How Surface Water Management Can Benefit Fish Conservation in Urban Streams

Carefully managing municipal wastewater effluent (hereafter effluent) is especially important in urban environments, where human activity can lead to alterations in water temperature, stream hydrology, and chemical composition that contribute to a decline in the biodiversity of fish and other organisms (Paul and Meyer 2001; Roy et al. 2005). In particular, the chlorination of effluent results in toxicity to many freshwater fish species (Tsai 1968; Roseboom and Richey 1977; Paller et al. 1983; Szal et al. 1991). However, long-term data have shown that management and abatement efforts can successfully limit impact to fish communities: mitigation of effluent over the past decades has led to significant improvements in fish abundance and biodiversity in waters receiving treated effluent (Parker et al. 2015; Gibson-Reinemer et al. 2017). Common examples of effluent mitigation include, among other approaches, cessation of chlorination (or dechlorination), increasing secondary and tertiary treatment, and abatement of combined sewer overflows (CSOs; Holeton et al. 2011). Some of these improvements (such as CSO abatement) have tended to be studied as part of aggregate improvements to systems receiving urban runoff, if they are related to fish communities at all (Plum and Schulte-Wülwer-Leidig 2014; Parker et al. 2015; Gibson-Reinemer et al. 2017).

Low-order streams have a special vulnerability to decreases in biodiversity because of their outsized importance to overall fish biodiversity and their sensitivity to pollution and urbanization (Karr et al. 1985; Gomi et al. 2002; Meyer et al. 2007). Because of the very tight link between freshwater stream biota, including fish, and watershed land use and land change patterns, these biota also respond to management of the landscape (Gomi et al. 2002; Meyer et al. 2007; Mueller et al. 2018). The responsiveness of fish diversity to a wide array of threats also makes them valuable for demonstrating the benefits of land use planning as well as pollution mitigation (Karr 1987; Vander Vorste et al. 2017). For example, a long-term (1957–2013) set of data from throughout the Illinois River Waterway showed dramatic increases in fish community biodiversity and abundance, which the authors attributed to systemic improvements following the implementation of the Clean Water Act of 1972 (Parker et al. 2015; Gibson-Reinemer et al. 2017). At the same time, there has been an impact on streams from altering the landscape via urbanization. The generally accepted understanding of a stream's response to urbanization is that as urbanization increases, biodiversity generally decreases because disruption-tolerant fish species replace more intolerant species (Paul and Meyer 2001; Roy et al. 2005; Steffy and Kilham 2006; Wenger et al. 2008, 2009; Brown et al. 2009). The mechanism implicated in this shift is the increase of stormwater runoff from impervious surfaces such as roofs, streets, parking lots, and other paved spaces. However, researchers developed this conceptual model from instances where humans converted forested or native vegetation to impervious land cover (Wenger et al. 2009). It is not as clear from previous research whether the conversion of agricultural land to impervious surfaces would have the same effects.

Our study analyzes changes in fish communities over time in a low-order, highly urbanized, low-gradient stream in an agricultural region: central Illinois, USA. Using a long-term set of data for this stream, we sought to analyze the impact on fish communities of 1) cessation of effluent chlorination, 2) CSO abatement, and 3) increased impervious surfaces within the watershed due to urbanization. Based on past research, we predicted that water management changes (chlorination cessation and CSO abatement) would most likely increase fish abundance and diversity, whereas increasing percent impervious surfaces would be predicted to primarily decrease the presence of sensitive species.

Sugar Creek is a small tributary of the Illinois River System, with its headwaters and upper watershed almost entirely within the twin cities of Bloomington-Normal, IL (Figure 1). The watershed for the region of Sugar Creek that we studied lies largely within the urban area of Bloomington, with study sites at stream sections ranging from second (sites 1 and 2) to third (sites 5 and 6) to fourth order (all remaining sites). Downstream of the city limits, the stream size is uniformly fourth order, and the surrounding landscape is heavily agricultural, until it converges with the Sangamon River (Bloomington-Normal Water Reclamation District [BNWRD] 2017). Although the twin cities have separate drinking water supplies, their wastewater is jointly managed by the BNWRD. The town of Normal performs maintenance of the riparian corridor for Sugar Creek within its boundaries, whereas BNWRD manages it within the boundaries of Bloomington. The water reclamation district, along with other wastewater facilities in Illinois, is required by the Illinois Environmental Protection Agency to conduct “Facility Related Stream Surveys” of abiotic (water quality) and biotic (fish community) parameters upstream and downstream of wastewater treatment plants. The fish and water quality surveys that we analyzed occurred from 1985 to 2017. Two major changes in wastewater management took place during this 33-y time span that attempted to mitigate the negative effects of effluent: the cessation of chlorination beginning at the end of 1990 and the completion of a CSO abatement project in 2007–2008.

