Diversity and Distributions, (Diversity Distrib.) (2012) 18, 366–376

BIODIVERSITY Taxonomic and functional homogenization RESEARCH of an endemic desert fish fauna Thomas K. Pool* and Julian D. Olden

School of Aquatic and Fishery Sciences, ABSTRACT University of Washington, Seattle, WA 98195 Aim The highly endemic fishes of the arid Southwest USA have been heavily USA impacted by human activities resulting in one of the most threatened fish faunas in the world. The aim of this study was to examine the patterns and drivers of taxonomic and functional beta diversity of freshwater fish in the Lower Colorado River Basin across the 20th century.

Location Lower Colorado River Basin (LCRB).

Methods The taxonomic and functional similarities of watersheds were quantified to identify patterns of biotic homogenization or differentiation over the period 1900–1999. Path analysis was used to identify the relative influence of A Journal of Conservation Biogeography dam density, urban land use, precipitation regimes and non-native species richness on observed changes in fish faunal composition.

Results The fish fauna of the LCRB has become increasingly homogenized, both

taxonomically (1.1% based on bsim index) and functionally (6.2% based on Bray–Curtis index), over the 20th century. The rate of homogenization varied substantially; range declines of native species initially caused taxonomic differentiation ()7.9% in the 1960s), followed by marginal homogenization (observed in the 1990s) in response to an influx of non-native species introductions. By contrast, functional homogenization of the basin was evident considerably earlier (in the 1950s) because of the widespread introduction of non-native species sharing similar suites of biological traits. Path analysis revealed that both taxonomic and functional homogenization were positively related to the direct and indirect (facilitation by dams and urbanization) effects of non-native species richness.

Main conclusions Our study simultaneously examines rates of change in multiple dimensions of the homogenization process. For the endemic fish fauna of the LCRB, we found that the processes of taxonomic and functional homogenization are highly dynamic over time, varying both in terms of the magnitude and rate of change over the 20th century. Keywords *Correspondence: Thomas Pool, 1122 NE Boat Street, Seattle WA 98195, USA. Beta diversity, biotic homogenization, functional diversity, native fishes, E-mail: [email protected] watershed conservation. and Distributions

focus of biogeography research (Whittaker et al., 2005). INTRODUCTION Despite recent advances in this topic, there are still substantial Studying patterns of biological diversity has been the founda- gaps in our understanding of how plant and animal diversity tion of numerous ecological pursuits over the past two varies between ecosystems (i.e. beta diversity) and along centuries (Andrewartha & Birch, 1954; Stauffer, 1957). Today, environmental gradients over time (i.e. species turnover) environmental degradation coupled with the widespread (Anderson et al., 2011). Additionally, our ability to maintain

Diversity introduction of non-native species has made identifying the ecosystem functioning is hindered by our inability to predict mechanisms that preserve or erode biodiversity a primary how changes in species distributions may alter the patterns in

DOI:10.1111/j.1472-4642.2011.00836.x 366 http://wileyonlinelibrary.com/journal/ddi ª 2011 Blackwell Publishing Ltd Homogenization of an endemic fish fauna functional diversity (Condit et al., 2002; Harborne et al., 2006; investigations into the mechanisms underlining this process Melo et al., 2009; Baselga, 2010). As non-native species are scarce (but see Olden et al., 2008). continue to expand outside their historical ranges and the The aim of our paper is to study the spatiotemporal patterns distributions of native species are simultaneously reduced, of taxonomic and functional homogenization of freshwater fish enhancing our knowledge of the patterns and drivers of beta faunas in the Lower Colorado River Basin (LCRB; Fig. 1). The diversity is paramount. highly endemic fish fauna of this semi-arid basin (approximately An area of growing interest in conservation biogeography is three-quarters of the species occur only in the LCRB, Carlson & the study by which biotas increase their taxonomic, functional, Muth, 1989) continues to be impacted by human activities that or genetic similarity over time – called biotic homogenization threaten native species persistence and promote the widespread (McKinney & Lockwood, 1999; Olden, 2006). This dynamic introduction of non-native fishes (Minckley & Deacon, 1968; process depends on the identities of individual species being Olden & Poff, 2005; Rinne & Miller, 2006). Our study has two lost (e.g. native extirpation) and gained (e.g. non-native primary objectives. First, we test for taxonomic and functional introduction) across locations over time, thus making biotic homogenization of LCRB fish faunas over the last century and homogenization distinct from the wealth of studies focusing identify those species most responsible for changes in compo- on patterns and drivers of regional species richness. In a sition. As non-native species expand their range within a region landmark study, Rahel (2000) compared the similarity of and native species simultaneously experience range declines, we regional (state) fish faunas in the United States between hypothesize a transition from differentiation to homogenization present-day and pre-European settlement periods and found over time. Second, we assess the direct and indirect influence of that, on average, fish faunas have become more similar by landscape-scale factors (describing both natural and human 7.2%. Most striking was that 89 pairs of states with zero created environments) and non-native species introductions on historical similarity (no species in common) shared, on observed patterns of fish faunal homogenization. We expect that average, over 25 species; this pattern was particularly pro- non-native species may directly contribute to homogenization, nounced in desert regions of the American Southwest. Recent and factors such as increased dam density and urbanization may studies have demonstrated that taxonomic homogenization is indirectly influence community similarity by affecting patterns not an ecosystem or taxon-specific phenomenon, but occurs of invasions and extirpations. We hypothesize that increased across a broad spectrum of plants and animals (e.g. McKinney, dam density and urbanization in a watershed will be positively 2004; Rooney et al., 2004; Schwartz et al., 2006; La Sorte & associated with fish fauna homogenization (Olden et al., 2008). McKinney, 2006; Qian & Guo, 2010; reviewed in Olden, 2006). Our study is unique by providing the first simultaneous Previous studies have shed considerable insight into patterns investigation of the spatiotemporal patterns and drivers of both of biotic homogenization, but three important knowledge gaps taxonomic and functional homogenization. remain. First, research has primarily been limited to describing the magnitude and geography of change between historical and METHODS present-day composition of biotas, but the temporal dynamics of the process remain poorly understood (but see Clavero & Study area and fish inventory Garcı´a-Berthou, 2006; Spear & Chown, 2008; Taylor, 2010). Limited empirical evidence suggests that the initial conse- Our study examined distributional trends for 23 native species quence of non-native species introduction is to increase the and 39 non-native species (Table S1) according to 1,413,904 distinctiveness of biotas (Scott & Helfman, 2001); however, a fish records collected from 1900 to 1999 as part of the LCRB detailed investigation of the spatiotemporal patterns of biotic Aquatic Gap Analysis Project (Whittier et al., 2006). Following homogenization or differentiation is lacking. Second, studies to the methodology of Sowa et al. (2007), we divided the LCRB date have focused almost exclusively on changes in the into 386 watersheds that ranged from 200 to 1600 km2 in size; taxonomic composition of biotas (reviewed in Olden et al., an effective size range to represent landscape heterogeneity 2011), whereas aspects of functional or trait composition have influencing fish assemblages (Pool et al., 2010). A total of 73 received considerable less attention (but see Smart et al., 2006; watersheds fit our selection criteria by having sufficient Winter et al., 2008). Functional homogenization can occur sampling effort to capture changes in fish community com- when introduced species fill similar functional roles (e.g. position over time (i.e. minimum of three sampling events for ecological redundancy of traits into new geographical loca- each time period and a minimum of 30 total sampling events tions) or when native species with unique functional roles (e.g. from 1900 to 1999 for each watershed). Spatial and temporal trophic specialists) decline in range or abundance (McKinney heterogeneity of fish records existed in our dataset, as is & Lockwood, 1999; Olden & Rooney, 2006). The non-random common in studies examining community composition trends selection of ‘winning’ and ‘losing’ trait strategies could result in from compiled data bases (Brose et al., 2003). To account for the functional convergence of communities over time despite the effect of spatial and temporal sampling biases in our study, diverging taxonomies. Third, recent evidence points to fish records were summarized according to the following time human-induced homogenization of environmental regimes periods: 1900–1959, 1960–1969, 1970–1979, 1980–1989 and via urbanization, agricultural development, and river regula- 1990–1999; to derive representative species lists (see Olden & tion by dams (Poff et al., 2007; Winter et al., 2008); yet, formal Poff, 2005 for additional rationale). When constructing species

