Watersheds in southern Ontario are of high conservation concern due to their diverse fish communities, productive environments, and threats from numerous anthropogenic stressors. The Credit River watershed, located west of the Greater Toronto Area, has over 60 fish species, and multiple stressors including urbanization, climate change, and aquatic invasive species. This study examines fish community change in the Credit River watershed. Historical fish datasets collected in the watershed from 1954 to 2015 were analyzed to examine richness patterns, temporal trends in species distributions, and faunal similarity at the site and sub-watershed levels. Species richness increased over time at the site and sub-watershed level, displaying predictable richness patterns due to anthropogenic introductions and changes in sampling methods. Species distribution patterns remained largely stable over time with decreases in some species (e.g. Redside Dace) and increases in others (e.g. Largemouth Bass). Faunal similarity also increased over time at the site and sub-watershed level, indicating that the fish communities in the Credit River are homogenizing.
Freshwater ecosystems support a high proportion of global biodiversity. Despite this, anthropogenic processes, such as climate change and habitat destruction, have significantly disrupted freshwater ecosystems (Abell et al., 2008). Freshwater fishes are particularly susceptible to ecosystem disruption due to specific habitat requirements such as seasonal migratory corridors or specific water-quality requirements (Abell, 2002). Currently, North American freshwater fishes are declining in distribution and abundance at faster rates than tropical forests (Rasmussen and Ricciardi, 1999; Jelks et al., 2008).
Similar patterns of biodiversity loss are found within Canada today with approximately 30% of Canadian freshwater fish at risk of extinction (Mandrak, unpublished data). Provincially, Ontario has the highest freshwater fish diversity in Canada, containing 128 fish species and 17 established invasive fish species (Holm et al., 2009). Ontario is also the most populous province and is, therefore, more susceptible to freshwater fish biodiversity loss and community change (Staton and Mandrak, 2006; Chu et al., 2015). Thus, there is a need to understand the relationship between long-term freshwater fish community patterns and a changing climate and increasing anthropogenic stressors.
Understanding the current state of freshwater fish biodiversity requires an examination of community change over time. Fish communities are susceptible to natural and anthropogenic change over time. These disturbances can result in fluctuations in species abundance, changes in species distribution, or loss of rare native species (Bronte et al., 2003). For example, Gido et al. (2010) found that, due to anthropogenic water and land-use changes, the hydrologic regime of many Great Plain streams had substantially changed over the last century, resulting in fish community change. Moreover, fish composition in these streams shared less than 50% of species originally found in historical collections, with the greatest changes occurring in basins most fragmented by reservoirs. Anthropogenic disturbances causing fish community change are also present within the Great Lakes. Trebitz et al. (2009) found that anthropogenic stressors structured fish composition by impacting water quality and aquatic vegetation structure along Great Lakes coastal wetlands. Furthermore, the study determined that these stressors led to the homogenization of fish communities.
Homogenization is another emerging form of community change. It is a process by which regional communities become more similar in composition over time (Taylor, 2004). This process stems from increased introductions of non-native species and the extirpation or extinction of native species. Rahel (2000) determined that widespread introductions of non-native species increased the similarity in fish community composition across the United States. Specifically, he showed that, from European settlement to 2000, 89 pairs of states with no species in common now have an average of 25.2 species shared. Taylor (2004) examined homogenization in Canadian freshwater fish faunas and found low, but significant, homogenization across Canada, with British Columbia displaying the highest level. Moreover, Taylor (2010) found that, when comparing Canadian fish communities from 2000 to 2005, faunal homogenization increased across Canada by 1.8%. Currently, understanding fish faunal homogenization is considered one of the greatest challenges to the conservation of freshwater fishes (Olden et al., 2010).