Over the 33 y examined, testing for water quality parameters and fish community composition consistently occurred at 13 sampling sites (Figure 1) within the Sugar Creek watershed. Sites 1, 2, 4, 5, 6, 9, 14, and 15 are in areas with a high percentage of impervious surface coverage, with site 14 just upstream and site 15 just downstream of the effluent outflow. Sites 16, 16A, and 17–19 are considered by BNWRD to be in areas of stream recovery from the impacts of effluent (BNWRD 2017). Hereafter, we consider sites with numbers 14 and below as “upstream” and sites with numbers 15 and above as “downstream.” Significant stretches of Sugar Creek's streambed have been armored with either solid concrete or concrete lattice. Because armoring creates low-quality habitat for fish, the location of all of the sampling sites was in nonarmored sections of the stream, with sites 1, 2, and 4–6 located upstream of considerable lengths of armoring (BNWRD 2017). Because of a reliance on surface runoff, upstream sites experienced low or no flow during summer in several sampling years (BNWRD 2017).

BNWRD conducted annual surveys monthly from May to October every year, beginning in 1985 (Data S1, Supplemental Material). Water quality measurements, including water and air temperature, dissolved oxygen (DO), biological oxygen demand (BOD), pH, total ammonia, unionized ammonia, conductivity, turbidity, and discharge were taken according to established protocol (BNWRD 2017). For the fish surveys (performed by a team of two to three members), the team performed backpack electroshocking for 30 min at each site, for each sampling time, with all available fish habitats “sampled as equitably as possible” (BNWRD 2017). Following electroshocking, the team performed netting at riffle areas to target “small minnow and shiner species that are less susceptible to electroshocking than are larger species” (BNWRD 2017). In each case, the team used a minnow seine to perform a 15.24-m sweep. From 1985 to 1987, the seine was 6.10 m wide; beginning in 1988, the team adopted a 4.88-m-wide seine. From 1985 to 1988, the team carried all fish to the lab; however, beginning in 1989, only fish that could not be identified immediately were sent to the lab for identification. The intention of this change was to reduce the sampling impact on the fish population. Identification of all fish was via Smith (1979). Two sites (17 and 18) showed a significant relationship (F-test: F = 4.39–7.527; P = 0.01–0.044; n = 33) between species richness and discharge, depending on the month. This effect was primarily the result of high-flow events occurring in June that may have decreased capture probabilities for some species. To limit the impact of seasonal discharge on results, we aggregated all data by year.