Diversity and Distributions, 18, 366–376, ª 2011 Blackwell Publishing Ltd 367 T. K. Pool and J. D. Olden

Figure 1 Map of the Lower Colorado River Basin displaying the major rivers and basin boundaries. lists for each time period, we assumed that if a native species output (Peters, 1983). Our ecological traits included the was recorded within a watershed in any time period, then it following: (1) water temperature preference defined as cold was also present in all earlier time periods. Conversely, if a (10–17 C), cool (18–26 C), or warm water (> 26 C) based on non-native species was recorded within a watershed, we species distributions and perceived physiological optima, and assumed it remained established in all subsequent time periods (2) trophic guild defined as omnivore (approximately < 5% (sensu Fagan et al., 2005). A ‘historical’ time period was plant matter), invertivore, invertivore–piscivore, piscivore, conservatively defined as all watershed occurrences of native herbivore–detritivore (approximately > 25% plant matter). species between 1900 and 1999. Our life-history traits included the following: (1) maximum We used a traits-based approach to test for functional total body length categorized as 0–400 mm, 401–800 mm or homogenization (Olden & Rooney, 2006). We selected two > 800 mm; (2) age at maturation categorized as 0–1 years, ecological traits and six life-history traits (converted to categor- 1.1–2 years, and > 2 years; (3) fecundity defined as total ical trait states as needed in parentheses) reflecting the ecological number of eggs or offspring per breeding season and categorized niche for each species within the LCRB. Unlike plants where the as 0–1000 or > 1000); (4) egg size defined as mean diameter of traits of a given species may be directly linked to their function mature (fully yolked) ovarian oocytes and categorized as (e.g. leaf area and photosynthetic rate), the measurable traits of 0–1.49 mm, 1.5–3 mm or > 3 mm); (5) parental care defined fish species are often one step removed from the function as the total energetic contribution of parents to their offspring performed within their environment. For example, animal body and categorized as none, low or high care (codes were 1, 2, ‡ 3 size is a universal trait correlated with growth rate, which in turn respectively following Winemiller’s (1989) 1 to 8 coding); and is associated with mortality rates, longevity, and reproductive (6) reproductive guild categorized as bearers (internal), guarders