Community change can be best examined using long-term monitoring as it can describe the structure of ecological communities over time in response to potential drivers of change (Elliot, 1990). Long-term monitoring is also necessary for predicting and managing the future of fish communities (Gido et al., 2000). Moreover, if available, using historical datasets for long-term monitoring can be the most effective way to gain insight into the composition and drivers of a biological community (Shaffer et al., 1998). Despite potential differences in sampling methods and effort, Shaffer et al. (1998) concluded that using historical presence-absence data could provide vital knowledge on identifying species declines or community composition over time. Currently, the majority of historical analyses performed on freshwater fish communities in Canada have been undertaken at limited biological, spatial, and temporal scales (e.g. Steedman, 1988; Wichert, 1995; Mandrak, 1995). Therefore, there is currently an information gap in long-term Canadian freshwater fish patterns and processes at multiple scales.
Chu et al. (2015) classified watersheds in southern Ontario, including the Credit River, as having the greatest conservation priority in Canada because of their diverse fish communities, productive environments, and threats from numerous anthropogenic stressors. The Credit River watershed is one of high conservation concern (Chu et al., 2015). This watershed has over 63 fish species including the Endangered Redside Dace (Clintostomus elongatus) and the invasive Round Goby (Neogobius melanostomus) (Appendix). The 23 sub-watersheds under the jurisdiction Credit Valley Conservation (CVC), 21 in the Credit River watershed and two adjacent watersheds draining directly into Lake Ontario, have been under increasing urbanization pressure and also face other stressors including climate change and aquatic invasive species (AIS). Past studies have focused on Credit River fishes and associated anthropogenic effects (Martin, 1984), determining the biotic integrity of the watershed (Steedman, 1988), and the effects of individual stressors, such as sewage effluent, on fish communities (Wichert, 1995). However, no studies have examined long-term changes in the fish communities of the Credit River.
This study focuses on describing fish community change in the Credit River watershed. Historical fish datasets from the Credit River watershed were compiled over a 63-year time period, 1954-2015. These datasets were used to examine patterns of fish community change, including homogenization of fish communities and richness patterns at site and sub-watershed levels. Such information is useful to make sound management decisions and inform policy direction.
The Credit River watershed is located in southern Ontario west of Toronto (Figure 1). The river stretches 90 kilometers, meandering from the relatively natural headwaters in Orangeville southeast towards highly urbanized areas, including Brampton and Mississauga, before draining into Lake Ontario. Approximately 1500 kilometers of streams are present in its 21 sub-watersheds and two adjacent subwatersheds (CVC, 2015). The Credit River watershed is a very diverse landscape, consisting of urban forested and agricultural areas.
Fish community data
Historical fish community datasets for the Credit River watershed were obtained from multiple sources (Table 1). Data for the 1990s were excluded from the study due to limited and non-comparable datasets (i.e. different sites from other time periods, small sample size). Each data source was assigned to one of six time periods. The number of sites sampled in each time period varied from approximately 60 to over 200 and the number of sub-watersheds sampled ranged from 15 to 23 (Table 1). Fishes were sampled by seining (1954, 1965) and backpack electrofishing (1975, 1984-1987, 1999-2006, 2007-2015). Gear type, effort, and abundance by species were not readily available for all the datasets; therefore, all analyses were based on presence-absence data.
For each time period, species presence-absence matrices were constructed by site and sub-watershed. Sites sampled in different time periods were considered the same site if found within a 350 m buffer of each other in the same waterbody and stream order. To develop sub-watershed presence-absence matrices, sites were assigned to a sub-watershed using a geographic information system.
Fish community change
To determine temporal trends in species richness, mean richness was summarized at site and sub-watershed levels. To examine the relationship between species richness and time periods, Kruskal-Wallis and pairwise Tukey and Kramer tests used to identify pair-wise differences between periods as the data were not normally distributed. The relationship between species richness and distance from the Credit River mouth was assessed. To test the primary and interaction effects between richness and distance from mouth for each time period, an analysis of covariance (ANCOVA) was used and a post-hoc Tukey test used to identify differences among periods. Species area-curves were created to examine the relationship between species richness and sub-watershed area. To test the primary interaction effects between richness and area for each period, an ANCOVA was used and a Tukey test was used to identify differences among periods. ANCOVAs and Tukey tests were conducted using R-software. To account for potential sampling bias on the effect of the number of sites sampled on richness, Pearson and partial correlations were used to test the strength of the relationship between number of sites sampled and richness, controlling for sub-watershed area.
