Freshwater Biology (2017) 62, 600–614

doi:10.1111/fwb.12889

Co-varying impacts of land use and non-native brown trout on fish communities in small streams , SCOTT A. WISSINGER*, BRANDON C. GOELLER*,2 AND LESLIE O. RIECK*,3 MARK A. KIRK*,1 *Biology and Environmental Science Departments, Allegheny College, Meadville, PA, U.S.A.

SUMMARY 1. Evaluating the biological integrity of stream ecosystems requires a clear understanding of biological responses to anthropogenic stressors. Co-variation of stressors with natural landscape gradients has been shown to complicate the ability of biological assessments to detect communitylevel responses to anthropogenic stressors. Similarly, the co-varying occurrence of non-native species would also likely confound the ability of biological assessments to accurately determine biological integrity, although these relationships have been less studied. 2. We compared fish communities in 99 wadeable tributaries (with catchments of 10–35 km2) of the upper Allegheny River watershed, Pennsylvania, U.S.A., to disentangle the effects of human-induced stressors associated with agricultural development, the presence of non-native brown trout (Salmo trutta) and background variation in community composition associated with natural landscape features. 3. Multivariate analyses using both taxonomic and trait-based data revealed that environmental gradients sorted fish species into different thermal communities (coolwater versus warmwater). Layered over the shift in thermal communities was the presence of large-bodied (piscivorous) brown trout, which along with agricultural land use, was a consistent determinant of community diversity and presence–absence of nine common native species. Small-bodied taxa (minnows, darters) were the most responsive to the presence of these large non-native predators. 4. Our results highlight how the presence of a non-native species can modify the interactions between natural and anthropogenic stressor gradients and therefore confound the ability to detect signals between land use stressors and local stream community composition. The complicated interaction between brown trout and other anthropogenic stressors documented herein is likely to be a widespread phenomenon given the global introductions of many salmonid species into systems with varying anthropogenic stressor gradients. Overall, understanding how non-native species influence stream community composition should lead to biological assessments with greater capabilities for detecting non-native effects on biological integrity relative to the more frequently evaluated land use stressors. Keywords: biological assessment, brown trout, non-native species, stream fishes, thermal regimes

Introduction Stream ecosystems are among the most imperilled systems globally with community-level threats from biodiversity loss (Xenopoulos & Lodge, 2006; Hawkins et al., 2015), biotic homogenisation (Gido, Schaefer & Falke,

2009; Rahel, 2010) and non-native species (Scott & Helfman, 1999; Mooney & Cleland, 2001; Gozlan et al., 2010). Perhaps the single greatest threat to stream ecosystems is the intensification of landscape-level development, the cumulative effects of which degrade in-stream conditions at multiple spatial scales (Fausch et al., 2002; Allan, 2004;

Correspondence: Mark A. Kirk, Biology and Environmental Science Departments, Allegheny College, Meadville, PA 16335 U.S.A. E-mail: [email protected] 1

Present address: Department of Zoology and Physiology, University of Wyoming, Laramie, WY, U.S.A.

2

Present address: Centre for Integrative Ecology, School of Biological Sciences, University of Canterbury, Christchurch, New Zealand.

3

Present address: School of Environment and Natural Resources, Ohio State University, Columbus, OH, U.S.A.

600

© 2017 John Wiley & Sons Ltd

Impacts of land use and trout on fish communities Stephenson & Morin, 2009). Anthropogenic alterations at the landscape level change the physical (damming, channelisation, habitat alterations), hydrological (water withdrawals, canals, changes in flow regime) and chemical (temperature, point and non-point source pollution) nature of streams, which in turn alters the composition and structure of stream communities (Argent & Carline, 2004; Diana, Allan & Infante, 2006; Niyogi et al., 2007; Infante & Allan, 2010). Concerns about the anthropogenic impacts on stream ecosystems have stimulated the development of biological assessment protocols, which are designed to evaluate the cumulative and interactive effects of human-induced stressors on the biological ‘health’ of streams (Karr, 1991; Hawkins, 2006). These protocols often rely on community-level responses of biological indicators (e.g. fish, invertebrates, algae) for monitoring the combined impacts of landscape- and local-scale disturbances (Fausch, Karr & Yant, 1984; Southerland et al., 2007; Marzin, Verdonschot & Pont, 2013; Johnson & Angeler, 2014). In particular, stream fish occupy a range of functional guilds (e.g. trophic, reproductive, habitat, life history) that respond to multiple environmental stressors (Angermeier, Smogor & Stauffer, 2000; Wang et al., 2008; Frimpong & Angermeier, 2009), thus providing a good indicator for the effects of human disturbance on biotic communities. However, developing efficient biological criteria can be difficult without sufficient knowledge of the background processes of community assembly, given the potential interactions between anthropogenic stressors and natural background variation (Allan, 2004; Heino et al., 2007). The ecological processes responsible for structuring freshwater fish communities operate at regional, landscape and local scales, all of which can be difficult to control (Jackson & Harvey, 1989; Jackson, Peres-Nato & Olden, 2001; Fausch et al., 2002; Grenouillet, Pont & Hẽrissẽ, 2004; Kautza & Sullivan, 2012). Anthropogenic stressors at the landscape level often co-vary with natural landscape gradients and thus complicate the ability to detect biological responses, particularly in regions with low-to-moderate impacts; a problem emphasised by Allan (2004). As a result, the effects of anthropogenic stressors on community composition can be masked by changes along natural gradients (Delong & Brusven, 1998; Heino et al., 2007; Wagenhoff et al., 2011). Most studies that consider how anthropogenic impacts and natural gradients co-vary have focused on landscape-level features and less attention has been given to other potential stressors, such as the reach-scale presence or absence of non-native species. The negative effects of non-native species in freshwater ecosystems have been extensively documented © 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614

601

(McIntosh, 2000; Rahel, 2010; MacRae & Jackson, 2001; Mooney & Cleland, 2001; Wissinger, McIntosh & Greig, 2006; Gozlan et al., 2010). Although non-native species metrics have been previously acknowledged as being important for biological assessments, their usage has been solely as a response metric to the more frequently used anthropogenic stressors associated with landscape disturbance factors (Angermeier, Smogor & Stauffer, 2000; Wang et al., 2008; Lyons, 2012). However, it is important to consider how non-native species co-vary with potential landscape factors and the resultant confounding interpretations for biological integrity, which has been less frequently acknowledged (Scott & Helfman, 1999). Understanding the roles of landscape-level disturbances, non-native predators and their potential interactions will improve the ability of biological assessments to determine the true anthropogenic disturbances compromising biological integrity. The objective of this study was to evaluate the co-variation of two separate anthropogenic stressors (agricultural development and non-native species), while also controlling for natural landscape features, in the upper (northern) Allegheny River basin. We hypothesised that the effects of land use on stream fish communities (defined herein as reach-scale assemblages) would co-vary with the presence of large-bodied (hence piscivorous) non-native brown trout (Salmo trutta: Salmonidae). We tested this hypothesis by evaluating patterns of association among fish community composition, agricultural development and the presence of a non-native predator using a series of multivariate and univariate analyses at the community (richness, diversity and ecological trait composition) and species level (presence–absence). Specifically, we predicted that changes in land use (conversion of forested into mixed agricultural–forest landscapes) would result in a shift from coolwater to warmwater communities (Lyons et al., 2009; Kanno, Vokoun & Beauchene, 2010; Lyons, 2012), which in turn would also be correlated with the presence or absence of large-bodied brown trout. We also predicted that the presence of large-bodied brown trout would negatively affect the presence of small-bodied species, and hence reduce community diversity.

Methods Study area Stream fish communities were surveyed in second- to third-order, wadeable tributaries of four sub-basins in the upper Allegheny River in northwest Pennsylvania,

M. A. Kirk et al.

602

NY OH

PA

Brokenstraw

Oil French

Allegheny River

Fig. 1 Map of the 99 sampling sites distributed within the four different sub-basins in the Upper Allegheny River. Black dots represent sampling sites, dark black lines represent the different subbasins and the thick dark grey line represents the mainstem of the Allegheny River. Inset map shows the Allegheny River basin relative to the states of Pennsylvania (PA), Ohio (OH) and New York (NY).