368 Diversity and Distributions, 18, 366–376, ª 2011 Blackwell Publishing Ltd Homogenization of an endemic fish fauna

(nest spawners or substratum choosers), or non-guarders regulating river flow regimes, fragmentation of waterways, and (brood hiders or open substratum spawners) following Balon the homogenization of flow conditions (Poff et al., 2007). (1975). Trait assignments were constructed utilizing compre- Percent urban development represents the threat of short- and hensive texts of state fish faunas, primary literature, state agency long-term ecological impacts of land-use conversion to reports, university reports, and graduate theses (Olden et al., impervious surfaces (Wheeler et al., 2005). Mean annual 2006) attempting to account for trait uncertainties by using precipitation of a watershed is a principal component of local categories. climate within watersheds that influences the magnitude of runoff and degree of habitat connectivity in intermittent and ephemeral rivers that dominate the LCRB (Propst et al., 2008). Statistical analysis Data sources for these variables included the National Inven- tory of Dams (USEPA, 2006), National Land Cover Database Compositional similarity of fish faunas (MRLC, 2001), and Climate Source (USDA-NRCS 2007). Species presence/absence data were used to calculate taxo- Non-native species richness within watersheds was also used as nomic compositional similarity (CS) among watersheds across a variable in our models to assess the strength of the time periods based on the bsim index (Lennon et al., 2001). relationship between invasion and faunal homogenization. This index focuses on compositional differences by minimizing We used path analysis – a special case of structural equation the influence of variation in species richness (Koleff et al., modelling – to examine the direct and indirect relationships 2003). Similarity matrices were produced for each time period between taxonomic and functional DCS and the factors (e.g. historical, 1900–1959, 1960–1969, etc.), and the changes described earlier (Mitchell, 1993; Shipley, 2000). Traditional in pairwise bsim values were calculated as DCS = CStime period of statistical methods are unable to treat individual parameters as interest –CShistorical. When the DCS value for a watershed is both independent and dependent within the same analysis positive, it indicates biotic homogenization, whereas a negative when assessing the influence of multiple parameters in a value indicates basin differentiation (Olden & Rooney, 2006). model. Path analysis allows this type of model structure Our quantification of functional homogenization followed facilitating specific hypothesis testing, making it unique from the same approach as above. We calculated a watershed-by-trait multiple regression techniques. matrix by multiplying the watershed-by-species matrix and the We created three alternative path models for both our species-by-trait matrix and dividing by watershed richness for taxonomic and functional analysis. Each set of models each time period. The result is a watershed-by-trait matrix where consisted of a full model comparing the relative influence of each cell represents the relative proportion of species in each our predictor variables (dam number, % urban land-use, and watershed exhibiting each trait state. The functional similarity mean annual precipitation) and non-native species richness on for each watershed-by-trait matrix was calculated according to the magnitude of the taxonomic and functional DCS. In the Bray–Curtis index (CSBC), accommodating our continuous addition, we examined two nested models from each full data in this portion of our analysis. Following the same approach model to assess the relative importance of the direct and we used for the taxonomic analysis, changes in pairwise indirect effects of the predictor variables by constraining some functional compositions between watersheds for each time model paths to zero and contrasting the model path values (see period were calculated. We tested for differences in watershed Light & Marchetti, 2007). No significant collinearity of our

DCSBC and DCSsim (i.e. difference in community similarity model variables were found with all variance inflation factors between two time periods) using a Student’s one-sample t-test. (VIF) values under 5 (Menard, 1995). Model chi-squares and Species contributions associated with compositional changes associated P values were used to determine the goodness-of-fit over time were calculated by sequentially removing each species of our models. This test quantified the fit of the path structure from the data matrix and repeating the aforesaid analysis. and the probability that the observed and expected correlation Calculating the change in DCS between these analyses identified matrices differed by more than would be expected by chance individual species as having a homogenizing (positive values) or (P-values greater than 0.05 indicate a good model). Our differentiating (negative values) influence on the community analysis was based on correlation matrices with all the path composition during each study period. All analyses were coefficients in the figures being standardized values. All of our performed in R 2.51 (R Development Core Team, 2007). model analysis was performed using R 2.51 utilizing the SEM library (R Development Core Team, 2007).

Predictors of fish faunal change RESULTS We modelled taxonomic and functional DCS as a function of three variables that have been identified as important deter- Spatiotemporal patterns in taxonomic and functional minants of fish species occurrence in the LCRB (Pool et al., ) homogenization 2010). These variables included dam density (noÆkm 2), urban development (% of total area), and mean annual precipitation Fish faunas of the LCRB have become taxonomically homog- (mm) from 1971 to 2000; calculated for the upstream enized, albeit slightly, over the 20th century increasing from a contributing watershed. Dam density represents the effects of mean historical similarity among watersheds of 30.8% to the

Diversity and Distributions, 18, 366–376, ª 2011 Blackwell Publishing Ltd 369 T. K. Pool and J. D. Olden

present-day 31.9%. By contrast, the basin showed considerable t72 = )5.06, P < 0.001). The same native and non-native functionally homogenization, increasing 6.2% from a mean species that played prominent roles in driving taxonomic historical similarity among watersheds of 45.2% to the present- homogenization also shaped the functional homogenization of day 51.4%. Although taxonomic and functional homogeniza- fish faunas; native species accounted for initial watershed tion was the general pattern for the entire basin, compositional faunal similarity followed by non-native species driving faunal changes of individual watersheds were much more spatially similarity by the end of the 1990s (Table 1). variable. For example, higher than average homogenization Changes in taxonomic and functional similarity were was evident for watersheds in the geographically distant Virgin positively correlated across all time periods, with the majority (n = 7 watersheds) and Gila (n = 8 watersheds) river basins of watersheds becoming increasingly homogenized by the end (Fig. 1). Historically, these basins had only 4 native species in of the 1990s (Fig. 3). An initially weak relationship between common (Catostomus clarkii, desert sucker; Catostomus lati- taxonomic and functional homogenization of fish faunas pinnis, flannelmouth sucker; Agosia chrysogaster, longfin dace; became progressively stronger with each decade. In the Rhinichthys osculus, speckled dace), but currently they share 1950s, most watersheds were taxonomically differentiating twenty species (four native and sixteen non-native species). while functionally homogenizing (67 of 73 watersheds), but by The LCRB witnessed a dynamic change in the both the the 1990s, almost half of watersheds exhibited both forms of taxonomic and functional composition of its watersheds homogenization (31 of 73 watersheds; Fig. 3). throughout our study time period. Taxonomic similarity among watersheds with their historic faunal compositions Predictors of fish faunal change over time initially decreased by as much as 7.9% in the 1960s (i.e. differentiation), followed by a 1.1% increase in the 1990s The best fitting models for both taxonomic (v2 = 5.72, 2 (Fig. 2; 1950: t72 = 10.12, P < 0.001; 1960 t72 = 41.51, P < d.f. = 6, P = 0.45) and functional (v = 4.48, d.f. = 6,

0.001; 1970 t72 = 40.11, P < 0.001; 1980 t72 = 3.32, P < 0.001; P = 0.61) DCS revealed a positive direct effect of non-native

