Influence of a high‐head dam as a dispersal barrier to fish community structure of the Upper Mississippi River

In river systems, high‐head dams may increase the distance‐decay of fish community similarity by creating nearly impermeable dispersal barriers to certain species from upstream reaches. Substantial evidence suggests that migratory species are impacted by dams, and most previous studies in stream/river networks have focused on small streams and headwaters. Here, we assess whether a high‐head dam (Lock and Dam 19; LD 19) on a large river, the Upper Mississippi River (UMR), substantially alters fish community structure relative to variability expected to occur independent of the dam's effect as a fish dispersal barrier. Using fish catch per unit effort data, we modelled the distance‐decay function for the UMR fish community and then estimated the similarity that would be expected to occur across LD19 and compared it with measured similarity. Measured similarity in the fish community above and below LD19 was close to the expected value based on the distance‐decay function, suggesting LD19 does not create an abrupt transition in the fish community. Although some migratory fish species no longer occur above LD19 (e.g., skipjack herring, Alosa chrysochloris), these species do not occur in high abundance below the dam and so do not drive variation in fish community structure. Instead, much of the variation in species structure is driven by the loss/gain of species across the latitudinal gradient. Lock and Dam 19 does not appear to be a clear transition point in the river's fish community, although it may function as a meaningful barrier for particular species (e.g., invasive species) and warrant future attention from a management perspective.


| INTRODUCTION
Changes in species presence and abundance across spatial gradients cause a decay in the similarity of communities with geographic distance, known in the literature as the distance-decay function (Morlon et al., 2008;Soininen, McDonald, & Hillebrand, 2007). Put simply, similarity in community structure between any two sites decreases as the physical distance between those sites increases because of (a) changing environmental conditions, (b) dispersal barriers, and (c) ecological drift and/or organisms' limited dispersal abilities (Soininen et al., 2007). Dispersal barriers can limit the range of certain species and create abrupt transitions in community structure (Araújo et al., 2013;Martinez, Chart, Trammell, Wullschleger, & Bergersen, 1994), potentially creating clear boundaries for management action. Humans strongly influence dispersal barriers, both by circumventing barriers (Rothlisberger, Finnoff, Cooke, & Lodge, 2012) and by creating new ones (Mammoliti, 2002).
Dam construction is one of the most fundamental impacts humans can have on natural river systems (Dynesius & Nilsson, 1994 Gehrke, Gilligan, & Barwick, 2002;Fullerton et al., 2010;Wilcox, 1999), which decreases the similarity of communities above and below the dam. Even non-migratory fish that do not lose habitat directly may be affected by dam construction (Araújo et al., 2013), which often create impounded areas that were not previously present in the river network. These impounded areas foster non-native species that become established and expand to upstream and downstream reaches (Gao, Zeng, Wang, & Liu, 2010;Mammoliti, 2002;Martinez et al., 1994;Santucci, Gephard, & Pescitelli, 2005). Independent of their ability to function as a barrier to species dispersal, dams may eliminate flood plain habitats (Koel, 2004), alter flow regimes (Galat & Lipkin, 2000), increase open water surface area, and impound sediment, altering habitat availability and structure immediately above the dam (Bhowmik & Adams, 1989).