U.S.A. (French Creek, Oil Creek, Brokenstraw Creek and small tributaries that drain directly into the Allegheny River; Fig. 1). All sub-basins and most direct tributaries occur within the glaciated Appalachian Plateau, a physiographic region that historically drained north towards Lake Erie prior to the last ice maximum (Cooper, 1983). The upper Allegheny is regionally renowned for having retained most of its native fish diversity, which is likely a result of the low-to-moderate levels of landscape-level disturbance (Cooper, 1983; Argent, Carline & Stauffer, 1998; Argent et al., 2003). The landscape of the region is predominately an agricultural–second-growth forest mixture with increasing second-growth forest along a west-to-east gradient (Whitney & DeCant, 2003). Sub-basins differed in landscape-level characteristics, with French Creek having tributaries with the highest per cent agriculture at the catchment scale and those on the mainstem of the

Allegheny River having the highest natural forest cover (Table 1). Many tributaries of the northern Allegheny River have not experienced the degree of urbanisation, oil and gas extraction and/or mining as compared to the lower (southern) Allegheny and Ohio Rivers (Argent et al., 2003; Whitney & DeCant, 2003). Instead, current impacts of the region are predominately from non-point source stressors associated with low-intensity agriculture (Paulsen et al., 2008; Brown & Froemke, 2012). However, all four sub-basins historically experienced legacy effects during the 19th century associated with intensive agricultural development, timber harvest and/or oil exploration (c. 50–80% deforestation across the region, Whitney & DeCant, 2003). Although the region has not been dramatically affected by the influx of non-native species (Argent & Carline, 2004), exotic brown trout have established naturally reproducing populations in many coldwater tributaries as a result of stocking practices for recreational fishing (Wagner et al., 2013). Despite the general recognition that the upper Allegheny has a relatively intact fish fauna, very little research has been conducted to identify whether these potential stressors are important drivers of fish communities in this region (Argent & Carline, 2004).

Fish community surveys Fish communities were sampled at 99 sites (subset of c. 220 sites surveyed to develop biological assessment tools for this region) during a 10-year period from 2006 to 2015. The 99 streams analysed herein had catchment sizes restricted to 10–35 km2, a size range that corresponds to transitional, coolwater tributaries in this region (Kanno et al., 2010) and controls for changes in community structure along the stream size gradient (Vannote et al., 1980). Surveys were conducted during the months of May–June or August–October under baseflow conditions using pulsed DC backpack electrofishing

Table 1 Mean (SD) landscape characteristics of 99 streams in the four study basins of the Upper Allegheny River, Pennsylvania, U.S.A. Latitude and longitude correspond to sampling location. French Creek tributaries (n = 68) Stream Stream Stream Stream Stream Stream

catchment area (km2) forest cover (%) agriculture (%) gradient (m km 1) longitude latitude

17.41 60.16 32.89 12.29 79.98 41.72

(6.95) (11.23) (11.51) (4.52) (0.16) (0.17)

Oil Creek tributaries (n = 13) 21.26 71.82 24.60 11.15 79.67 41.69

(6.24) (15.64) (15.33) (3.2) (0.10) (0.10)

Allegheny River tributaries (n = 10) 27.04 86.24 9.29 16.53 79.54 41.51

(4.89) (10.92) (8.73) (5.91) (0.12) (0.10)

Brokenstraw Creek tributaries (n = 8) 18.33 77.40 16.46 13.56 79.47 41.85

(6.04) (7.83) (7.11) (5.99) (0.13) (0.06)

© 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614

Impacts of land use and trout on fish communities (Smith-Root Model 12; 15–120 Hz; 200–1000 V, SmithRoot, Vancouver). Surveys were pooled across seasons because no seasonal effects were detected in a priori analyses. First, a quantitative sample was conducted within a 50 m reach by moving upstream in a zig-zag fashion from bank to bank and herding fish upstream to a barrier, such as a blocking seine or natural geomorphic barrier. The quantitative sample involved at least two passes to ensure removal and enumeration of most individuals for an accurate estimate of relative abundance. We resampled 21 sites using a 100 m reach to ensure our quantitative sampling distance was sufficient and found very similar estimates of community composition (e.g. richness, relative abundance) between the two sampling lengths. We required that the quantitative sample include at least 100 individuals of Cyprinidae (i.e. small-bodied minnows) to obtain an accurate estimate of relative abundances for all species. Capturing 100 individuals of this family was easily accomplished for nearly all sites (mean = 158.2) and rarely required extending the distance of the quantitative sample (i.e. all sites involved similar degrees of effort). A previous study also concluded that capturing 120–150 total individuals was important for conducting biological assessments in larger-sized rivers (Hughes & Herlihy, 2007). Furthermore, studies in the eastern U.S.A. provided a rationale for eliminating samples with fewer than 30–50 total individuals (Angermeier, Smogor & Stauffer, 2000; Kanno et al., 2010). Overall, we believe the minimum estimate of 100 individual minnows is both a conservative and rigorous requirement for accurately assessing community composition. To complement this quantitative sample, a qualitative survey was conducted across a separate 100–150 m reach to survey additional habitats that may have been underrepresented in the quantitative survey. Hereafter, references to measures of relative abundance refer to only the quantitative sample (50 m reach) and references to presence/absence of species refer to the combined quantitative/qualitative sample (150–200 m reach). To ensure that an accurate representation of the community was achieved at each site, we chose sampling reaches for both the quantitative and qualitative samples that included all stream habitat types (riffles, runs, pools; Schlosser, 1987). This included sampling multiple units of each habitat type for both samples; although the number of habitat units would have been lower in the quantitative sample. The total length of salmonid species (brown trout and native brook trout, Salvelinus fontinalis) was measured to the nearest millimetre. All fish were identified in the field to the species level except for © 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614

603

ammocoetes of two lamprey species (genus Icthymyzon) and juvenile redhorse (genus Moxostoma), which were lumped at the genus level. We retained small individuals of problematic taxa (e.g. juvenile Moxostoma spp., Notropis spp.) for scrutiny in laboratory aquaria or retention as voucher specimens.

Environmental data We used geographic information systems (GIS; ARCMAP by ERSI on ArcGIS Desktop 9.3.1, ESRI, Redlands) for basin delineations and determining the upstream land use cover at all sites. Land use data were obtained for sites sampled during 2006–2011 (n = 41) from the 2006 National Land Cover Dataset (NLCD; Fry et al., 2011), and those sampled in 2012–2015 (n = 58) from the 2012 NLCD (Homer et al., 2015). Comparing the land cover between these two datasets for a subset of 40 streams revealed negligible changes in land use between years (r = 0.98). The NLCD land use categories were condensed into three primary groups: natural land use (several forest types and wetlands), agricultural land use (cultivated crops and pasture) and developed land use (roads and urbanisation). Given that the primary land use stressor of interest in our region was agricultural, we excluded streams that had developed land use >25%. Additional landscape-level variables were calculated to control for other potential differences in landscape factors across streams and sub-basins. The influence of stream size on fish communities has been thoroughly documented (Vannote et al., 1980; Xenopoulos & Lodge, 2006), and the catchment area (km2) upstream of the sampling site was calculated to control for the observed variation in fish communities among the range of the tributary sizes in this study (10–35 km2). Because there are substantial differences in fish habitats between high-gradient (riffle habitat, cobble-rock substrate) and low-gradient streams (run-pool habitat, finer substrates; Vannote et al., 1980; Argent et al., 2003; Grossman et al., 2010), we calculated channel slope (i.e. stream gradient) for each catchment, defined as metres in elevation drop from source to site based on national hydrography dataset flow lines. Finally, we calculated the distance from the sample site to the confluence of the mainstem of each of the major study basins (French Creek, Brokenstraw Creek, Oil Creek, Allegheny River) as a measure of stream connectivity to control for potential species immigration from downstream source pools (Hitt & Angermeier, 2008). In addition to the landscape covariates discussed above, spatial position within the stream network has an