1990 t72 = )0.57, P = 0.57). Native species such as speckled species richness on fish faunal similarity, whereas dam number dace and desert sucker were most responsible for changes in and urbanization showed minimal direct effects but strong watershed similarity historically, but by the end of the 1990s, indirect effects (Fig. 4). The full models incorporating direct the non-native Lepomis cyanellus (green sunfish), Pimephales paths between all variables were also supported (Taxonomic promelas (fathead minnow), and Cyprinus carpio (common v2 = 4.67, d.f. = 4, P = 0.32; Functional v2 = 3.35, d.f. = 4, carp) were the most influential species driving a transition P = 0.50; Fig. S1). A reduced model excluding a linkage from taxonomic differentiation to homogenization (Table 1). between non-native species richness and DCS performed Despite the fact that the majority of non-native species were poorly (taxonomic v2 = 20.34, d.f. = 5, P < 0.01; functional present by the end of the 1960s (36 out of 39 species), v2 = 29.19, d.f. = 5, P < 0.01; Fig. S1). Dam number and taxonomic homogenization within the basin did not occur for mean annual precipitation were significantly correlated with another three decades, reflecting the considerable time course non-native species richness in all models, indicating that their over which these species spread throughout the basin. In potential role in driving faunal homogenization is likely to be contrast to patterns in taxonomy, functional homogenization indirect. was evident as early as the 1960s and continued to gradually increase to 6.2% by the turn of the century (Fig. 2; 1950: DISCUSSION t72 = )6.58, P < 0.001; 1960 t72 = )4.17, P < 0.001; 1970 t72 = )5.08, P < 0.001; 1980 t72 = )4.65, P < 0.001; 1990 In the present study, we quantified both patterns and drivers of changes in the taxonomic and functional composition of an 10 endemic fish fauna. We found that fish assemblages of the 1950 1960 1970 1980 1990 LCRB have become homogenized over the 20th century in terms of taxonomy (species composition) and function (trait 5 composition); however, the rate and magnitude of change varied substantially as have the roles of the native and non- 0 native species driving those processes. Notably, changes in species composition resulted in taxonomic differentiation

similarity (%) –5 decades preceding the eventual marginal homogenization of the fish fauna. Declining native species distributions in the Change in mean faunal middle of the century primarily drove the initial changes in –10 faunal similarity, for example speckled dace historically Figure 2 Changes in mean pairwise similarity among watersheds occurred in 83% of our study watersheds but experienced a for each time period. The bars represent the mean change in range decline of 34% by the end of the 1960s. The introduction taxonomic (grey) and functional (white) similarity of watersheds, and proliferation of non-native species mid-century continued with negative values representing overall basin differentiation and into the present day, thus playing an increased role in positive values representing basin homogenization. promoting basin homogenization. For example, non-native

370 Diversity and Distributions, 18, 366–376, ª 2011 Blackwell Publishing Ltd Homogenization of an endemic fish fauna

Table 1 The mean contribution of the top five species (native and non-native) to the homogenization of the Lower Colorado River Basin by decade.

Mean species Mean species Taxonomic Status importance (%) Functional Status importance (%)

1950 Speckled dace Native 7.0 Speckled dace Native 4.0 Desert sucker Native 2.8 Sonora sucker Native 3.5 Roundtail chub Native 1.3 Desert sucker Native 3.5 Sonora sucker Native 0.6 Roundtail chub Native 3.2 Longfin dace Native 0.4 Longfin dace Native 2.3

1960 Speckled dace Native 4.6 Speckled dace Native 3.6 Desert sucker Native 2.4 Sonora sucker Native 2.4 Sonora sucker Native 0.6 Desert sucker Native 2.1 Green sunfish Non-native 0.4 Roundtail chub Native 2.1 Roundtail chub Native 0.3 Common carp Non-native 1.7

1970 Speckled dace Native 3.0 Speckled dace Native 2.7 Desert sucker Native 1.7 Sonora sucker Native 2.1 Green sunfish Non-native 1.0 Desert sucker Native 2.0 Common carp Non-native 0.7 Fathead minnow Non-native 1.8 Rainbow trout Non-native 0.7 Roundtail chub Native 1.8

1980 Desert sucker Native 1.6 Speckled dace Native 2.7 Speckled dace Native 1.4 Fathead minnow Non-native 2.4 Green sunfish Non-native 1.4 Desert sucker Native 2.3 Common carp Non-native 1.0 Sonora sucker Native 2.2 Rainbow trout Non-native 0.9 Common carp Non-native 2.1

1990 Green sunfish Non-native 2.1 Fathead minnow Non-native 2.7 Fathead minnow Non-native 1.7 Common carp Non-native 2.5 Common carp Non-native 1.4 Green Sunfish Non-native 2.5 Desert sucker Native 1.2 Desert sucker Native 2.3 Red shiner Non-native 0.9 Speckled dace Native 2.3

Scientific names for species are provided in Table S1. fathead minnow was first introduced in the early 1950s (Olden reduction of ecosystem resistance to environmental change & Poff, 2005) and spread into 65% of our study watersheds by (Olden et al., 2004), simplification of food web structure the end of the century. In a study of the Iberian Peninsula fish (Beisner et al., 2003), and an increased potential for further faunas, Clavero & Garcı´a-Berthou (2006) found a comparable non-native species invasion (Grosholz & Tilman, 2005). In temporal transition in community similarity with initial the LCRB, we found that fish faunas were functionally differentiation followed by a homogenization in fish species homogenized to a greater extent than the taxonomic com- composition. Our results indicate that fish faunal change is position even in areas where native taxonomic richness had highly dynamic and can transition between periods of differ- not significantly changed during the 20th century. The entiation and homogenization depending on the time period functional homogenization of fishes occurred by mid-century of investigation and the relative contributions of native species when the basin was taxonomically differentiating in part loss and non-native species gain. As we look to the future of because of the widespread introduction of non-native species the LCRB, if non-native species continue to spread and native such as sunfishes (Family: Centrarchidae). The diverse array species ranges decline, we expect continued taxonomic of sunfishes that accounted for taxonomic differentiation homogenization of fish faunas unless considerable conserva- within the basin (i.e. green sunfish; bluegill; Micropterus tion action is taken to reverse this trend. salmoides, largemouth bass) also shared suites of traits (Olden Functional homogenization of regional diversity can result et al., 2006) that led to the simultaneous functional homog- in both ecological and evolutionary consequences, including a enization of fish faunas. Many of those non-native species