Few studies have explored the causes and drivers of the distancedecay function in large rivers. In Soininen et al.'s (2007) review of published distance-decay relationships, none were from the main stem of large river systems. Most previous studies in stream/river networks have focused on small streams and headwaters (e.g., Hitt & Angermeier, 2008;Datry, Pella, Leigh, Bonada, & Hugueny, 2016, but see Araújo et al., 2013;Vitorino Júnior, Fernandes, Agostinho, & Pelicice, 2016). The Upper Mississippi River (UMR) is a large river that extends 1,070 km from St. Anthony Falls, Minnesota, downstream to St. Louis, Missouri. Reaches between Minneapolis, Minnesota, and Clinton, Iowa, lie within the Driftless Area, an area unaffected by Pleistocene glaciation (Fremling, 2005). These reaches have the greatest habitat diversity due to steep forested slopes, bluffs, and rock cliffs preventing agricultural and urban development and thus encouraging natural flood plain habitat development (Theiling et al., 2000). In southern reaches, channelization and agricultural development have greatly simplified habitat structure reducing connectivity and increasing fragmentation (Theiling et al., 2000 weather.gov/climate; Theiling et al., 2000).
A total of 29 navigational locks with accompanying dams have been constructed in the UMR to create a waterway suitable for commercial and recreational boat traffic (U.S. Army Corps of Engineers, 2015). The lock and dam structures break the UMR into 28 navigational "pools," although the term "pool" here refers to the reach of river between lock and dam structures, not a habitat type (often there is very little impounded area associated with these lock and dam structures; U.S. Army Corps of Engineers, 2015). During high flow, 27 of these dams are "open" and do not pose any meaningful barrier to fish passage (Theiling & Nestler, 2010). However, open conditions at most dams occur less than 25% of the time (Theiling & Nestler, 2010), so the dams do appear to have reduced longitudinal connectivity to riverine biota (Fremling, 2005;Wilcox, Stefanik, & Kelner, 2004 (Coker, 1929;Theiling & Nestler, 2010;Wilcox et al., 2004). Concerns about the impact of LD19 on fish communities in the UMR have been longstanding (see review in Wilcox et al., 2004), and the presumption is that some migratory species such as the skipjack herring, no longer occur above LD19 because of the dam's presence. However, Chick, Pegg, and Koel (2006) used a distance-decay analysis to conclude that although upstream and downstream pools in the UMR had differing fish communities, fragmentation due to the navigation system as a whole did not appear to create abrupt changes in the fish community structure across the UMR.
In this study, we used electrofishing data from 11 UMR pools to characterize the distance-decay relationship in the UMR main-stem fish community. Although this relationship was previously observed by Chick et al. (2006), sampling in this past study was limited for Pools 19 and 20 and slightly skewed towards pools occurring above LD19 (eight of 14 pools). Chick et al. (2006) interpreted a strong distancedecay in similarity to be evidence itself of the weak impact of the navigation dams on fish community structure. Here, we also model this distance-decay relationship to predict similarities between Pools 19 and 20 (on either side of LD19). If LD19 has a big impact on community structure, then we would expect Pools 19 and 20 to be less similar than the distance-decay relationship would imply.