We used the data from reports filed for 1985–2017 and calculated the annual species richness, annual mean catch rate (all individuals person⁻¹·h⁻¹), and number of stream-specific, pollution-intolerant species common to Illinois and the Midwest: Rock Bass Ambloplites rupestris; Highfin Carpsucker Carpiodes velifer; Steelcolor Shiner Cyprinella whipplei; Fantail Darter Etheostoma flabellare; Orangethroat Darter Etheosoma spectabile; Banded Darter Etheosoma zonale; Northern Hogsucker Hypentelium nigricans; Smallmouth Bass Micropterus dolomieu; Shorthead Redhorse Moxostoma macrolepidotum; Stonecat Noturus flavus; and Slenderhead Darter Percina phoxocephala. Multiple sources for the region attest to most of these fish as being intolerant or moderately intolerant of pollution such as turbidity and sedimentation (Miltner et al. [2004], and sources therein; Grabarkiewicz and Davis [2008], and sources therein). Furthermore, Karr et al. (1986) defined Steelcolor Shiner and Highfin Carpsucker specifically as central Illinois species intolerant of pollution. The BNWRD reports also included a calculation for the annual value of an alternative index of biotic integrity (AIBI) following the protocol described by Karr (1981), with some modifications designed to better suit the nature of the collections made at Sugar Creek (BNWRD 2017). Modifications made to calculate the AIBI were based on the fish assemblage found in Sugar Creek. Specifically, the modifications serve the following functions: 1) making the calculation annually rather than for individual sampling times, due to season-dependent periods of low flow at some sites; 2) defining the metric for number of individuals as an average of seining and electrofishing results; and 3) defining the scores for the number of sunfish, sucker, darter, and pollution-intolerant species metrics based on the number of species specific to Sugar Creek (BNWRD 2017). The modifications act in a manner consistent with the approach described in Karr et al. (1986) for adapting the IBI to specific streams. To estimate fish biodiversity, we calculated the Shannon–Wiener diversity index (H′; Gibson-Reinemer et al. 2017).

We determined the geographic coordinates of each site from Google Earth (Google, Mountain View, CA) by using the written site descriptions found in the annual reports. We obtained a digital elevation model of Illinois from the Illinois State Geological Survey (Luman et al. 2003) and used the elevation data to generate subcatchments for each site with the Watershed tool in ArcMap v.10.5 (ESRI, Redlands, CA) following a standard method (Zhu 2016). We downloaded raster files of impervious surface percentages for 2001, 2006, 2011, and 2016 from the U.S. Geological Survey Multi-Resolution Land Characteristics Consortium (USGS 2019). We used ArcMap to calculate a weighted average of total percentage of impervious surfaces within each subcatchment area for each year that was available.

To determine long-term trends in the fish community metrics of interest (i.e., mean annual catch rate, annual species richness, annual number of intolerant species, annual H′, and AIBI), we performed linear regressions as a function of time (1985–2017). Catch rate and H′ were based on annual averages, whereas species richness and number of intolerant species were annual totals. AIBI was an annual score, as described previously. In addition to calculating trends for each individual site, we also averaged each metric for upstream and downstream sites for each year. This initial analysis provided a background for understanding overall, long-term changes in the fish community metrics. To determine whether there were any short-term effects of the effluent mitigation efforts, we compared fish community metrics (as described above) for the 6-y periods before and after each mitigation effort (cessation of chlorination and completion of CSO abatement) for significant differences. This time frame was chosen because there were only 6 y of data before cessation of chlorination. Water quality parameters were calculated as medians to account for below- and above-detection limit values during certain seasons. To make sure the data met assumptions of normality and homoscedasticity, we conducted a Shapiro–Wilk test and variance ratio test, respectively. We tested the assumption of homoscedasticity by using the rule of thumb that var_max/var_min > 1.5 indicated heteroscedasticity, where var_max is the mean with the highest variance and var_min is the mean with the lowest variance. In the case that one or both means did not meet the assumption of normality, or that the metric itself was discrete rather than continuous (as in the case of AIBI), a Wilcoxon rank-sum test was substituted for the t-test. These comparisons were also performed for the averages of upstream and downstream sites (described previously).

Because BNWRD collected these data against a background of overall improvements in the river system due to implementation of the Clean Water Act of 1972 (Gibson-Reinemer et al. 2017), we preformed additional analyses to determine whether our observed short-term effects could be more conclusively attributed to changes in effluent management. To this end, we separated the data into five consecutive 6-y time periods (1985–1990, 1991–1996, 1997–2002, 2003–2008, and 2009–2014) that coincided with the short-term periods described above. Because most of the data did not meet the assumptions to perform an analysis of variance, we preformed the nonparametric Kruskal–Wallis test in a pairwise manner to compare all time periods to one another and look for significant differences (α = 0.05). We performed this test for each site as well as the upstream and downstream sites grouped together. In addition, we compared sites grouped as upstream and downstream on a year-by-year basis, with each year compared with the preceding and following 3 y. This approach allowed us to identify any significant shifts in fish community metrics and determine whether they coincided with changes in effluent management. These comparisons were done only for upstream and downstream sites pooled for replication, and because the Kruskal–Wallis test has a relatively low power, significance was set at α = 0.10.