To assess temporal trends in distribution across sub-watersheds for each species, the slope of the relationship between the proportion of the 23 sub-watersheds in which the species was found and the time period was calculated (Buckwalter et al., 2018). Based on the slopes, species were assigned to one of five temporal distribution categories: spreaders, decliner-limited, stable-widespread, stable-limited, and single occurrence (Table 2). Due to the small sample size (six time periods), the slopes could not be tested for significant difference from a slope of 0.
Change and homogenization in fish communities over time were examined. Change in fish communities was measured in two ways: an inter-decadal pairwise analysis and a pairwise historical analysis (Table 3). Both analyses calculate the mean similarity between the same sites and sub-watersheds across time periods. An inter-decadal analysis compares changes in fish communities, at the same site or watershed, between subsequent time periods (e.g. 1 vs. 2, 2 vs. 3). A historical analysis compares changes in fish communities at the same site or watershed between each time period and the original period (e.g. 1 vs. 2, 1 vs. 3). Homogenization was measured as the mean similarity of fish communities separately for each time period for sites and sub-watersheds (Rahel, 2000, Taylor, 2004, Taylor, 2010). The mean similarity represents the mean community similarity across all sites or sub-watersheds for each individual period. To examine the relationship between mean similarity and time periods, ANOVA and pairwise Tukey tests were used to identify pair-wise differences between periods for the sub-watershed data as it was normally distributed. Kruskal-Wallis and pairwise Tukey and Kramer tests used to identify pair-wise differences between periods for the site data as they were not normally distributed. Statistical analyses were completed using the Ade-4 and LAVAAN packages in R-.
Over the 63-year time period, a total of 63 fish species were recorded in the Credit River watershed (Appendix). In total, fishes were collected at over 500 sites across all 23 sub-watersheds. Of the 63 species, 10 species are not native, including the Common Carp (Cyprinus carpio), Goldfish (Carassius auratus), and Round Goby (Neogobius meanostomus) (Appendix). Other notable species recorded included the Endangered Redside Dace, reintroduced Atlantic Salmon (Salmo salar), and coldwater Brook Trout (Salvelinus fontinalis).
At both site and sub-watershed levels, mean species richness generally increased over time. At the site level, mean species richness decreased between period one and two and between period five and six despite increasing throughout all other periods (Figure 2) The increase was significant over time (Kruskal-Wallis chi-squared = 197.7, p < 0.05), with mean species richness in periods 1 and 2, periods 3 and 4, and periods 5 and 6 not significantly different from one another (Figure 2) At the sub-watershed level, mean species richness decreased between period one and two, then increased during the remaining periods (Figure 3). The change in species richness was significant over time (Kruskal-Wallis chi-squared = 27.279, p < 0.05), with more pairwise complex interactions; in general, species richness was similar for the first three periods and the second three periods (Figure 3).
At the site level, species richness decreased significantly with distance from mouth for each time period (F = 65.199, p < 0.05) and differed significantly across time periods (F = 69.589, p < 0.05) (Figure 4). Based on Tukey tests, there were no significant differences in species richness between periods one and two, one and three, three and four, and five and six. Across all time periods, species richness increased significantly with watershed area (F = 86.2, p < 0.05) (Figure 5). Species richness differed significantly across (F = 7.2, p < 0.05), but not between (F = 1.999, p = 0.08), time periods. For all six time periods, species richness was correlated to the number of sites (Table 4). When controlling for area using a partial correlation, there was no significant relationship between species richness and number of sites sampled for four (1-2, 5-6) of the six periods (Table 4).