Diversity and Distributions, 18, 366–376, ª 2011 Blackwell Publishing Ltd 371 T. K. Pool and J. D. Olden

(a) (b) r P 1950 40 = 0.1 = 0.399 Taxonomic homogenization Taxonomic homogenization Functional differentiation Functional homogenization 30

(%) 20

Sim 10

CS 0 –30 –10–10 10 30 50

Taxonomic differentiation Taxonomic differentiation –20 Functional differentiation Functional homogenization –30 Taxonomic Δ Taxonomic Functional ΔCSBC (%)

(c) r = 0.57 P < 0.01 (d) 1970 r = 0.65 P < 0.01 1960 40 40 30 30 (%) (%) 20 20 Sim Sim 10 10

0 CS 0 Δ –30 –10–10 10 30 50 –30 –10–10 10 30 50 –20 –20 –30 –30 Taxonomic Taxonomic ΔCS Taxonomic Functional ΔCSBC (%) Functional ΔCSBC (%)

(e) 1980 r = 0.69 P < 0.01 (f) 1990 r = 0.72 P < 0.01 40 40 30 30 (%) 20 (%) 20 Sim 10 Sim 10

CS 0 0 –30 –10–10 10 30 50 –30 –10–10 10 30 50 –20 –20 –30 –30 Taxonomic Δ Taxonomic Taxonomic ΔCS Taxonomic Functional ΔCSBC(%) Functional ΔCSBC (%)

Figure 3 Bivariate relationship between DCSBC and DCSsim compositional similarity of individual watersheds between historical and intermediate time periods. Each quadrant represents a different combination of taxonomic and functional change of each watersheds fish fauna (a). Each intermediate time period is represented by watersheds mean change in similarity with each watersheds historical watershed compositions (b–f). were introduced to the LCRB to promote recreational fishing The link between taxonomic and functional fish homogeni- and subsequently spread to become widely distributed zation has proven to be temporally dynamic, but several clear throughout the basin. Similarly, Smart et al. (2006) found trends emerged from our analysis. Despite the weak association that plant functional composition in Britain increased over a between taxonomic and functional community similarity in the 20-year period, while taxonomic similarity concurrently 1950s and 1960s, by the end of our study we found that the declined. Some of the non-native species introduced into majority of watersheds exhibiting taxonomic homogenization the LCRB were also functionally redundant with native were similarly homogenized in terms of their trait composition. species occupying the basin (Olden et al., 2006). The non- This relationship suggests that correlated patterns of taxonomic native Gambusia affinis (western mosquitofish), possessing and functional homogenization may become increasingly traits similar to that of the native Poeciliopsis occidentalis (Gila predictable over time. A recent study by Baiser & Lockwood topminnow), established in many areas that historically did (2011) used simulated data to show that an increasing ratio of not support the native topminnow, further increasing the trait states to species richness strengthened the correlation functional similarity between formerly distinct faunas. Inter- between the taxonomic and functional similarity of bird estingly, by the end of the 1990s, the widespread establish- assemblages in the USA. These findings support our results; ment of a few functionally distinct non-native species in that, the weak relationship between fish taxonomic and (fathead minnow, common carp, and green sunfish) played functional similarity in the early time periods may have been the dominate roles in driving both the taxonomic and driven by the trait redundancy associated with introduced sport functional homogenization of the basin. fishes within the basin. The introduction and proliferation of