To model the impacts of urbanization, we calculated the amount of change in the annual mean values of DO, pH, unionized ammonia, conductivity, turbidity, discharge, percent impervious surfaces, and fish community metrics by using linear regression for the time period 2001–2016. We assumed the slope of each linear regression to represent the average rate of change for this time period and used it to calculate the absolute change for the 16-y period. We compared all variables (as estimated change) to one another using a Pearson's correlation test as a first pass. If Pearson's test indicated a significant (α = 0.10) correlation between a variable and change in impervious surfaces, we used linear regression as a second pass to determine the strength, directionality, and significance of the relationship, with change in impervious surfaces as the independent variable. We used residual analysis to identify potential curvature and performed nonlinear regression when appropriate, following the approach taken by Syed et al. (2016). We tested linear, logarithmic, exponential, second-order polynomial, and sigmoidal (Olmstead and LeBlanc 2005) curves for nonlinear relationships with change in impervious surface percentage.

Very few water quality parameters showed a significant difference when comparisons were made between periods before and after the two effluent mitigation efforts (Tables S1 and S2, Supplemental Material). Dissolved oxygen increased at three upstream sites following CSO abatement, which led to a significant increase in the upstream average. These sites were well upstream of the effluent outflow. Two sites downstream of the outflow also showed significant increases in pH following completion of CSO abatement. This led to a significant increase in pH for the downstream average. It is unclear how CSO abatement could have led to either of these results, and it is likely the result of some other, unknown factor.

Most fish community metrics significantly increased over the 33-y time span (Figure 2; Table 1). Species richness and AIBI increased significantly at all but two sites (sites 2 and 6), both upstream. Change occurred more rapidly at downstream sites than at upstream sites, according to slope (Table 1), and most relationships were relatively strong, although r² values ranged from 0.124 to 0.752 for individual sites. The number of intolerant species followed a similar pattern, except a third upstream site (site 1) also did not show a significant relationship to time. Relationships between time and number of intolerant species were relatively strong, with r² values ranging from 0.416 to 0.818 for individual sites. It is noteworthy that, in every case, the site directly downstream of the effluent outflow (site 15) had the strongest relationship and one of the largest slopes.

Annual diversity (H′) showed a different pattern over time (Figure 2; Table 1). Upstream sites either showed no trend (sites 1, 2, 4, and 14), a significantly positive trend (sites 5 and 9), or a significantly negative trend (site 6) with time, whereas downstream sites were all positively correlated with time, except for one site (site 19). The strength of the significant relationships was variable, with r² values ranging from 0.143 to 0.659. Annual mean catch rate of fish only showed significant positive trends at one downstream site (site 15) and two upstream sites (sites 1 and 6), whereas a third upstream site (site 4) had a significant negative trend during this time. However, these trends were all relatively weak, with r² values ranging from 0.170 to 0.474. The high variability of catch rate across space and time indicates that factors other than time likely explain more of this variability. Therefore, improvements to the Sugar Creek system do not appear to have improved fish abundance as measured by catch rate.

Annual species richness and diversity significantly increased at both upstream and downstream sites in the 6 y following the cessation of chlorination (Table 2). However, the two sites directly upstream of the effluent outflow (sites 9 and 14) and the site directly downstream of it (site 15) contributed the most to this outcome, although the upstream site 5 also significantly increased in richness at this time (Table S3, Supplemental Material). Upstream sites demonstrated a clear, rapid increase in both diversity and species richness in 1990 that led to a new, relatively constant value, whereas downstream sites showed a longer term increase that only began in 1990 (Figure 3). This supports the idea that cessation of chlorination, rather than system-wide improvements, led to increases in species richness and diversity at upstream sites and to a lesser extent at downstream sites. Mean catch rate, AIBI, and number of intolerant species did not show any consistent, significant changes when comparing the 6 y before with the 6 y after cessation of chlorination (Table 2). Although the number of intolerant species did significantly increase at upstream sites as a whole, this appears to be part of a long-term trend, rather than an immediate effect of cessation of chlorination. Again, the site directly downstream of the effluent outflow, site 15, had a significant increase in both AIBI and catch rate after the cessation of chlorination.