Based on the slopes of the number of watersheds by decade relationship, the distributions of 68% of species were stable-limited, 13% spreaders, 11% stable-widespread, 5% single occurrence, and 3% decliner-limited (Appendix).
The inter-decadal pairwise analyses based on sites and sub-watersheds both displayed positive slopes, indicating an increase in fish faunal homogenization over time. At the site level, community similarity generally increased, except between periods 4 and 5 (Figure 6a). At the sub-watershed level, community similarity generally increased, except between periods two and three (Figure 6b). The historical pairwise comparison analyses displayed a positive slope at the site-level, indicating an increase in community similarity over time, and a negative slope at the sub-watershed level, indicating a decrease in community similarity over time. At the site-level, community similarity increased for every time period, except period five (Figure 6c). At the sub-watershed level, community similarity increased during the first three time periods, then decreased during the last two periods (Figure 6d).
Homogenization, based on the mean similarity for all possible combinations within each time period, displayed a positive slope and significant differences between time periods, indicating an increase in homogenization over time (Figure 6e) at the site (F = 7.107, p < 0.05) and sub-watershed (Kruskal-Wallis chi-squared = 3018.668, p < 0.05) levels. At the site level, homogenization increased between periods 1 and 3, was the same for periods 4 and 6, and highest for period 5 (Figure 6e). At the sub-watershed level, homogenization was the same for periods 1-3, 5 and 6, and higher for period 4, which was not significantly different than period 2 (Figure 6f).
Over time, the fish communities in the Credit River have changed, have become more homogenized, and display predictable richness. Despite variation in effort and number of sites and sub-watersheds sampled, the Credit River watershed demonstrated predictable species richness patterns over the 63-year time scale. At both site and sub-watershed levels, mean species richness generally increased from time period one to six. Anthropogenic introductions of non-native species explain the increased richness. The highest increase in richness occurred between periods three and five. During this time, multiple introductions occurred in the Credit River, including Round Goby and Common Carp. These introductions occurred due to anthropogenic processes. The Round Goby was introduced through ballast-water transfer from Ponto-Caspian ports and has secondarily spread throughout the Great Lakes (Ricciardi and MacIassac, 2000), including to the mouth of the Credit River. In addition, Round Goby was introduced into the headwaters of the West Credit River, likely through bait-bucket release (CVC, 2015). Finally, Atlantic Salmon stocking began in 2006, while stocking of non-native sportfishes, such as Chinook Salmon (Onocrhynchus tshawytscha) and Rainbow Trout (Oncorhynchus mykiss) has been ongoing for over 30 years (CVC, 2015). In addition, the increased richness could be influenced, in part, by differences in gear type used over time from seining (periods 1-2) to electrofishing (periods 3-6). Electrofishing has produced higher estimates of species richness and diversity than seining in a montane river in Mexico, indicating that seining alone does not provide a complete perspective on fish communities (Mercado-Silva and Escandon-Sandoval, 2008). Furthermore, Poos et al. (2007) found that electrofishing was more effective at sampling harder to detect species, such as species at risk, compared to seining. For example, native fishes, such as the Finescale Dace (Chrosomus neogaeus), Spotfin Shiner (Cyprinella spiloptera), and Iowa Darter (Etheostoma exile), were not detected until period three, when electrofishing was first used.
For all six time periods, species richness significantly declined as distance of the site from the Credit River mouth increased and sub-watershed area decreased. This trend is consistent with the river continuum concept, which states that river systems have predictable characteristics and operate to conform to the most probable position of the system (Vannote et al., 1980). Additionally, many of the headwater streams are classified as coldwater streams that are generally low in species diversity, including the Brook Trout (CVC, 2015). Sampling bias was assessed by examining species-area relationships at the sub-watershed level. If richness was significantly correlated to number of sites sampled in a sub-watershed, sampling bias would be present. For periods (1-2, 5-6), the number of sites sampled had no effect on species richness when controlling for sub-watershed area. The sampling bias found in period three and four may be the result of fewer sites sampled compared to the other periods.