604

M. A. Kirk et al.

important role in stream community structure (Grenouillet et al., 2004; Hitt & Angermeier, 2008; Kautza & Sullivan, 2012). We completed a principal coordinates of neighbours matrix (PCNM) in the software R (version 3.0.3; R Development Core Team, 2014) to generate spatial variables from a Euclidean distance matrix based on the latitude and longitude of each site. The resulting spatial axes were used to determine whether similarity in community composition was associated with site proximity. We selected PCNM because the resulting spatial axes could be used as predictor variables for both regression and constrained ordination methods (e.g. canonical correspondence analysis [CCA], redundancy analysis [RDA]; Borcard & Legendre, 2002). Residual variation in PCNM variables did not display any spatial trends, which is an important assumption for this method. Finally, an important assumption of this work is that stream temperatures track the changes in land use, as hypothesised in the objectives. Although we sampled stream temperature and other water chemistry parameters (e.g. pH, alkalinity, nitrogen, phosphorus, hardness) at each site, these data represent only point measurements and were not standardised for collection date (i.e. eliminating temporal variability in temperature would have required synoptic sampling). Hence, all associated water chemistry data were censored. Despite the potential temporal variability in temperature data, there was a negative correlation between stream temperature for natural land cover (n = 99, r = 0.32, P = 0.001) and a positive correlation with catchment area (n = 99, r = 0.30, P = 0.002), which leads us to infer that these temperature changes occur along the land use (e.g. streams with higher natural cover have colder temperatures) and stream size (e.g. smaller streams have colder temperatures) gradients evaluated herein.

Multivariate statistics: taxonomic and trait-based composition We characterised fish communities using both multivariate (taxonomic and trait-based) and univariate (species richness, diversity and presence/absence) descriptors. Multivariate analyses included constrained ordination using CCA for examining the relationships between community composition (taxonomic and ecological trait data) and the environmental variables of interest. The most commonly employed constrained ordination methods are CCA and RDA. Unimodal responses require the use of CCA, whereas linear relationships are more appropriate for RDA. Based on the results of detrended correspondence analysis (DCA), we determined that

CCA was the more appropriate method since the gradient length for the first DCA axis was >3 SD (Ter Braak & Prentice, 1988; Kautza & Sullivan, 2012). Separate CCAs were performed for taxonomic data and ecological trait-based data. Trait-based classifications represented thermal guilds, trophic guilds, habitat guilds and tolerance guilds with a total of 11 different classifications among those four trait groups. Species thermal guilds were determined based on cold_, cool_ and warmwater preferences, as defined by Lyons (2012; see Appendix S2). Trophic guilds were categorised as piscivores, invertivores or generalist omnivores. Classification of habitat guilds was defined as either fluvial dependent (require free-flowing lotic habitats), benthic specialist (bottom dwelling) or open water generalists (lentic habitats). Finally, tolerance guilds were defined as either tolerant or sensitive to environmental degradation. Trophic, habitat and tolerance classifications for each species were assigned based on Barbour et al. (1999). The CCAs were performed on log-transformed relative abundance data from quantitative samples, which allowed for the visualisation of sample sites, species and trait classifications in ordination space. Relative abundance data were selected to provide a quantitative comparison of fish communities since most univariate measurements of community composition were associated with metrics of presence/absence from the qualitative and quantitative samples. Species that occurred at ≤5% of all sites were excluded to reduce the influence that rare species can have on multivariate techniques (Niyogi et al., 2007; Gido et al., 2009). The variables included in the CCA were agricultural land use, the number of large, potentially piscivorous adult brown trout (piscivorous defined as >150 mm; McIntosh, 2000), stream gradient, catchment area, distance to mainstem, spatial variables from the PCNM and latitude and longitude to control for potential geographic gradients or changes across sub-basins (see Fig. 1). A forward selection procedure was used to select the best subset of environmental variables for explaining variation in the taxonomic and trait-based data following Blanchet, Legendre & Borcard (2008). Variables were included based on a Monte Carlo permutation test (999 reiterations) and were included in the CCA model if P < 0.05. A Monte Carlo permutation test was also used to test the statistical significance of the canonical axes. The CCAs and forward selection models were completed using R (v. 3.0.3). We used a multivariate analysis of variance (MANOVA; adonis function, R v. 3.0.3) to tease apart the variation in species composition (based on Bray–Curtis © 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614

Impacts of land use and trout on fish communities similarity distances with brown trout removed from the community matrix) associated with both land cover and the presence/absence of large adult brown trout. We focused primarily on these two predictors because of their relevance to the primary objectives of the work and because they were the most frequently selected variables in the model selection processes (see Results). Non-metric multidimensional scaling (NMDS) was used to visualise composition differences as a function of land cover and large trout. We also conducted indicator species analysis (multipatt function, R v. 3.0.3) to identify species that were most associated with the land use and stream size gradients explored in this study. The indicator value is calculated as the product of the species relative frequency and relative average abundance, thus providing a measure of both fidelity and dominance. All 99 sites were split into one-third percentile groups along a land use gradient with the lower 33% being a forest–agriculture mixture (natural land cover <58%), and the upper 33% being predominately natural land cover (natural cover >75%). The stream size gradient consisted of the lower 33% as headwater sites (catchment areas <13 km2) and the upper 33% were medium size streams (areas >23 km2). The presumption was that landscapes with higher natural cover and small headwater streams would share similar indicator species representative of cold- and coolwater communities.

Univariate statistics: community level and species level Six univariate metrics of community diversity were calculated, which constituted elements of fish species richness (qualitative and quantitative samples) and equitability (e.g. evenness of spp. abundance from quantitative samples). The metrics included: (i) total species richness, (ii) richness of darter species (Etheostoma and Percina spp.), (iii) richness of cyprinid species (smallbodied minnows), (iv) richness of centrarchid species (e.g. sunfish and bass), (v) Simpson diversity and (vi) Shannon diversity, as described by Johnson & Angeler (2014). Richness was only calculated for these three taxonomic groups because each group had ≥5 species. General linear models (GLM function, R v. 3.0.3) were then used to evaluate whether agricultural land use and non-native brown trout were associated with fish community metrics (six univariate metrics above) and the presence/absence of nine species, which were assessed using multiple logistic regression (GLM function with binomial distribution, R v. 3.0.3). To ensure enough meaningful variation, we defined common species as © 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614

605

being present at ≥50% of sites, but absent from ≥20% of sites, and present within all four sub-basins. These common species represent a range of different morphological types and occupy different ecological trait guilds that likely reflect differences in stream habitat conditions. The nine species that met these criteria were pumpkinseed (Lepomis gibbosus), longnose dace (Rhinichthys cataractae), central stoneroller (Campostoma anomalum), redside dace (Clinostomus elongatus), common shiner (Luxilus cornutus), bluntnose minnow (Pimephales notatus), fantail darter (Etheostoma flabellare), rainbow darter (Etheostoma caeruleum) and northern hogsucker (Hypentelium nigricans). Family names for each species are provided in Table 3. An information theoretic approach (Burnham & Anderson, 2002; Burnham, Anderson & Huyvaert, 2011) was used to identify the most parsimonious set of predictor variables from an a priori set of 18 candidate models for the community metric and presence/absence models (Appendix S1). The model variables matched those used in the CCAs with two exceptions. First, we used different sub-groupings of either variables of interest (agricultural land use, brown trout) or control variables (three landscape covariates, PCNM variables, latitude and longitude). Secondly, the number of adult brown trout could not be used for linear models due to a highly non-normal, left-skewed distribution (i.e. many zeroes) and was instead replaced with presence/absence data. A priori analyses considering three different piscivorous brown trout categorical variables (absent, present in low numbers, present in high numbers) found no differences between either the high- or low-density variables and thus does not add any new information already provided by the binary present/absent variable. Variance inflation factors (VIF) indicated acceptable collinearity for predictor variables in all models (VIF’s were <3; Zuur, Ieno & Elphick, 2010). Variables with non-normal distributions were log(x + 1) transformed prior to analyses to meet the assumption of homoscedasticity for linear models. Akaike information criterion (AIC) scores corrected for small sample sizes (AICc) were used to compare candidate models with the best model, which was determined with respect to both DAICc (lower AICc being better fit) and the estimated Akaike weights (wi). Since Akaike weights of all candidate models sum to 1, we only included wi ≥ 0.1 for interpretation (Gido et al., 2009). We also calculated an evidence ratio (ER; Burnham et al., 2011) in order to compare the likelihood of the hypotheses associated with the top ranked model and the hypotheses associated with the second ranked

M. A. Kirk et al.

606

model, which was simply calculated as the top models wi divided by the second ranked models wi. Performance of the multiple logistic regression models was assessed using area under the curve (AUC) estimates of receiver operating characteristic plots, in which values range from 0.5 to 1.0, with values of 0.5 indicating poor accuracy and values of 1 indicating perfect accuracy (Buisson, Blanc & Grenouillet, 2008).