Species richness, number of intolerant species, and diversity all significantly increased at downstream sites in the 6 y after completion of the CSO abatement project (Table 3), consistent with long-term trends (Table 1). The number of intolerant species experienced a large increase in 2012 (Figure 3) that, in turn, reflected more frequent collections of Orange Throat Darter, Banded Darter, and Steelcolor Shiner that began in 2010, 2011, and 2012, respectively. However, this followed some smaller expansions of intolerant fish in the years preceding CSO abatement, including the Northern Hognose Sucker (2001), the Banded Darter (2003), and Smallmouth Bass (2005–2007), as seen in Figure 3. By contrast, no metrics significantly increased at upstream sites during this time period, except for some gains in the number of intolerant species that followed long-term trends. In fact, diversity significantly declined at upstream sites on average during this time period (Table 3), whereas the long-term pattern showed variable change at those sites (Figure 3). Mean catch rate significantly decreased at downstream sites on average in the years following the CSO abatement, primarily because of two sites (sites16 and 18).

Analysis of data from 2001 to 2016 found that our sites had between 14.34 and 42.97% impervious surfaces within their subcatchments. The magnitude of change over this time period also varied within subcatchments, with increases ranging from 1.3 (site 19) to 11.1% (site 1). Although variable, upstream sites had higher absolute percentages of impervious surfaces and significantly higher (two-tail t-test: t = 3.21; P = 0.012; n₁ = 7; n₂ = 6) increases in percent impervious surfaces over this period. The results of linear regressions over time indicated that all sites experienced a significant increase in impervious surface at the level of α = 0.10. However, change was strongly significant (α = 0.05) only at sites further upstream (sites 1, 2, 4, 5, and 9). The percentage of impervious surfaces increased primarily in upstream regions of Sugar Creek, suggesting that effects from impervious surfaces cannot be completely separated from effects caused by stream size.

The change in percent impervious surfaces significantly (α = 0.10) correlated to concurrent changes in median air temperature, water temperature, and BOD throughout Sugar Creek. These relationships were all positive, but the relationships with air temperature (P = 0.045, r² = 0.317) and water temperature (P = 0.061, r² = 0.285) were weak, and individual sites did not show significant changes over this time (Tables S4 and S5, Supplemental Material), suggesting the correlations are illusory. The relationship between percent impervious surfaces and BOD was relatively strong (P = 0.008; r² = 0.487), and BOD in Sugar Creek significantly (α = 0.10) decreased over this time at 10 of the 13 sites. Only upstream sites 1, 2 and 4, the sites with the highest percentages of impervious surfaces, did not show this relationship. This means that BOD decreased throughout Sugar Creek, but the sites most impacted by expanding impervious surfaces showed no change. The total percent impervious surfaces of subcatchments did not correlate to any trends in water quality metrics during this time.

Of all fish community metrics, only the change in number of pollution-intolerant species was significantly (F-test: F = 6.3; P = 0.014; n = 13) correlated to change in the percentage of impervious surfaces, and no metrics changed in a manner that correlated with total percent impervious surfaces. This relationship was negative and nonlinear, with a three-parameter sigmoidal curve with a lower bound of zero as described by Syed et al. (2016) fitting the data significantly better (F-test: F = 24.02–31.42; P < 0.001; n = 13) than all other nonlinear regressions that were modeled (linear, logarithmic, exponential, and second-order polynomial). According to this relationship, subcatchments with the lowest increase in the percentage of impervious surfaces averaged an increase of 3.42 intolerant species from 2000 to 2016 (Figure 4), but sites with the highest increases in the percentage of impervious surface demonstrated no change in the number of intolerant species found. This result corresponds with increases in intolerant species at downstream sites with little percent impervious surface increase, as described previously.