The 63 fish species captured over the six time periods generally displayed one of five temporal trends in species distribution at the sub-watershed level. The majority of the species captured had a stable-limited distribution (68%). These species include habitat-specific species like the Rainbow Darter (Etheostoma caeruleum), which are commonly found in small rivers with heterogeneous substrate (Harding et al., 1998). Species with a stable-widespread distribution (11%) included ubiquitous species, such as Common Shiner (Luxilus cornutus), Longnose Dace (Rhinichthys cataractae), and White Sucker (Catostomus commersonii), which are abundant with large geographic ranges (Thompson, 2001). Species assessed as spreaders (13%) consisted of stocked species, such as Rainbow Trout, exhibiting range expansion to over 50% of sub-watersheds. Species characterized as having a decliner-limited distribution (3%) included rare and difficult to detect species. For example, the Endangered Redside Dace was found in less than 30% of sub-watersheds in period one to less than 10% in period five. This coincides with the severe decline in conservation status, from Special Concern (1987) to Threatened (2000) to Endangered (2007) due to increases in urbanization (Poos et al., 2007). Other decliners included habitat specialists, such as the Blacknose Shiner (Notropis heterolepis), which requires clear, heavily vegetated habitat (Scott and Crossman, 1973). Using a linear slope to determine temporal trends in species distribution may not be the most appropriate model for some species. For example, the Brassy Minnow (Hybognathus hankinsoni) displayed a Poisson distribution while the Rainbow Trout demonstrated a logistic distribution. Testing multiple fits for each species could improve the accuracy of species temporal-trend characterization over time.
Fish communities are becoming generally more similar across sub-watershed and within sites and sub-watersheds in the Credit River over time. Homogenization generally increased over time at the site, but not sub-watershed, level. Fish community similarity increased at both site and sub-watershed levels in the inter-decadal pairwise analyses, and at the site, but not sub-watershed, level in the historical analysis. The increase in fish community similarity and homogenization over time could be expected based on the environment of the watershed. The Credit River watershed is of high conservation concern due to increasing urbanization, a changing climate, and the introduction of AIS (Chu et al., 2015). Marchetti et al. (2006) found that urbanization was positively correlated to the invasion of non-native fishes and the endangerment of native fishes in California. Blair (2001) determined that increased urbanization favors the persistence of fewer native bird and butterfly species, loss of rare native species, and introduction of AIS. Therefore, these studies concluded that increased urbanization can lead to, or establish, favourable environments for faunal homogenization. In the Credit River watershed, rates of urbanization have been increasing, with population and road-density data displaying linear increases over time (Allen and Mandrak, unpublished data).
Climate change has also been linked to homogenization by influencing the distribution of predator species. The climate of the Credit River has warmed across the periods (Allen and Mandrak, unpublished data). Therefore, this watershed represents an area where species distributions may already be altered by climate, resulting in the expansion and subsequent homogenizing effects of certain predatory species, such as Largemouth Bass (Micropterus salmoides), a species spreading in the Credit River watershed. Jackson and Mandrak (2002) predicted that the northward expansion of Smallmouth Bass could result in the loss of cyprinid assemblages and lead to homogenization in the fish communities. Furthermore, it was determined that lakes inhabited by Smallmouth Bass, in comparison to non-bass lakes, had an average of 2.3 fewer small-bodied species (MacRae and Jackson, 2006). This reduction in species was attributed to predation from the bass, leading to the potential homogenization of the lake.