| Study area
Pool 19 is the longest pool in the UMR (74.5 km of total 1,070 km) and is geomorphically diverse. The lower half of the pool is characterized by great width (over 2.5 km below Fort Madison, Iowa), lacustrine habitat with low current velocities, static water levels, and extensive (6,800 ha) shallow-water areas with floatingleaved vegetation. The upper portion of the pool is riverine with extensive side channels and shallow backwaters. Pool 20 (35.2 km) is characterized as straight, narrow, and riverine with high current velocities, lotic conditions, sparse vegetation, and limited off-channel or lacustrine habitat. Pool 20 consists of mostly main channel (80.4%) habitat, followed by side channel (19%) habitat, and a small percentage of backwater (0.6%) habitat ( Figure 1). Pool 19 has proportionally less main channel (29.3%) and side channel (17.4%) habitat than Pool 20, in addition to more backwater (5.6%) and impounded habitats (47.7%; Figure 1).

| Data collection
Many sites in the UMR are regularly monitored by government agencies and universities, and descriptions of those sites can be found in Ratcliff, Gittinger, O'Hara, and Ickes (2013) and Fritts et al. (2017).  Figure 2), although these data are limited to main channel border (i.e., shallow water areas along channel edges) sites. Both LTRM and LTEF monitoring follow methods originally developed for the LTRM program (Gutreuter, Burkhardt, & Lubinski, 1995; see more detail below). Data from the LTRM and LTEF programs were obtained from the LTRM online database (http://www.umesc.usgs. gov/data_library/fisheries/fish_page.html) and from the Illinois Natural History Survey (INHS), Illinois River Biological Station, Havana, IL, USA, personnel, respectively.
To add to the existing network of data, we extensively sampled fish communities in Pools 19 and 20. Although LTEF has been sampling Pools 19 and 20 since 2009, their protocol limits sampling to one main channel site per five river miles. However, LTRM procedures select sites occurring within particular habitat types. Habitat types were identified based on aquatic areas designated by Wilcox (1993).
Forty-eight sites (24 main channel and 24 side channel) were selected using a stratified random sampling approach according to LTRM protocol (Gutreuter et al., 1995) to collect fish community samples, for both Pools 19 and 20 in 2013 and 2014. The main channel conveys the majority of the river discharge and in most reaches includes the navigation channel ( Figure 1). Side channels are large channels that carry less flow than the main channel ( Figure 1). Inaccessible areas, including those with private or no physical access, were omitted from consideration. If insufficient water depth or obstructions were present, the randomly selected site was replaced with the nearest randomly selected alternative site. In Pool 19 during 2013, one randomly selected site was sampled on two separate dates in the same time period, bringing the total for that year to 49 sites (Table 1). In some cases, it was not possible to sample (e.g., flooding, dangerous water conditions, inclement weather, and equipment malfunction) as was the case at some sites in Pool 20 during 2013, bringing the total for that year to only 43 sites (Table 1).
Fish collection methodology generally followed LTRM standardized electrofishing methods described in Gutreuter et al. (1995; https://www.umesc.usgs.gov/documents/reports/1995/95p00201. pdf). Each electrofishing run lasted 15 min and spanned a 200-m stretch of shoreline, which was consistent for both main channel and side channel sites. The electrofishing boat was operated by a pilot and two dip netters. Dip netters collected fish as they appeared, regardless of size or species. Fish were placed in a holding tank until the run was completed, and then enumerated, recorded, and released back into the river. Fish were collected three times each year (2013 and 2014; June 15 to July 31, August 1 to September 15, and September 16 to October 31) using new random sites each time interval.
Pulsed DC daytime boat electrofishing was used to sample fish using LTRM standardized electrofishing specifications. Power goals from the electrofishing boat and sampling design emulated LTRM protocols to achieve comparable fish catch rates and standardization across all pools in the UMR (Gutreuter et al., 1995). Western Illinois University Institutional Animal Care and Use Committee approval was obtained before commencement of this study (WIU 13-13-r). All fish were acquired and used in accordance with federal, state, and local laws and regulations.

