.J. (1985) Aquatic ecosystem response to flow modification: an overview of the issues. Symposium on Small Hydropower and Fisheries. Environmental Sciences Division, Oak Ridge National Laboratory, Ramada Renaissance Hotel, Denver, CO. SAS (1996) SAS Users Guide: Statistics, Version 6.12, 584. SAS Institute Inc., Cary, NC. Schlosser I.J. (1982) Fish community structure and function along two habitat gradients in a headwater stream. Ecological Monographs, 52, Schlosser I.J. (1987) A conceptual framework for fish communities in small warmwater streams. In: Community and Evolutionary Ecology of North American Stream Fishes (eds W.J. Matthews & D.S. Heins), pp Oklahoma University Press, Norman, OK. Schlosser I.J. (1990) Environmental variation, life history attributes, and community structure in stream fishes: implications for environmental management and assessment. Environmental Management, 14, Shannon, C.E., & Weaver, W. (1949) The Mathematical Theory of Communication. University of Illinois Press, Urbana, IL. Sheldon A.L. (1968) Species diversity and longitudinal succession in stream fishes. Ecology, 49, Smogor R.A., Angermeier P.L. & Gaylord C.K. (1995) Distribution and abundance of American Eels in Virginia streams: tests of null models across spatial scales. Transactions of the American Fisheries Society, 124, Strickland R.R. (1985) Distribution and habitats of fishes in the Mohaka River. Fisheries Environmental Report No. 55. New Zealand Ministry of Agriculture and Fisheries. Taranaki Catchment Commission (1984) Water Quality: Taranaki Ring Plain Water Resources Survey. Taranaki Catchment Commission. Taranaki Regional Council (1998) Some statistics from the Taranaki Regional Council database (FWB) of fresh water macroinvertabrate surveys performed during the period from January 1980 to 31 December Technical Report No Taranaki Regional Council. Taylor M.J. & Main M.R. (1987) Distributions of freshwater fishes in the Whakapohai to Waita River area, South Westland. Fisheries Environmental Report No. 77. New Zealand Ministry of Agriculture and Fisheries. Townsend C.R. & Crowl T.A. (1991) Fragmented population structure in a native New Zealand fish: an effect of introduced brown trout? Oikos, 61, Vannote R.L., Minshall G.W., Cummins K.W., Sedell J.R. & Cushing C.E. (1980) The river continuum concept. Canadian Journal of Fisheries and Aquatic Sciences, 37, Werner E.E. & Gilliam J.F. (1984) The ontogenetic niche and species interactions in size-structured populations. Annual Reviews of Ecological Systems, 15, Williams B.K. (1983) Some observations on the use of discriminant analysis in ecology. Ecology, 64, Zalewsky M. & Nayman R. (1985) The Regulation of Riverine Fish Communities by a Continuum of Abiotic Biotic Factors. Butterworth Scientific, London. Zalewsky M., Frankiewicz P., Przbylski M., Banbura J. & Nowak M. (1990) Structure and dynamics of fish communities in temperate rivers in relation to abiotic biotic regulatory continuum concept. Polish Archive Hydrobiologia, 37, (Manuscript accepted 3 July 2000)