Since the 1970s, the combination of introduced AIS, such as Common Carp and Round Goby, and the consistent stocking of sportfishes, like Brown Trout (Salmo trutta) and Chinook Salmon, could be a primary cause of the increased faunal homogenization. Of 233 known invasive fish species, 18 account for 49% of global introductions (Rahel, 2007). These species include the Common Carp and Goldfish, both found in the Credit River. Invasive fish species in the Credit River have increased from none to seven in some sub-watersheds between periods one and six and, therefore, appear to be a primary cause of homogenization. Alternatively, Eby et al. (2006) determined that stocking over time can lead to increased homogenization in fish communities. This is caused by increased richness in stocked species, which tend to be predators, or introduced stocked species altering native populations. The introduction of sportfishes, rather than the extirpation of native species, has also played a large role in altering fish communities (Rahel, 2010). Both the introduction of invasive species and stocking of sportfishes may play a role in increased homogenization in the fish communities of the Credit River watershed.
The faunal similarity analyses, based on the methods of Rahel (2000) and Taylor (2004), indicated a small increase in faunal homogenization at the site level, but no change at the sub-watershed level. Based on freshwater fish faunas, Taylor (2004) found minor homogenization across the Canadian provinces and territories, with a 1.2% increase in faunal similarity, while Rahel (2000) found greater homogenization with a 7.2% increase across the American states. This is similar to the results of this study, finding small increases in homogenization across periods at the site level in the Credit River. The lack of homogenization at the sub-watershed level could stem from the multiple introductions of AIS and stocked species in some, but not all, of the sub-watersheds. Despite finding homogenization in the Canadian freshwater fish fauna, Taylor (2004) also found that aquatic regions in British Columbia became significantly less similar due to localized increases in non-native species. Alternatively, Marchetti et al. (2001) determined that homogenization can be dependent on spatial scale, with watersheds showing little to no levels of homogenization compared to larger scales (e.g. provinces, states). Therefore, the lack of homogenization found at the sub-watershed level in this study may be the result of the localized introductions of invasive species, such as Round Goby, and the spatial scale of this study. The fish communities have become generally more similar between adjacent decades. The slow, but consistent, increase in homogenization suggests that the effects of invasive fish species, stocking, urbanization, and climate could be driving the Credit River watershed to high levels of homogenous fauna in the near future. Taylor (2010) determined that, over a five-year time scale, fish communities consistently exhibited faunal homogenization at the smallest (provincial) and largest scale (Canada).
The historical analyses displayed a small increase in homogenization at the site level and decreased homogenization at the sub-watershed level. The decrease in homogenization at the sub-watershed level can be potentially explained through changes in sampling methods and the introduction of invasive fishes since the historical time period. Comparing seining data from time period one to electrofishing data (period 3-6) could result in lower species similarities as a result in differences in capture efficiency. This stems from electrofishing producing higher estimates of species richness and diversity compared to seines (Mercado-Silva and Escandon-Sandoval, 2008). Seines are also less efficient at capturing fast-moving and benthic fish species (Mercado-Silva and Escandon-Sandoval, 2008). Furthermore, during period one sampling, the crew ended sampling upon the capture of an indicator species (H. Regier, University of Toronto, pers comm., 2015). Therefore, the decline in species similarity in the historical sub-watershed-level analyses could be attributed to both introductions and changes in sampling method.
Species richness and distribution demonstrated predictable trends in the Credit River watershed from 1954 to 2015. The homogenization results indicate that Credit River fish communities are generally becoming more similar over time. Future studies should assess what drivers are leading to increasing rates of homogenization in the Credit River and how fish communities will respond.
We would like to thank: Henry Regier for his insight into the 1954 sampling methods; Credit Valley Conservation, and Jon Clayton, Phil Bird, and Deborah Martin-Downs in particular, for providing historical fish datasets used in this report and for their continued guidance on this project; and the Mandrak Lab at the University of Toronto Scarborough for their help throughout this project. The NSERC CREATE Great Lakes program provided funding.
Supplementary data for this article can be accessed on the publisher’s website.
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/uaem.