Our work demonstrates a significant improvement in fish community richness and diversity in this urban stream since 1985, a change that coincides with implementation of the Clean Water Act of 1972, and supports research in other reaches of the Illinois River System (Parker et al. 2015; Gibson-Reinemer et al. 2017). Previous work has shown that improvements in wastewater management have increased fish diversity (H′) and sport fish catch rate in the Illinois River and some of its larger tributaries (Gibson-Reinemer et al. 2017). At some locations on the Illinois River, species richness and native species metrics also increased (Parker et al. 2015). Within our small tributary of the Illinois River, we likewise found long-term increases to those same or comparable metrics. However, we found that species richness and number of intolerant species showed the most consistent increases across sites, followed by diversity, whereas AIBI showed less consistent increases. The overall impact of improved water management in Sugar Creek has done little to increase the catch rate of fish in the stream, but it has increased the number and type of species found.

The literature supports the idea that adding chlorine to effluent increases its toxicity, leading to direct reduction of fish populations in the field (Tsai 1968) by causing decreased spawning (Zillich 1972), avoidance of chlorinated waters (Osborne et al. 1981), reduced growth and increased illness (Grizzle et al. 1988), and outright mortality (Zillich 1972; Osborne et al. 1981; Grizzle et al. 1988; Szal et al. 1991). Studies assessing the resilience of freshwater fish communities following cessation of chlorination or active dechlorination of effluent are few. However, Paller et al. (1983) found that when chlorination of effluent ceased in two tributaries of the Kaskaskia River in Illinois, species richness and abundance both increased immediately downstream of effluent outflow. In addition, chlorine concentrations in those rivers showed a negative correlation to IBI score (Karr et al. 1985), with Paller et al. (1983) specifically noting an increase in less tolerant species at one site below the effluent outflow. Kleinssasser and Linam (1992) and Wise (1995) likewise saw increases in species richness downstream of effluent outflows following wastewater dechlorination on the Trinity River and Pecan Creek, respectively, both in Texas, USA. Wise (1995), in particular, noticed a shift toward multitaxa dominance and increased diversity in stretches where wastewater effluents were dechlorinated.

One difference between our results and those of some of the published work in this field is that we found increased species richness and diversity directly upstream of the effluent outflow. This is in contrast to Paller et al. (1983), who found no such improvements at upstream sites, whereas Kleinsasser and Linam (1992) did not examine changes at upstream sites at all following dechlorination. Although it may seem counterintuitive that chlorinated effluent could affect upstream sites, fish are known to avoid chlorinated waters (Zillich 1972), although they will sometimes make short excursions into chlorinated stretches (Osborne et al. 1981). Therefore, an area of continuously chlorinated water in a small stream can act as a barrier for upstream movement of fish (Osborne et al. 1981). Preventing this kind of movement would ultimately lower diversity in a low-order stream by excluding fish species that use headwaters as nurseries (Meyer et al. 2007). Therefore, it is likely that Paller et al. (1983) saw no impacts because, in the Kaskaskia River, fish were able to find a path upstream around the chlorinated effluent in a manner similar to that observed in other studies (Zillich 1972; Osborne et al. 1981; Kleinsasser and Linam 1992). This idea is supported by Wise (1995), who also found a significant increase in species richness at one site upstream of effluent outflow following dechlorination. Pecan Creek, the urban stream studied by Wise (1995), has more similarities in its size and urban character to our sites on Sugar Creek than to the other studies cited. Therefore, it is likely that dechlorination or cessation of the chlorination of effluent will improve fish communities both upstream and downstream of effluent outflows in lower order, urban streams.

To the best of our knowledge, this is the first study that has examined the response of fish community metrics in a small urban stream following a specific CSO abatement project. The impacts of CSOs on fish populations are poorly understood, in large part because of the challenge in assessing the amount of wastewater released during CSO events (Viviano et al. 2017). These events occur on both a highly stochastic time scale of hours to days and a highly localized geographic scale (Holeton et al. 2011). The assumed impacts of CSOs include increased loads of fecal coliforms, decreased DO, increased turbidity, and increased toxicant concentrations (see Farnham et al. [2017], and sources therein). In our study, although DO increased at multiple sites following the completion of the CSO abatement project. These were all sites upstream of effluent outflow, or far downstream. This makes the exact mechanism by which CSO abatement could have impacted Sugar Creek DO unclear. Parker et al. (2015) took a similar approach to our study in analyzing changes to the fish community of the upper Illinois River by using data recorded over a long period. They found that from 1983 to 2010, fish diversity in the reach closest to the highly urbanized city of Chicago, IL, negatively correlated to fecal coliform counts, suggesting that Chicago's decreased CSO volumes ultimately benefited that stretch of the river (Parker et al. 2015). Our data similarly indicate that completion of a CSO abatement project in Bloomington-Normal coincided with increases in fish species richness and diversity as part of long-term trends. Specifically, in the years immediately following CSO abatement, three intolerant fish species were found more consistently at sites downstream of the effluent outflow, the most dramatic expansion of intolerant fish species on record at those sites.