Results Multivariate results: taxonomic and trait-based composition In the multivariate analyses, 36 of 55 total species occurred at ≥5% of sites and were included for analysis. For both the taxonomic and trait-based CCAs, brown trout were selected as an important explanatory variable for community composition, whereas agricultural land use was not. The forward selection model included four PCNM variables (PCNM8, PCNM17, PCNM41 and PCNM49), latitude, longitude and the number of large brown trout in the taxonomic-based CCA (all P < 0.02). The first three axes of the taxonomic-based CCA were deemed significant (all P ≤ 0.013). Axis 1 explained 10.1% of the variation in the species matrix and was predominately associated with a longitudinal gradient from west to east and the number of large brown trout (Fig. 2). Axis 2 explained 3.9% of the variation in the species matrix and was associated with most of the PCNM variables. Axis 3 (data not shown) only

explained 2.5% of variation and was very similar to axis 1, representing a brown trout gradient. The cumulative variation explained by the first two axes was relatively low (14.0%), which is likely due to the high overall similarity of fish communities. Species associations from the CCA revealed a general tendency of warmwater and coolwater species to orient along axis 1 (Fig. 3). Coldwater and coolwater species such as brown trout, longnose dace, mottled sculpin and river chub (Nocomis micropogon) were positively correlated with axis 1, whereas warmwater species such as pumpkinseed, bluntnose minnow and central stoneroller were negatively correlated with axis 1. Almost half of all coolwater species exhibited considerable overlap with other warmwater species, in part because many of these coolwater species were present at every site. In contrast, brook trout, a coldwater species, was negatively associated along axis 1; a surprising result given that this species should have oriented to the opposite side of axis 1 with other coldwater species. The CCA for ecological traits revealed similar trends, with patterns of piscivory and temperature guilds corresponding to environmental gradients (Fig. 4). However, the cumulative variation explained by the first two axes of the trait-based CCA was considerably higher than the taxonomic CCA (33.7% versus 14.0%, respectively). Only three PCNM variables (PCNM8, PCNM17 and PCNM41), longitude and the number of large brown

1.0

GSU

BDT

SND

MOX

2

Axis 2 (3.9%)

PCNM49 PCNM41

1

Latitude

Browns

RCK BGL YBL GSD PDC PCK FDT JDTSSH SMB SCP RBD BSD HOG PKS CCK BND RSD STR ICT CSH BNM WHS

0.5

Axis 2 (3.9%)

Allegheny River Brokenstraw French Oil

SPT

0.0

PCNM8

0

–0.5

MIM

–0.5

0.0

FHM

Longitude

–1

BKT

LND BRT RCB ABL RSF

–1.0 –1.5

–2

–3

–2

–1

0

–1.0

0.5

1.0

1.5

Axis 1(10.1%)

PCNM17 1

2

3

Axis 1(10.1%) Fig. 2 Canonical correspondence analysis (CCA) biplot for 99 sample sites based on relative abundance data of 36 species. Variables shown in the biplot are those selected from the forward model selection process. Circles are placed around the three streams with allopatric, wild brook trout populations.

Fig. 3 Canonical correspondence analysis (CCA) plot based on relative abundance data of 36 species recorded at 99 sample sites. Variables from the biplot in Fig. 2 are removed to improve interpretability of species locations in ordination space. Ovals are placed around coolwater species, rectangles are placed around coldwater species and warmwater species are left open. Thermal designations are taken from Lyons (2012) and species abbreviations are provided in Appendix S2. © 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614

Impacts of land use and trout on fish communities 2.0

Pcnm17 OMNIVORE

0.2

Axis 2 (2.8 %)

PISCIVORE

OPEN 0.0

SENSITIVE

TOLERANT

1.5

Longitude Browns

COOL COLD

Pcnm8 BENTHIC

–0.2

WARM

INVERTIVORE

Pcnm41

FLUVIAL

NMDS Axis 1 Score

0.4

607

Brown trout absent (R2 = 0.05) Brown trout present (R2 = 0.12)

1.0 0.5 0.0 –0.5 –1.0 –1.5

–0.4

–2.0

–0.8

–0.6

–0.4

–0.2

0.0

0.2

0.4

0.6

0.8

Axis 1 (30.9%) Fig. 4 Canonical correspondence analysis (CCA) biplot for 11 ecological trait groups based on relative abundance data of 36 species recorded at 99 sample sites. Variables shown in the biplot are those selected from the forward model selection process. Ecological trait groups are capitalised and biplot predictor variables from the forward selection model are italicised.

trout were selected for inclusion (all P ≤ 0.01). The first canonical axis was the only one that significantly explained variation in trait groups (P < 0.001). Trait associations were most evident for the thermal and trophic guilds (Fig. 4), with piscivorous and coldwater species positively correlated along axis 1. Omnivorous generalists, open water generalists and warmwater species were all negatively correlated along axis 1. MANOVA revealed that sites with similar land use had different community structure with brown trout present than with brown trout absent, even after the removal of brown trout from the community matrix. Significant effects of both natural land cover (R2 = 0.02, P = 0.014) and the presence of large brown trout (R2 = 0.12, P < 0.001) were observed, despite weak explanatory power. However, the community changes associated with the presence/absence of large brown trout was more evident than land use when visualised in the NMDS (Fig. 5). The slope of a simple regression line indicated that changes in community composition along the land use gradient with large brown trout absent were lower than when present (R2 = 0.05 versus 0.12, respectively). These compositional differences were especially evident for streams with natural land cover values in the 60–80% range (Fig. 5). Cold- and coolwater taxa had indicator values that were significantly (P < 0.05) associated with predominately natural land uses (>75% forest). Brown trout, longnose dace and river chub were the three significant indicator species for streams with more natural cover of © 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614

40

60

80

100

Natural land use cover (%)

Fig. 5 Community composition from a non-metric multidimensional (NMDS) plotted as a function of natural land use cover and large brown trout presence and absence based on the multivariate analysis of variance (MANOVA).

the 35 species (mottled sculpin excluded), whereas ten species (predominately warmwater cyprinids and darter species) were significantly associated with streams with higher agricultural cover. Surprisingly, this pattern was not the same for the stream size gradient. Brown trout and river chub were significantly associated with medium-sized streams (>23 km2), whereas three of the five significant indicator species for small headwaters (<13 km2) were warmwater cyprinids (Appendix S2).

Univariate results: community level The top ranked models for all six fish community metrics included either agricultural land use or the presence/absence of large brown trout (Table 2). The top ranked model for total richness included both a positive relationship with agricultural land use and a negative relationship with the presence of large brown trout. Darter richness and centrarchid richness exhibited positive relationships with agricultural land use, while centrarchid richness was also associated with landscape covariates of gradient, area and distance to the mainstem. The top ranked model for cyprinid richness included just the brown trout term and indicated a negative association with the presence of large brown trout, which was 2.5 times stronger than the second ranked model (Table 2). Streams with large brown trout also had significantly lower measures of Shannon and Simpson diversity, indicating a tendency for community homogenisation. Overall, the explanatory power of these models was relatively low and highly variable (0.07 ≤ R2 ≤ 0.35).