372 Diversity and Distributions, 18, 366–376, ª 2011 Blackwell Publishing Ltd Homogenization of an endemic fish fauna

(a) dominating role in the homogenization of fish faunas at a large R2 = 0.267 Dam density 0.196 spatial scale. Prior research also suggests that dams in the Non-native LCRB promote non-native species by altering downstream 0.144 0.127 richness % Urban river flows (Propst & Gido 2004; Propst et al., 2008) and by –0.151 land-use supporting enriched non-native communities in impounded 0.443 0.020 reservoirs (Johnson et al., 2008). The work of Olden et al. R2 = 0.227 Annual (2008) similarly identified a suite of human disturbance precipitation 0.480 Taxonomic ΔCS factors, including dams, as drivers of fish faunal homogeniza- tion in Australian drainages. Non-native species may also contribute directly to fish faunal homogenization by spreading (b) R2 = 0.267 throughout the basin predating on and competing with native Dam density 0.196 Non-native fishes (Carpenter & Mueller, 2008). While our analysis 0.144 0.127 richness identified non-native species richness and dam density as % Urban –0.151 land-use important drivers of fish faunal homogenization in the LCRB, 0.567 0.020 further work incorporating additional ecological threats R2 = 0.333 Annual (Paukert et al., 2011) may facilitate a better understanding of 0.480 Functional precipitation ΔCS the full scope of factors influencing the homogenization process. Understanding the patterns and drivers of species turnover Figure 4 Path diagrams displaying the effects of dam density, in a rapidly changing environment is important for conserva- urbanization, mean annual precipitation and non-native species tion and restoration efforts (Whittaker et al., 2005; Hobbs taxonomic richness (black boxes) on the (a) taxonomic DCS and (b) functional DCS (white boxes) of Lower Colorado River Basin et al., 2006). Our study highlights the dynamic nature of the fish faunas. The values near each path are the standardized coef- homogenization process by showing that the rate of fish ficient estimates displaying the relative strength of each relation- compositional change may vary over long time-scales. Recog- ship, also represented by line thickness [coefficient estimates nition that the taxonomic and functional homogenization of cannot be compared between two non-hierarchical models such as faunas may be linked but are not synonymous with each other, model (a) and (b)]; with variation explained (R2) being reported particularly for areas in the early stages of environmental for non-native taxonomic richness and watershed DCS in each change and species invasion, emphasizes the need to focus on model. Solid arrows represent significant path coefficients both taxonomic diversity and ecosystem functioning as (P < 0.05), and dashed lines represent insignificant coefficients. conservation priorities (McGill et al., 2006). While a majority of studies to date have identified low levels of taxonomic non-native fishes with more diverse trait profiles in the latter homogenization within faunas, we have shown that significant part of the century increased the number of trait states, and thus changes in community structure can occur at time steps contributed to the strengthening of the correlation between typically not captured in historic to contemporary faunal patterns of taxonomic and functional homogenization. Ulti- comparisons. Furthermore, if freshwater environments con- mately, in areas where there are highly selective introductions of tinue to be altered in ways that favour the spread of non-native non-native species (e.g. functionally similar sport fishes) species and hinder endemic native species, additional homog- followed by less selective introductions of species (e.g. func- enization of regional biotas should be expected. Although the tionally diverse aquarium fishes), we contend the relative specific conservation implications of biotic homogenization taxonomic and functional similarity of faunas may provide are still emerging (Rooney et al., 2007), we believe that some insight into the stage of homogenization for that system. managing for native species that are highly imperilled and Multiple environmental drivers associated with land-use functionally distinct should be prioritized in addition to the change (urban, agriculture) and hydrological alteration are protection of watersheds that are minimally invaded (Strecker known to influence fish species composition in aquatic et al., in press). Future studies that identify prioritization areas ecosystems, yet the manner in which these factors directly or that may mitigate the homogenizing effect of non-native indirectly influence the homogenization process remains species are also needed to curtail further simplification of unclear (Marchetti et al., 2006; Olden et al., 2008). We found ecosystems. that non-native species richness directly and dam density indirectly (through non-native species) promoted taxonomic ACKNOWLEDGEMENTS and functional homogenization. These results indicate that factors associated with dams such as habitat fragmentation and The authors graciously acknowledge the comments of three altered flow regimes promote the establishment and spread of anonymous reviewers, which improved the final paper. Fund- non-native species, which subsequently contribute to the ing support was provided by the USGS National Gap Analysis homogenization process. Rahel (2000) examined taxonomic Program and the USGS Status and Trends Program. We also changes in fish composition across the USA and found thank Joanna Whittier and Craig Paukert for their contribu- comparable results indicating that non-native species play a tions to the LCRB Aquatic Gap Analysis Project.

Diversity and Distributions, 18, 366–376, ª 2011 Blackwell Publishing Ltd 373 T. K. Pool and J. D. Olden