We found no correlation between total percent impervious surfaces or changes in that percentage with species richness, abundance, or diversity. However, we identified an inverse relationship with the number of intolerant species with respect to change in percent impervious surfaces. Previous studies in other streams have established that increasing amounts of urbanization and percent impervious surface typically result in a decrease in sensitive species, whereas the responses of fish abundance and species richness can be mixed (Walsh et al. 2005). Closer examinations have shown that this is actually a complex response, with native species being replaced by invasive, tolerant species, without necessarily decreasing overall abundance and species richness (Steffy and Kilham 2006). For example, a study of the Etowah River in Georgia, USA, found a significant negative relationship between percent impervious surface within a watershed and the number of endemic species, fluvial specialists, and intolerant species, but no relationship with species richness or abundance (Roy et al. 2005). Similarly, in a comparison of nine different urban areas, Brown et al. (2009) found that although impervious surfaces (indexed as urban intensity) correlated to fish species traits in four of the cities under study, richness did not correlate in any. Our results reinforce the notion that the impacts of urbanization on a fish community can at present best be seen by looking at the traits of fish species, rather than their overall abundance and diversity.

We found that greater change in the percentage of impervious surfaces in a subcatchment resulted in a stark threshold effect on the number of intolerant species found at each site. Miltner et al. (2004) supported prior work that suggested that 8–20% impervious surfaces within a watershed causes “gross” impairment, whereas 25–65% impervious surfaces leads to “irreparable” impairment. However, a more recent review has suggested that such universal thresholds are problematic (Wenger et al. 2009). This is because the preservation of within-stream fish habitat and riparian vegetation can lead to a higher degradation threshold (Yoder et al. 1999), and, as we found, efforts to improve water quality can also improve fish abundance and species richness within the entire system (Gibson-Reinemer et al. 2017).

Our data indicated a system-wide improvement in the fish community and, specifically, the number of intolerant species found at each site. The observed effect therefore was not one of increasing impervious surfaces correlating with a decline in sensitive species, but rather increasing impervious surfaces preventing or inhibiting recovery at certain locations. Similarly, four intolerant fish species—Tricolor Shiner Cyprinella trichroistia, Etowah Darter Etheostoma etowahae, Speckled Madtom Noturus leptacanthus, and Bronze Darter Percina palmaris—from the Etowah River basin have also been shown to approach 0% occupancy after crossing impervious surface thresholds of 2, 4, ∼15, and 4%, respectively (Wenger et al. 2008). Our model predicts a similar sharp drop, with a 4.2% increase in impervious surfaces reducing the expected increase in the number of intolerant species by half. Although a universal threshold of percent impervious surfaces that controls recovery of intolerant species likely does not exist (Wenger et al. 2009), it is also clear that regional and local thresholds do. The key to using the threshold concept to design and evaluate planning and conservation outcomes lies in customizing the threshold to the local situation, in much the same way that regionalizing other bioassessment protocols (such as IBIs) increases their relevance.