608

M. A. Kirk et al.

Table 2 Top ranked multiple linear regression models (only Akaike weights ≥0.10) from Akaike information criterion (AIC) models for six univariate descriptors of fish community composition. Models were ranked based on differences in AIC corrected for small sample size (DAICc), Akaike model weights (wi) and evidence ratios (ER). Model (observed relationships)

R2

Total richness

AG (+), PISC ( ) AG (+) PISC ( ) AG (+) AG (+), PISC ( ) AG (+), GRAD ( ), DIST ( ), AREA (+) PISC (+), GRAD ( ), DIST ( ), AREA (+) PISC ( ) AG (+), PISC ( ) PISC ( ), LAT ( ), LONG (+) AG (+), GRAD ( ), DIST ( ), AREA (+) GRAD ( ), DIST ( ), AREA (+) PISC ( ), GRAD ( ), DIST ( ), AREA (+) PISC ( ), GRAD (+), DIST (+), AREA ( ) PISC ( ), PCNM variables PISC ( ) PISC ( ), GRAD (+), DIST (+), AREA ( ) PISC ( ) AG (+), PISC ( )

0.10 0.08 0.07 0.12 0.13 0.17 0.14 0.12 0.12 0.13 0.12 0.09 0.12 0.35 0.38 0.30 0.29 0.23 0.23

Darter richness

Cyprinid richness

Centrarchid richness

Simpson diversity

Shannon diversity

The decline in community evenness was particularly noticeable for two ubiquitous cyprinids, creek chub (Semotilus atromaculatus) and western blacknose dace (Rhinichthys obtusus), as well as common darter species (Etheostoma and Percina spp.). When large brown trout were absent, the relative abundance of blacknose dace + creek chub, and darter species made up a similar proportion of the community (32.1% versus 30.4%, respectively). However, when large brown trout were present, the relative abundance of darter species decreased by 150% and the relative abundance of blacknose dace and creek chub increased (12.4% versus 44.7%, respectively; Fig. 6). There was no correlation between the relative abundance of blacknose dace and creek chub and per cent agricultural land use (n = 99, r = 0.09, P = 0.376), and only a weak positive correlation between the relative abundance of darters and per cent agricultural land use (n = 99, r = 0.29, P = 0.003).

Univariate results: species level Logistic regression models revealed that agricultural land use and large brown trout were also important predictors for the presence–absence of individual species (Table 3). However, the AUC values for these models ranged from 0.59 to 0.88 and generally indicated weakto-average performance for most models. Piscivory not only had a negative effect in the top ranked models of four species (pumpkinseed, bluntnose minnow, common

Relative abundance from quantative sample (%)

Community metric

AICc 581.0 581.2 582.9 371.6 371.8 372.8 372.8 441.7 443.6 444.4 12.0 11.9 11.3 112.1 111.7 100.8 106.7 108.0 109.0

DAICc

wi

ER

0 0.11 1.84 0 0.19 1.15 1.17 0 1.81 2.64 0 0.01 0.62 0 0.46 1.33 0 1.34 2.32

0.30 0.28 0.12 0.26 0.24 0.15 0.15 0.52 0.21 0.14 0.23 0.22 0.17 0.30 0.24 0.15 0.43 0.22 0.13

1.07 – – 1.08 – – – 2.48 – – 1.05 – – 1.25 – – 1.95 – –

50

Blacknose dace/creek chub Darter spp. 40

30

20

10

0 Present

Absent

Piscivorous brown trout

Fig. 6 Relative abundance of blacknose dace and creek chub compared with relative abundance of all darter species of total fish abundance from the quantitative sample in relation to the presence of large brown trout (present/absent).

shiner and fantail darter) but also had a positive effect in the top ranked models of two species (longnose dace and northern hogsucker). Notably, the top models with large brown trout for common shiner and northern hogsucker were 1.9 and 2.4 times stronger than the second ranked models. Agricultural land use was the most important variable in the top ranked models for rainbow darters, central stonerollers and redside dace, whereas landscape covariates also had important effects on redside dace and central stonerollers. © 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614

Impacts of land use and trout on fish communities

609

Table 3 Top ranked multiple logistic regression models (only Akaike weights ≥0.10) from Akaike information criterion (AIC) models for nine common species captured during our study. Models were ranked based on differences in AIC corrected for small sample size (DAICc), Akaike model weights (wi) and evidence ratios (ER). Species

Model (observed relationships)

AUC

AICc

DAICc

wi

ER

Bluntnose minnow (Cyprinidae)

PISC ( ) LAT ( ), LONG (+) AG (+) AG (+), GRAD (+), DIST (+), AREA ( ) LAT ( ), LONG (+) AG (+) PISC ( ) PISC ( ), PCNM variables PISC ( ) PCNM Variables AG (+) PISC ( ) AG (+) LAT ( ), LONG ( ) GRAD (+), DIST (+), AREA (+) PISC (+) AG ( ), PISC (+) PISC (+) AG ( ) LAT (+), LONG (+) PISC ( ) PISC ( ), AG (+), LAT ( ), LONG ( ) PISC ( ), AG (+) PISC ( ), LAT ( ), LONG ( ) AG (+) PISC ( ) LAT (+), LONG ( ) AG (+), LAT (+), LONG ( ) AG (+), GRAD ( ), DIST ( ), AREA ( ) AG (+) PISC (+) LAT (+), LONG (+)

0.64 0.71 0.65 0.81 0.74 0.69 0.69 0.78 0.62 0.76 0.63 0.62 0.61 0.71 0.73 0.76 0.80 0.69 0.61 0.63 0.74 0.81 0.76 0.78 0.63 0.59 0.67 0.73 0.88 0.61 0.67 0.73

129.0 130.0 131.1 90.3 90.4 90.6 90.8 136.9 138.3 138.3 138.3 88.2 88.3 88.6 89.2 89.1 90.9 141.2 141.2 143.1 133.8 134.8 135.4 135.6 103.6 103.8 104.5 104.9 128.6 128.7 128.9 129

0 1.00 2.10 0 0.12 0.29 0.49 0 1.30 1.39 1.47 0 0.10 0.38 0.98 0 1.74 0 0.1 1.9 0 1.00 1.58 1.76 0 0.15 0.85 1.28 0 0.11 0.33 0.36

0.20 0.19 0.18 0.16 0.14 0.14 0.12 0.27 0.14 0.13 0.13 0.19 0.18 0.16 0.11 0.47 0.20 0.28 0.27 0.11 0.33 0.20 0.15 0.14 0.22 0.21 0.15 0.12 0.14 0.13 0.12 0.12

1.05 – – 1.14 – – – 1.93 – – – 1.06 – – – 2.35 – 1.04 – – 1.65 – – – 1.05 – – – 1.08 – – –

Central stoneroller (Cyprinidae)

Common shiner (Cyprinidae)

Fantail darter (Percidae)

Northern hogsucker (Catostomidae) Longnose dace (Cyprinidae) Pumpkinseed (Centrarchidae)

Rainbow darter (Percidae)

Redside dace (Cyprinidae)

Discussion There is general consensus that the mechanisms of freshwater community organisation (fish, invertebrates, algae) are highly interactive and occur at multiple spatial scales (Jackson et al., 2001; Stephenson & Morin, 2009; Kautza & Sullivan, 2012; Marzin et al., 2013). In our study, variation in fish community composition between small tributaries of the upper Allegheny River was associated with anthropogenic factors that co-vary at both landscape (land use gradients) and local scales (non-native predators). Species sort based on physiological tolerances for water temperature (i.e. cold- and coolwater versus warmwater communities; Lyons et al., 2009; Buisson et al., 2008; Kanno et al., 2010; Lyons, 2012) that change along an environmental gradient. However, species also respond to the presence or absence of large brown trout (a non-native predator) at the reach scale. In addition, the effects of large brown trout co-vary to some degree © 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614

with the observed effects from agricultural land use (i.e. brown trout typically occur in streams with relatively high natural cover and low agricultural land use). In this study, communities and individual species exhibited clear separations in ordination space that was associated with shifts in thermal preferences (Figs 3 & 4). These temperature-based changes in communities are likely to be driven by some interaction between landscape- and local-level processes. Landscape-level land use and reach-scale riparian condition are considered to be the primary determinants of in-stream conditions at the reach scale (Argent & Carline, 2004; Wagenhoff et al., 2011; Marzin et al., 2013), with the relative importance of each varying among regions (Kautza & Sullivan, 2012). For example, agricultural development associated with riparian clearing, increased siltation and turbidity and alterations in the proportion of groundwater–surface water runoff can all be associated with changes in stream temperatures (Allan, 2004; Diana et al., 2006;