Novel ecosystems: theoretical and management aspects of REFERENCES the new ecological world order. Global Ecology and Bioge- Anderson, M.J., Crist, T.O., Chase, J.M., Vellend, M., Inouye, ography, 15, 1–7. B.D., Freestone, A.L., Sanders, N.J., Cornell, H.V., Comita, Johnson, P.T.J., Olden, J.D. & Vander Zanden, M.J. (2008) L.S., Davies, K.F., Harrison, S.P., Kraft, N.J.B., Stegen, J.C. & Dam invaders: impoundments facilitate biological invasions Swenson, N.G. (2011) Navigating the multiple meanings of b into freshwaters. Frontiers in Ecology and the Environment, 6, diversity: a roadmap for the practicing ecologist. Ecology 357–363. Letters, 14, 19–28. Koleff, P., Gaston, K.J. & Lennon, J.J. (2003) Measuring beta Andrewartha, H.G. & Birch, L.C. (1954) The distribution and diversity for presence-absence data. Journal of Animal Ecol- abundance of animals. University of Chicago Press, Chicago, ogy, 72, 367–382. IL. La Sorte, F.A. & McKinney, M.L. (2006) Compositional sim- Baiser, B. & Lockwood, J.L. (2011) The relationship between ilarity and the distribution of geographical range size for functional and taxonomic homogenization. Global Ecology assemblages of native and non-native species in urban floras. and Biogeography, 20, 134–144. Diversity and Distributions, 12, 679–686. Balon, E.K. (1975) Reproductive guilds of fishes – Proposal Lennon, J.J., Koleff, P., Greenwood, J.D. & Gaston, K.J. (2001) and definition. Journal of the Fisheries Research Board of The geographical structure of British bird distributions: Canada, 32, 821–864. diversity, spatial turnover and scale. Journal of Animal Baselga, A. (2010) Multiplicative partition of true diversity Ecology, 70, 966–979. yields independent alpha and beta components; additive Light, T. & Marchetti, M.P. (2007) Distinguishing between partition does not. Ecology, 91, 1974–1981. invasions and habitat changes as drivers of diversity loss Beisner, B.E., Ives, A.R. & Carpenter, S.R. (2003) The effects of among California’s freshwater fishes. Conservation Biology, an exotic fish invasion on the prey communities of two lakes. 21, 434–446. Journal of Animal Ecology, 72, 331–342. Marchetti, M.P., Lockwood, J.L. & Light, T. (2006) Effects of Brose, U., Martinez, N.D. & Williams, R.J. (2003) Estimating urbanization on California’s fish diversity: differentiation, species richness: sensitivity to sample coverage and insensi- homogenization and the influence of spatial scale. Biological tivity to spatial patterns. Ecology, 84, 2364–2377. Conservation, 127, 310–318. Carlson, C.A. & Muth, R.T. (1989) The Colorado River: lifeline McGill, B.J., Enquist, B.J., Weiher, E. & Westoby, M. (2006) of the American Southwest. Proceedings of the international Rebuilding community ecology from functional traits. large river symposium, Canadian Special Publication of Trends in Ecology and Evolution, 21, 178–185. Fisheries and Aquatic Sciences, 106, 220–239. McKinney, M.L. (2004) Measuring floristic homogenization by Carpenter, J. & Mueller, G.A. (2008) Small nonnative fishes as non-native plants in North America. Global Ecology and predators of larval razorback suckers. The Southwestern Biogeography, 13, 47–53. Naturalist, 53, 236–242. McKinney, M.L. & Lockwood, J.L. (1999) Biotic homogeni- Clavero, M. & Garcı´a-Berthou, E. (2006) Homogenization zation: a few winners replacing many losers in the next mass dynamics and introduction routes of invasive freshwater fish extinction. Trends in Ecology and Evolution, 14, 450–453. in the Iberian Peninsula. Ecological Applications, 16, 2313– Melo, A.S., Rangel, T. & Diniz, J.A.F. (2009) Environmental 2324. drivers of beta-diversity patterns in New-World birds and Condit, R., Pitman, N., Leigh, E.G., Chave, J., Terborgh, J., mammals. Ecography, 32, 226–236. Foster, R.B., Nunez, P., Aguilar, S., Valencia, R., Villa, G., Menard, S. (1995) Applied logistic regression analysis: Sage Muller-Landau, H.C., Losos, E. & Hubbell, S.P. (2002) Beta- University Series on Quantitative Applications in the Social diversity in tropical forest trees. Science, 295, 666–669. Sciences. Sage, Thousand Oaks, CA. Fagan, W.F., Kennedy, C.M. & Unmack, P.J. (2005) Quanti- Minckley, W.L. & Deacon, J.E. (1968) Southwestern fishes and fying rarity, losses, and risks for native fishes of the lower enigma of endangered species. Science, 159, 1424–1432. Colorado River Basin: implications for conservation listing. Mitchell, R.J. (1993) Path analysis: Pollination. In: Design and Conservation Biology, 19, 1872–1882. Analysis of Ecological Experiments. (ed. by Scheiner S.M. & Grosholz, E.D. & Tilman, G.D. (2005) Recent biological Gurevitch J.), pp. 211–231. Chapman & Hall, New York. invasion may hasten invasional meltdown by accelerating Multi-Resolution Land Characteristics Consortium – National historical introductions. Proceedings of the National Academy Land Cover Database [MRLC]. (2001). EROS Data Center. of Sciences USA, 102, 1088–1091. Available at: http://www.mrlc.gov/index.php (accessed 1 Harborne, A.R., Mumby, P.J., Zychaluk, K., Hedley, J.D. & February 2009). Blackwell, P.G. (2006) Modeling the beta diversity of coral Olden, J.D. (2006) Biotic homogenization: a new research reefs. Ecology, 87, 2871–2881. agenda for conservation biogeography. Journal of Biogeog- Hobbs, R.J., Arico, S., Aronson, J., Baron, J.S., Bridgewater, P., raphy, 33, 2027–2039. Cramer, V.A., Epstein, P.R., Ewel, J.J., Klink, C.A., Lugo, Olden, J.D. & Poff, N.L. (2005) Long-term trends of native and A.E., Norton, D., Ojima, D., Richardson, D.M., Sanderson, non-native fish faunas in the American Southwest. Animal E.W., Valladares, F., Vila, M., Zamora, R. & Zobel, M. (2006) Biodiversity and Conservation, 28, 75–89.