In summary, we used 33 y of standardized long-term monitoring data to show that fish diversity, species richness, number of pollution-intolerant species, and AIBI improved dramatically as water quality improved in the Sugar Creek system. These increases were strongest at sites downstream of the effluent outflow. Analysis of years directly before and after cessation of wastewater chlorination and completion of a CSO abatement project supports the conclusion that these effluent mitigation efforts benefitted fish communities. Cessation of chlorination corresponded to a 14% increase in Shannon diversity and a 20% increase in species richness across all sites, whereas CSO abatement corresponded with increases in Shannon diversity (11%), species richness (10%), and number of intolerant species (15%) at sites downstream of the outflow only. Increases in impervious surfaces likely prevented the recovery of intolerant species, primarily at upstream sites. Taken together, this strongly suggests that by implementing these types of changes, municipal wastewater managers can strongly benefit environmental quality and species conservation efforts. However, in an urban stream system such as Sugar Creek, where human impact may come from multiple, unrelated stressors, these benefits can appear muted or masked. Thus, we suggest that more research into the interacting effects that wastewater management and urbanization have on stream communities would improve the success of urban planning and wastewater management efforts.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author for the article.

Table S1. Comparisons between select water quality measurements at 13 sites in Sugar Creek (Bloomington-Normal, IL, USA) before (1985–1990) and after (1991–1996) cessation of chlorination. All measurements are recorded as mean (standard deviation). ≈ indicates no significant difference between the two time periods; > indicates the 1985–1990 mean is significantly larger than the 1991–1996 mean; < indicates the 1991–1996 mean is significantly larger than the 1985–1990 mean; DO = dissolved oxygen; Up = average for sites upstream of effluent outflow; Down = average for sites downstream of effluent.

Available: https://doi.org/10.3996/JFWM-20-051.S1 (21.51 MB DOCX)

Table S2. Comparisons between select water quality measurements at 13 sites in Sugar Creek (Bloomington-Normal, IL, USA) before (2003–2008) and after (2009–2014) completion of a CSO abatement project. All measurements are recorded as mean (standard deviation). ≈ indicates no significant difference between the two time periods; > indicates the 2003–2008 mean is significantly larger than the 2009–2014 mean; < indicates the 2009–2014 mean is significantly larger than the 2003-2008 mean. DO = dissolved oxygen; Up = average for sites upstream of effluent outflow; Down = average for sites downstream of effluent.

Available: https://doi.org/10.3996/JFWM-20-051.S2 (21.3 MB DOCX)

Table S3. Comparisons between fish community measurements (annual species richness, annual alternative index of biotic integrity [AIBI], and mean catch rate] at 13 sites in Sugar Creek (Bloomington-Normal, IL, USA) during five 6-y time periods. All measurements are recorded as mean (standard deviation). Means with the same letter are not significantly different from other means recorded at the same site according to a nonparametric Kruskal–Wallis test (α = 0.05). Up = average for sites upstream of effluent outflow; Down = average for sites downstream of effluent.

Available: https://doi.org/10.3996/JFWM-20-051.S3 (19.6 MB DOCX)

Table S4. Comparisons between fish community measurements (annual number of intolerant species, annual Shannon–Weiner diversity index [H′]) at 13 sites in Sugar Creek (Bloomington-Normal, IL, USA) during five 6-y time periods. All measurements are recorded as mean (standard deviation). Means with the same letter are not significantly different from other means recorded at the same site according to a nonparametric Kruskal–Wallis test (α = 0.05). Up = average for sites upstream of effluent outflow; Down = average for sites downstream of effluent.

Available: https://doi.org/10.3996/JFWM-20-051.S4 (18.33 MB DOCX)

Table S5. Outcomes of linear regressions with respect to time (2001–2016, based on monthly samples May–October each year) for water quality measurements at 13 sites in Sugar Creek (Bloomington-Normal, IL, USA). Regressions are reported as slope (P value). DO = dissolved oxygen; BOD = biological oxygen demand.

Available: https://doi.org/10.3996/JFWM-20-051.S5 (21.16 MB DOCX)

Reference S1. [BNWRD] Bloomington-Normal Water Reclamations District. 1985–2017. Biosurvey Report. Bloomington-Normal Water Reclamations District, Bloomington, IL.

Available: https://doi.org/10.3996/JFWM-20-051.S6 (85.98 MB ZIP)

We acknowledge A. Casper for providing extremely useful comments to an early draft of this paper. We also acknowledge the extensive data provided by BNWRD for use in this paper. Three anonymous reviewers and the Associate Editor provided comments that greatly improved an earlier version of this manuscript

Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.