610

M. A. Kirk et al.

Infante & Allan, 2010; Niyogi et al., 2007; Johnson & Angeler, 2014), thus driving observed shifts in thermal groups. Importantly, the observed changes appear to be independent of the natural changes in thermal groups typically associated with stream size gradients (Wang et al., 2008; Kanno et al., 2010). Although indicator species for the land use categories matched fairly well for thermal groups (i.e. warmwater species were associated with agricultural land use), the indicator species for small headwaters (areas <13 km2) and medium-sized streams (>23 km2) did not (see Appendix S2). Coolwater species (e.g. brown trout) were indicative of medium-sized streams and warmwater minnows (e.g. common shiner, fathead minnow; Cooper, 1983) were often encountered in headwater streams. This was surprising because cold- and coolwater communities typically predominate in small headwater streams in this region (Lyons et al., 2009; Kanno et al., 2010; Beauchene et al., 2014). One potential explanation for these nonintuitive outcomes is that the narrow stream size gradient evaluated in these analyses does not accurately reflect differences in communities (i.e. warmwater species might not become uncommon until streams were <10 km2 in size). A second explanation is that land use had a stronger effect on community composition than natural changes in the stream continuum. Trends towards community homogenisation along the stream size continuum have been previously observed in agriculturally impaired streams (Delong & Brusven, 1998). Such changes in community composition from land use effects reflect a shift in trait-based characteristics rather than changes in species richness and diversity (Angermeier et al., 2000; Frimpong & Angermeier, 2009), as observed in Fig. 4. Indeed, biological assessments based on thermal traits have become a popular method for controlling background variation in temperate streams that encompass thermal transitions (Southerland et al., 2007; Lyons et al., 2009; Lyons, 2012; Beauchene et al., 2014). A third explanation for the observed non-intuitive patterns between thermal groups and land use effects is associated with the differential susceptibility of species to the presence of non-native brown trout. For example, native brook trout did not sort with the other coldwater salmonid species (non-native brown trout) in multivariate space (Fig. 3). Prior to deforestation during European settlement in this region, brook trout were abundant and the dominant predators in cold- and coolwater streams (Hudy et al., 2008; DeWeber & Wagner, 2015). Displacement of native brook trout by introduced brown trout is a well-described pattern in eastern North

America (McKenna, Slattery & Clifford, 2013; Wagner et al., 2013), as are the competitive and predatory advantages of brown trout (Fausch & White, 1981, 1986; Hoxmeier & Dieterman, 2013), except in cold, headwater streams (Ohlund et al., 2008). Given that non-native predatory species often have greater impacts on prey communities than co-evolved native piscivores (Fausch & White, 1981; Scott & Helfman, 1999; Quist & Hubert, 2004; Lepori et al., 2012), brown and brook trout should have different effects on fish communities. Indeed, three streams with allopatric populations of native, piscivorous brook trout did not ordinate in the same multivariate space as non-native, piscivorous brown trout streams (circles; Fig. 2). These patterns may also be confounded by a stream size or temperature effect, which we are currently exploring in a complementary study on trout distributions in first-order headwater streams (Kirk & Wissinger, unpubl. data). Whereas the effects of non-native salmonids on native salmonids have been extensively documented, the effects of non-native salmonids on other native fish are less well documented (McIntosh, 2000; Turek et al., 2014). Salmonids are generally considered aggressive, highly territorial feeders that have stronger foraging interactions and higher energetic demands compared with warmwater predators (Fausch & White, 1986; Schlosser, 1987). In our study, small-bodied fish (cyprinids, darters) appear to be the most susceptible to the presence of large brown trout (Tables 2 & 3), indicating the potential for size-selective predation (Power, Matthews & Stewart, 1985; Schlosser, 1987). The negative association between small-bodied species and brown trout has been previously documented independent of temperature changes (Lyons et al., 2009) and land use practices (Townsend & Crowl, 1991). The impacts of agricultural development may thus prevent the establishment of these coldwater, non-native predators and consequently allow the establishment of species that would otherwise be susceptible to interactions with brown trout. In our study, brown trout altered community structure in a relatively species-rich system. Hence, brown trout effects on aquatic taxa are likely to be more pronounced in regions where large-bodied piscivores were historically absent and/or that have naturally depauperate communities (e.g. New Zealand and Australia; Townsend & Crowl, 1991; McIntosh, 2000; Wissinger et al., 2006; Sowersby, Thompson & Wong, 2015). An important limitation of our results is acknowledging that the correlations we document do not imply causation, since we did not directly evaluate brown trout piscivory. The responses of longnose dace and northern © 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614

Impacts of land use and trout on fish communities hogsucker (see Table 3) highlight the potential for collinearity with brown trout and habitat features (i.e. the positive association is likely due to a shared affinity for streams with similar habitat characteristics). However, we hypothesise the ability of these species to co-occur with large brown trout is related to lower predation susceptibility. Longnose dace are microhabitat specialists that are encountered in shallow riffle habitats not inhabited by large brown trout. Northern hogsucker is the only species in our models that reaches a body size comparable to large brown trout, thus attaining a size refuge from predation (e.g. Turek et al., 2014). In contrast, species like common shiner and bluntnose minnow that do not reach large body sizes and occupy deeper pool habitats should be more susceptible to interactions with large brown trout. Overall, our results are consistent with the notion that species susceptibility to predators will depend upon habitat preferences (Schlosser, 1987; McIntosh, 2000; Sowersby et al., 2015). Further research is warranted to validate these results and identify the mechanisms by which non-native brown trout influence the composition of fish communities. Our results highlight the complicated nature of how the local presence of a non-native predator can interact and co-vary with the processes of anthropogenic stressors at the landscape level (Allan, 2004). In the context of biological assessment, the presence of brown trout can be interpreted as both a positive and negative indicator of biological integrity. In one sense, the presence of naturalised populations of brown trout can serve as a quick indicator of relatively undisturbed, in-stream conditions (i.e. forested landscapes with low-to-moderate impacts). Prior to European settlement, most tributaries in our study region would have been embedded in predominately forested landscapes (Whitney & DeCant, 2003). The resulting fish faunas would have therefore mostly resembled coolwater communities where wild brown trout are a current indicator of. In general, trout streams are often identified as being high-quality habitats that are associated with minimum anthropogenic disturbances and are often considered positive biological indicators (Wang et al., 2008). Conversely, the addition of non-native brown trout will likely alter aquatic community composition such that communities will not resemble those expected at reference sites that do not contain these non-native predators (see Angermeier et al., 2000; Hawkins, 2006; Hawkins et al., 2015). This is the general consensus of most biological assessments, that non-native salmonids should not be considered as an important response metric in bioassessment protocols. Given the global © 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614

611

introductions of many salmonid species and the welldocumented effects of non-native piscivores (MacRae & Jackson, 2001; McIntosh, McHugh & Budy, 2011), we suspect this complicated issue of co-varying effects from landscape-level stressors and non-native species is a wide spread problem for the development of biological assessments. This problem is likely true for many stream systems across North America, where non-native salmonids have been introduced into pristine, high-quality streams after either the extirpation or local declines in native salmonids (Quist & Hubert, 2004). Clearly more consideration needs to be given to whether the presence of self-recruiting populations of non-native salmonids is viewed positively or negatively in the context of biological assessment. In particular, ‘reference-condition’-based biological assessments should consider accounting for both landscape-level disturbances and the possible occurrence of non-native species at more local scales (e.g. Hawkins, 2006). When establishing biological criteria for assessing stream integrity, a priori knowledge of the mechanisms acting on community organisation and the level of covariation among them should help guide decisions in metric selection to improve the strength of biological assessments (Southerland et al., 2007; Wang et al., 2008). Rather than variation between just landscape-level features, we observed associations between natural stream gradients, human-induced landscape changes and a non-native species. The co-varying impacts of landscapelevel alterations and non-native species will likely confound interpretations of biological integrity aimed at inferring cumulative impacts of land use disturbances at multiple scales (landscape and local). Simultaneous consideration of non-native species impacts at the local scale, alongside anthropogenic disturbances at the landscape scale, should vastly improve the effectiveness of biological assessments in elucidating human impacts on stream ecosystems (Scott & Helfman, 1999).