374 Diversity and Distributions, 18, 366–376, ª 2011 Blackwell Publishing Ltd Homogenization of an endemic fish fauna

Olden, J.D. & Rooney, T.P. (2006) On defining and quanti- Rooney, T.P., Olden, J.D., Leach, M.K. & Rogers, D.A. (2007) fying biotic homogenization. Global Ecology and Biogeogra- Biotic homogenization and conservation prioritization. phy, 15, 113–120. Biological Conservation, 134, 447–450. Olden, J.D., Poff, N.L., Douglas, M.R., Douglas, M.E. & Schwartz, M.W., Thorne, J.H. & Viers, J.H. (2006) Biotic Fausch, K.D. (2004) Ecological and evolutionary conse- homogenization of the California flora in urban and quences of biotic homogenization. Trends in Ecology and urbanizing regions. Biological Conservation, 127, 282–291. Evolution, 19, 18–24. Scott, M.C. & Helfman, G.S. (2001) Native invasions, Olden, J.D., Poff, N.L. & Bestgen, K.R. (2006) Life-history homogenization, and the mismeasure of integrity of fish strategies predict fish invasions and extirpations in the assemblages. Fisheries, 26, 6–15. Colorado River Basin. Ecological Monographs, 76, 25–40. Shipley, B. (2000) Cause and correlation in biology: a user’s Olden, J.D., Kennard, M. & Pusey, B. (2008) Species Invasions guide to path analysis, structural equations and causal infer- and the changing biogeography of Australian Freshwater ence. Cambridge University Press, Cambridge. Fishes. Global Ecology and Biogeography, 17, 25–37. Smart, S.M., Thompson, K., Marrs, R.H., Le Duc, M.G., Olden, J.D., Lockwood, J.L. & Parr, C.L. (2011) Species inva- Maskell, L.C. & Firbank, L.G. (2006) Biotic homogenization sions and the biotic homogenization of faunas and floras. and changes in species diversity across human-modified Conservation biogeography (ed. by R.J. Whittaker and R.J. ecosystems. Proceedings of the Royal Society B: Biological Ladle), pp. 224–243. Wiley-Blackwell, Oxford. Sciences, 273, 2659–2665. Paukert, C.P., Pitts, K.L., Whittier, J.B. & Olden, J.D. (2011) Sowa, S.P., Annis, G., Morey, M.E. & Diamond, D.D. (2007) A Development and assessment of a landscape-scale ecological gap analysis and comprehensive conservation strategy for threat index for the Lower Colorado River Basin. Ecological riverine ecosystems of Missouri. Ecological Monographs, 77, Indicators, 11, 304–310. 301–334. Peters, R.H. (1983) The ecological implications of body size. Spear, D. & Chown, S.L. (2008) Taxonomic homogenization in Cambridge University Press, Cambridge. ungulates: patterns and mechanisms at local and global Poff, N.L., Olden, J.D., Merritt, D.M. & Pepin, D.M. (2007) scales. Journal of Biogeography, 35, 1962–1975. Homogenization of regional river dynamics by dams and Stauffer, R.C. (1957) Haeckel, Darwin and ecology. Quarterly global biodiversity implications. Proceedings of the National Review of Biology, 32, 138–144. Academy of Sciences USA, 104, 5732–5737. Strecker, A.L., Olden, J.D., Whittier, J.B. & Paukert, C.P. Pool, T.K., Olden, J.D., Whittier, J.B. & Paukert, C.P. (2010) Defining conservation priorities for freshwater fishes Environmental drivers of fish functional diversity and according to taxonomic, functional, and phylogenetic composition in the Lower Colorado River Basin. Canadian diversity. Ecological Applications. doi:10.1890/11-0599.1. Journal of Fisheries and Aquatic Sciences, 67, 1791–1807. Taylor, E.B. (2010) Changes in taxonomy and species distri- Propst, D.L. & Gido, K.B. (2004) Responses of native and butions and their influence on estimates of faunal homog- nonnative fishes to natural flow regime mimicry in the San enization and differentiation in freshwater fishes. Diversity Juan River. Transactions of the American Fisheries Society, and Distributions, 16, 676–689. 133, 922–931. United States Department of Agriculture – National Resources Propst, D.L., Gido, K.B. & Stefferud, J.A. (2008) Natural Conservation Service [USDA] (2007). PRISM Climate flow regimes, nonnative fishes, and native fish persistence Mapping Project–Precipitation. Mean monthly and annual in arid-land river systems. Ecological Applications, 18, precipitation digital files for the continental U.S. USDA-NRCS 1236–1252. National Cartography and Geospatial Center, Ft. Worth, TX. Qian, H. & Guo, Q.F. (2010) Linking biotic homogenization to Available online at http://www.ncgc.nrcs.usda.gov/products/ habitat type, invasiveness and growth form of naturalized datasets/climate/. alien plants in North America. Diversity and Distributions, U.S. Environmental Protection Agency [USEPA] (2006) 16, 119–125. National Inventory of Dams. U.S. Environmental Protection R Development Core Team (2007) R: a language and envi- Agency [USEPA], Washington, DC. Available online at: ronment for statistical computing. R Foundation for Statistical http://www.nicar.org/data/dams. Computing, Vienna, Austria. Available at: http://www. Wheeler, A.P., Angermeier, P.L. & Rosenberger, A.E. (2005) R-project.org. Impacts of new highways and subsequent landscape urban- Rahel, F.J. (2000) Homogenization of fish faunas across the ization on stream habitat and biota. Reviews in Fisheries United States. Science, 288, 854–856. Science, 13, 141–164. Rinne, J. & Miller, D. (2006) Hydrology, geomorphology and Whittaker, R.J., Arau´ jo, M.B., Paul, J., Ladle, R.J., Watson, management: implications for sustainability of native J.E.M. & Willis, K.J. (2005) Conservation Biogeography: southwestern fishes. Reviews in Fisheries Science, 14, 91–110. assessment and prospect. Diversity and Distributions, 11,3– Rooney, T.P., Wiegmann, S.M., Rogers, D.A. & Waller, D.M. 23. (2004) Biotic impoverishment and homogenization in Whittier, J.B., Paukert, C.P. & Gido, K. (2006) Development of unfragmented forest understory communities. Conservation an aquatic GAP for the Lower Colorado River Basin. Gap Biology, 18, 787–798.

Diversity and Distributions, 18, 366–376, ª 2011 Blackwell Publishing Ltd 375 T. K. Pool and J. D. Olden

Analysis Bulletin No. 14. USGS/BRD/Gap Analysis Program, BIOSKETCHES Moscow, Idaho. Winemiller, K.O. (1989) Patterns of variation in life-history Thomas K. Pool is a PhD candidate in the School of Aquatic among South-American fishes in seasonal environments. and Fishery Sciences at the University of Washington. His Oecologia, 81, 225–241. research focuses on using trait-based and taxonomic Winter, M., Kuhn, I., Nentwig, W. & Klotz, S. (2008) Spatial approaches to identify biotic and abiotic drivers of non-native aspects of trait homogenization within the German flora. species invasion. Specifically, he is interested in the conserva- Journal of Biogeography, 35, 2289–2297. tion biogeography of arid system fishes and developing predictive models that can be applied towards native fish SUPPORTING INFORMATION conservation efforts.

Additional Supporting Information may be found in the online Julian D. Olden is an Assistant Professor in the School of version of this article: Aquatic and Fishery Sciences at the University of Washington. His research focuses on the ecology and management of Figure S1 Path diagrams displaying the full model (a), the invasive species, environmental flows, biogeography of fresh- nested passenger model (b) and the nested driver model (c). water fishes, and the development of conservation strategies in Table S1 Native (n = 21) and non-native (n = 36) fish in the natural and built environments. Lower Colorado River Basin that were examined in our study. Author contributions: Both authors contributed ideas that led As a service to our authors and readers, this journal provides to the manuscript concept; Thomas Pool led the data analysis supporting information supplied by the authors. Such mate- and both authors contributed significantly to the writing of the rials are peer-reviewed and may be re-organized for online paper. delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing Editor: Anthony Ricciardi files) should be addressed to the authors.

376 Diversity and Distributions, 18, 366–376, ª 2011 Blackwell Publishing Ltd