Acknowledgments This research was funded by Allegheny College (2005–2015) and in recent years (2012–2015) by National Fish and Wildlife Federation grants to Scott Wisssinger in support of the Unassessed Waters Initiative of the Pennsylvania Fish and Boat Commission. We thank Lynette Gardner, Elizabeth Goetz, Kirsten Ressel, Anna Zimmerman, Katie Robbins, Scott Kirk, Jared Balik and Susan Washko who contributed to data collection and preliminary analyses. Chris Shaffer provided invaluable assistance with GIS. Finally, we would like to thank Milt

612

M. A. Kirk et al.

Ostrofsky, Angus McIntosh, Amanda Delvecchia, Beno^ıt Demars and two anonymous reviewers who invested significant time and energy into reviewing earlier drafts and helping to improve this manuscript.

References Allan J.D. (2004) Landscapes and riverscapes: the influence of land use on stream ecosystems. Annual Review of Ecology, Evolution and Systematics, 35, 257–284. Angermeier P.L., Smogor R.A. & Stauffer J.R. (2000) Regional frameworks and candidate metrics for assessing biotic integrity in Mid-Atlantic highland streams. Transactions of the American Fisheries Society, 129, 962–981. Argent D.G., Bishop J.A., Stauffer J.R., Carline R.F. & Myers W.L. (2003) Predicting freshwater fish distributions using landscape-level variables. Fisheries Research, 60, 17–32. Argent D.G. & Carline R.F. (2004) Fish assemblage changes in relation to watershed land use disturbance. Aquatic Ecosystem Health and Management, 7, 101–114. Argent D.G., Carline R.F. & Stauffer J.R. (1998) Changes in the distribution of Pennsylvania fishes: the last 100 years. Journal of Pennsylvania Academy of Sciences, 72, 32–37. Barbour M.T., Gerritsen J., Snyder B.D. & Stribling J.B. (1999) Rapid Bioassessment Protocls for Use in Streams and Wadealbe Rivers: Periphyton, Benthic Macroinvertebrates and Fish, 2nd Edn. EPA 841-B-99-002. U.S. Environmental Protection Agency, Office of Water, Washington. Beauchene M., Becker M., Bellucci C.J., Hagstrom N. & Kanno Y. (2014) Summer thermal thresholds of fish community structure in Connecticut streams. North American Journal of Fisheries Management, 34, 119–131. Blanchet F.G., Legendre P. & Borcard D. (2008) Forward selection of explanatory variables. Ecology, 89, 2623–2632. Borcard D. & Legendre P. (2002) All-scale spatial analysis of ecology data by means of principal coordinates of neighbor matrices. Ecological Modelling, 153, 51–68. Brown T.C. & Froemke P. (2012) Nationwide assessment of nonpoint source threats to water quality. BioScience, 62, 136–146. Buisson L., Blanc L. & Grenouillet G. (2008) Modelling stream fish species distribution in a river network: the relative effects of temperature versus physical factors. Ecology of Freshwater Fish, 17, 244–257. Burnham K.P. & Anderson D.R. (2002) Model Selection and Multi-Model Inference, 2nd edn. Springer-Verlag, New York. Burnham K.P., Anderson D.R. & Huyvaert K.P. (2011) AIC model selection and multimodel inference in behavioral ecology: some background, observations, and comparisons. Behavioral Ecology and Sociobiology, 65, 23–35. Cooper E.L. (1983) Fishes of Pennsylvania and the Northeastern United States. The Pennsylvania State University Press, University Park.

Delong M.D. & Brusven M.A. (1998) Macroinvertebrate community structure along the longitudinal gradient of an agriculturally impacted stream. Environmental Management, 22, 445–457. DeWeber J.T. & Wagner T. (2015) Predicting brook trout occurrence in stream reaches throughout their native range in the Eastern United States. Transactions of the American Fisheries Society, 144, 11–24. Diana M., Allan J.D. & Infante D. (2006) The influence of physical habitat and land use on stream fish assemblages in Southeastern Michigan. American Fisheries Society Symposium, 48, 359–374. Fausch K.D., Karr J.R. & Yant P.R. (1984) Regional application of an index of biotic integrity based on stream fish communities. Transactions of the American Fisheries Society, 113, 39–55. Fausch K.D., Torgeren C.E., Baxter C.V. & Li H.W. (2002) Landscapes to riverscapes: bridging the gap between research and conservation of stream fishes. BioScience, 52, 483–498. Fausch K.D. & White R.J. (1981) Competition between brook trout (Salvelinus fontinalis) and brown trout (Salmo trutta) for positions in a Michigan stream. Canadian Journal of Fisheries and Aquatic Sciences, 38, 1220–1227. Fausch K.D. & White R.J. (1986) Competition among juveniles of Coho salmon, brook trout, and brown trout in a laboratory stream, and implications for Great Lakes tributaries. Transactions of the American Fisheries Society, 115, 363–381. Frimpong E.A. & Angermeier P.L. (2009) FishTraits: a database of ecological and life-history traits of freshwater fishes of the United States. Fisheries, 34, 487–495. Fry J., Xian G., Jin S., Dewitz J., Homer C., Yang L. et al. (2011) Completion of the 2006 National Land Cover Database for the conterminous United States. Photogrammetric Engineering and Remote Sensing, 77, 858–864. Gido K.B., Schaefer J.F. & Falke J.A. (2009) Convergence of fish communities from the littoral zone of reservoirs. Freshwater Biology, 54, 1163–1177. Gozlan R.E., Britton J.R., Cowx I. & Copp G.H. (2010) Current knowledge on non-native freshwater fish introductions. Journal of Fish Biology, 76, 751–786. Grenouillet G., Pont D. & Hẽrissẽ C. (2004) Within-basin fish assemblage structure: the relative influence of habitat versus stream spatial position on local species richness. Canadian Journal of Fisheries and Aquatic Sciences, 61, 93– 102. Grossman G.D., Ratajczak R.E., Farr M.D., Wagner C.M. & Petty J.T. (2010) Why there are fewer fish upstream. American Fisheries Society Symposium, 73, 63–81. Hawkins C.P. (2006) Quantifying biological integrity by taxonomic completeness: its utility in regional and global assessments. Ecological Applications, 16, 1277–1294. Hawkins C.P., Mykra H., Oksanen J. & Vander Laan J.J. (2015) Environmental disturbance can increase beta © 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614

Impacts of land use and trout on fish communities diversity of stream macroinvertebrate assemblages. Global Ecology and Biogeography, 24, 483–494. Heino J., Mykra H., Hamalainen H., Aroviita J. & Muotka T. (2007) Responses of taxonomic distinctness and species diversity indices to anthropogenic impacts and natural environmental gradients in stream macroinvertebrates. Freshwater Biology, 52, 1846–1861. Hitt N.P. & Angermeier P.L. (2008) River-Stream connectivity affects fish bioassessment performance. Environmental Management, 42, 132–150. Homer C.G., Dewitz J.A., Yang L., Jin S., Danielson P., Xian G. et al. (2015) Completion of the 2011 National Land Cover Database for the conterminous United StatesRepresenting a decade of land cover change information. Photogrammetric Engineering and Remote Sensing, 8, 345– 354. Hoxmeier R.J.H. & Dieterman D.J. (2013) Seasonal movement, growth and survival of brook trout in sympatry with brown trout in Midwestern US streams. Ecology of Freshwater Fish, 22, 530–542. Hudy M., Thieling T.M., Gillespie N. & Smith E.P. (2008) Distribution, status, and land use characteristics of subwatersheds within the native range of brook trout in Eastern United States. North American Journal of Fisheries Management, 28, 1069–1085. Hughes R.M. & Herlihy A.T. (2007) Electrofishing distance needed to estimate consistent index of biotic integrity (IBI) scores in raftable Oregon rivers. Transactions of the American Fisheries Society, 136, 135–141. Infante D.M. & Allan J.D. (2010) Response of stream fish assemblages to local-scale habitat as influenced by landscape: a mechanistic investigation of stream fish assemblages. American Fisheries Society Symposium, 73, 371–397. Jackson D.A. & Harvey H.H. (1989) Biogeographic associations in fish assemblages: local vs. regional processes. Ecology, 70, 1472–1484. Jackson D.A., Peres-Nato P.R. & Olden J.D. (2001) What controls who is where in freshwater fish communities – the roles of biotic, abiotic, and spatial factors. Canadian Journal of Fisheries and Aquatic Sciences, 58, 157–170. Johnson R.K. & Angeler D.G. (2014) Effects of agricultural land use on stream assemblages: taxon-specific responses of alpha and beta diversity. Ecological Indicators, 45, 386– 393. Kanno Y., Vokoun J.C. & Beauchene M. (2010) Development of dual fish multi-metric indices of biological condition for streams with characteristic thermal gradients and low species richness. Ecological Indicators, 10, 565–571. Karr J.R. (1991) Biological integrity: a long-neglected aspect of water resource management. Ecological Applications, 1, 66–84. Kautza A. & Sullivan S.M.P. (2012) Relative effects of localand landscape-scale environmental factors on stream fish assemblages: evidence from Idaho and Ohio, USA. Fundamental and Applied Limnology, 180, 259–270. © 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614

613

Lepori F., Benjamin J.R., Fausch K.D. & Baxter C.V. (2012) Are invasive and native trout functionally equivalent predators? Results and lessons from a field experiment. Aquatic Conservation: Marine and Freshwater Ecosystems, 22, 787–798. Lyons J. (2012) Development and validation of two fishbased indices of biotic integrity for assessing perennial coolwater streams in Wisconsin, U.S.A. Ecological Indicators, 23, 402–412. Lyons J., Zorn T., Stewart J., Seelbach P., Wehrly K. & Wang L. (2009) Defining and characterizing coolwater streams and their fish assemblages in Michigan and Wisconsin, USA. North American Journal of Fisheries Management, 29, 1130–1151. MacRae P.S. & Jackson D.A. (2001) The influence of smallmouth bass (Micropterus dolomieu) predation and habitat complexity on the structure of littoral zone fish assemblages. Canadian Journal of Fisheries and Aquatic Sciences, 58, 342–351. Marzin A., Verdonschot P.F.M. & Pont D. (2013) The relative influence of catchment, riparian corridor, and reach-scale anthropogenic pressures on fish and macroinvertebrate assemblages in French Rivers. Hydrobiologia, 704, 375–388. McIntosh A.R. (2000) Habitat and size-related variations in exotic trout impacts on native galaxiid fishes in New Zealand streams. Canadian Journal of Fisheries and Aquatic Sciences, 57, 2140–2151. McIntosh A.R., McHugh P.A. & Budy P. (2011) Brown trout (Salmo trutta). In: A Handbook of Global Freshwater Invasive Species (Ed. R.A. Francis), pp. 285–296. Earthscan, London. McKenna J.E., Slattery M.T. & Clifford K.M. (2013) Broadscale patterns of brook trout responses to introduced brown trout in New York. North American Journal of Fisheries Management, 33, 1221–1235. Mooney H.A. & Cleland E.E. (2001) The evolutionary impact of invasive species. Proceedings of the National Academy of Sciences of the United States of America, 98, 5446–5451. Niyogi D.K., Koren M.K., Arbuckle C.J. & Townsend C.R. (2007) Stream communities along a catchment land-use gradient: subsidy-stress responses to pastoral development. Environmental Management, 39, 213–225. Ohlund G., Nordwall F., Degerman E. & Eriksson T. (2008) Life history and large-scale habitat use of brown trout (Salmo trutta) and brook trout (Salvelinus fontinalis) – implications for species replacement patterns. Canadian Journal of Fisheries and Aquatic Sciences, 65, 633–644. Paulsen S.G., Mayio A., Peck D.V., Stoddard J.L., Tarquinio E., Holdsworth S.M. et al. (2008) Condition of stream ecosystems in the US: an overview of the first national assessment. Journal of the North American Benthological Society, 27, 813–821.

614

M. A. Kirk et al.

Power M.E., Matthews W.J. & Stewart A.J. (1985) Grazing minnows, piscivorous bass, and stream algae: dynamics of a strong interaction. Ecology, 66, 1448–1456. Quist M.C. & Hubert W.A. (2004) Bioinvasive species and the preservation of cutthroat trout in the western United States: ecological, social, and economic issues. Environmental Science and Policy, 7, 303–313. R Development Core Team (2014) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna. Available at: http://www. R-project.org. Rahel F.J. (2010) Homogenization, differentiation, and the widespread alteration of fish faunas. American Fisheries Society Symposium, 73, 371–397. Schlosser I.J. (1987) The role of predation in age- and sizerelated habitat use by stream fishes. Ecology, 68, 651–659. Scott M.C. & Helfman G.S. (1999) Native invasions, homogenization, and the mismeasure of integrity of fish assemblages. Fisheries, 26, 6–15. Southerland M.T., Rogers G.M., Kline M.J., Morgan R.P., Boward D.M., Kazyak P.F. et al. (2007) Improving biological indictors to better assess the condition of streams. Ecological Indicators, 7, 751–767. Sowersby W., Thompson R.M. & Wong B.B.M. (2015) Invasive predator influences habitat preferences in a freshwater fish. Environmental Biology of Fish, 99, 1–7. Stephenson J.M. & Morin A. (2009) Covariation of stream community structure and biomass of algae, invertebrates and fish with forest cover at multiple spatial scales. Freshwater Biology, 54, 2139–2154. Ter Braak C.J.F. & Prentice I.C. (1988) A theory of gradient analysis. Advances in Ecological Research, 18, 271–317. Townsend C.R. & Crowl T.A. (1991) Fragmented population structure in a native New Zealand fish: an effect of introduced brown trout? Oikos, 61, 347–354. Turek K.C., Pegg M.A., Pope K.L. & Schainost S. (2014) Potential population and assemblage influences of nonnative trout on native nongame fish in Nebraska headwater streams. Ecology of Freshwater Fish, 25, 99–108. 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, 130–137.

Wagenhoff A., Townsend C.R., Philips N. & Matthaei C.D. (2011) Subsidy-stress and multiple-stressor effects along gradients of deposited fine sediment and dissolved nutrients in a regional set of streams and rivers. Freshwater Biology, 56, 1916–1936. Wagner T., DeWeber J.T., Detar J. & Sweka J.A. (2013) Landscape-scale evaluation of asymmetric interactions between brown trout and brook trout using two-species occupancy models. Transactions of the American Fisheries Society, 142, 353–361. Wang L., Brenden T., Seelbach P., Cooper A., Allan D., Clark R. et al. (2008) Landscape based identification of human disturbance gradients and reference conditions for Michigan streams. Environmental Monitoring Assessment, 141, 1–17. Whitney G.G. & DeCant J.P. (2003) Physical and historical determinants of the pre- and post-settlement forests of northwestern Pennsylvania. Canadian Journal of Forest Resources, 33, 1683–1697. Wissinger S.A., McIntosh A.R. & Greig H.S. (2006) Impacts of introduced brown and rainbow trout on benthic invertebrate communities in shallow New Zealand lakes. Freshwater Biology, 51, 2009–2028. Xenopoulos M.A. & Lodge D.M. (2006) Going with the flow: using species-discharge relationships to forecast losses in fish biodiversity. Ecology, 87, 1907–1914. Zuur A.F., Ieno E.N. & Elphick C.S. (2010) A protocol for data exploration to avoid common statistical problems. Methods in Ecology and Evolution, 1, 3–14.

Supporting Information Additional Supporting Information may be found in the online version of this article: Appendix S1. Set of 18 candidate models used for all regression models. Appendix S2. Thermal guild designations, indicator species analyses, and Scores for the 36 fish species along the three axes of the canonical correspondence analysis (CCA) in Fig. 3. (Manuscript accepted 23 November 2016)

© 2017 John Wiley & Sons Ltd, Freshwater Biology, 62, 600–614