| Statistical analyses
Like all collection methods, electrofishing is only effective for sampling a subset of the fish community, so we limited our analyses to species that are either known to be easily captured by electrofishing or are considered common species. Previously published research has shown that electrofishing has power > 0.80 to detect a 20% interannual abundance change in at least one habitat type for 16 species in the UMR (Lubinski, Burkhardt, Sauer, Soballe, & Yin, 2001; Table 2). We term these the "highly catchable" species. We restricted the analyses to either (a) those 16 "highly catchable" species or (b) each species that made up >1% of the catch in any pool, termed the ">1% catch" species (Table 2)  16  structure between pools was estimated using the Bray-Curtis Dissimilarity index as implemented in the R package "vegan" (Bray & Curtis, 1957;Dixon, 2003;R Development Core Team, 2014).
Dissimilarity is 1 − similarity, so for clarity, we converted everything to similarity for all data presentation and will refer to similarity hereafter. In the Bray-Curtis similarity index, higher values indicate more similarity (i.e., two communities are more alike). Bray-Curtis similarity between pools was calculated for each year. To minimize differences in fish community due to habitat (which influences electrofishing effectiveness), statistical procedures were performed separately for main channel and side channel habitats. In addition, statistical procedures were not performed separately among sampling time periods.
All samples within a pool were standardized for a 15-minsampling run and averaged to generate a pool-specific estimate of CPUE.
All data were square root transformed and standardized (using the "wisconsin" standardization) prior to analysis (Dixon, 2003). To assess how robust our similarity measurements were to small differences in the number of samples collected, we performed resampling on our data at the level of the electrofishing run. For each resampling, 10% of electrofishing runs were discarded randomly, and similarity indices were calculated for all pairwise pool combinations using the remaining data. We repeated this resampling procedure 500 times and then calculated mean and standard deviation of similarity for all site-year combinations. We estimated the correlation between similarities calculated using the different species lists, habitat types, or sampling year to assess whether these differences strongly influenced our conclusions about among-pool variation in fish community structure. That is, if we used the highly catchable species list, did our calculated similarity vary among pool combinations the same way as when we used the >1% catch species list?
Pearson's correlation coefficients were calculated using the base R function "cor()." Similarities were then related to physical distance using a simple linear regression model implemented using a Bayesian framework (McCarthy, 2007). Longitudinal distance was calculated as river km between the midpoint of one pool to the midpoint of another pool (Table S1). Distance and similarity were standardized prior to estimating regressions so that standardized slopes could be calculated (Hair, Anderson, Tatham, & Black, 1998). Priors (Bayesian probability distributions) for the slope were informative to reflect the assumption that sites further apart would be more different than sites closer to each other (although uninformative priors only slightly altered these estimates). We then used this regression to calculate the "predicted" similarity due to distance between Pools 19 and 20. If the actual similarity between Pools 19 and 20 was lower than the predicted similarity, then we considered this evidence that the dam is causing larger than expected impacts to the fish community. Similarities between 19 and 20 were included in the regression, but removal of those similarities did not alter the conclusions; as a result, we have retained them in the regressions. This analysis was undertaken separately for 2013 and 2014 and for side channel and main channel habitats.

| Similarity among pools on the Mississippi River
Calculated similarities in fish assemblages among pools were relatively insensitive to (a) data resampling, (b) choice of species list, (c) analysis year, and (d) habitat type (Table 3). (a) The resampling analysis indicated the mean similarity values calculated from the resampling procedure had a standard deviation (SD) equal to or less than 0.03 (the index ranges from 0 to 1;

| The relationship between distance and similarity among pools on the Mississippi River
The distance-similarity model showed a strong relationship between physical distance among locations and similarity of fish communities throughout the UMR (Pools 4-26). That is, pools separated by a greater distance had less similar fish communities. Because similarity values were not strongly influenced by the choice of species list, we used only the >1% catch list to generate the distance to similarity relationship. In both years and in both habitats, the association between distance and similarity had a standardized slope > 0.8 and R 2 values ranging from .66-.80 depending on year and habitat (Figure 3). The standardized slopes were very similar, with 95% confidence intervals that were credible and always had substantial overlap.   and 20 in 2014 is 1.7 standard deviations lower than the neighbouring pool average (0.64), but in 2013 the similarity between Pools 19 and 20 is very close to the average (0.74, Table 4). As a result, there is not clear evidence that Lock and Dam 19 is associated with a substantially larger similarity than occurs in other neighbouring pools.

| DISCUSSION
Our results showed a strong relationship between physical distance among locations and similarity of fish communities, consistent with previous research in this and other ecosystems (Araújo et al., 2013;Chick et al., 2006;Soininen et al., 2007). Previous researchers have focused on environmental conditions and dispersal barriers as potential drivers of distance-decay in community similarity (Morlon et al., 2008). In the UMR, pools separated by the largest physical distances are characterized by very different geomorphic features (i.e., environmental conditions) that likely contribute to differences in fish community structure. Upper reaches (Pools 4-13) are characterized by large backwater areas with lacustrine habitat and abundant vegetation.
Ictiobus spp., channel catfish (Ictalurus punctatus), white bass (Morone chrysops), freshwater drum (Aplodinotus grunniens), gizzard shad (Dorosoma cepedianum), common carp (Cyprinus carpio), and silver carp (Hypophthalmichthys molotrix) are dominant species in lower reaches preferring these conditions (Pflieger, 1997;Upper Mississippi River Conservation Committee, 2004). In addition, some species appear to be restricted to either the northern or southern reaches due to the species' thermal limitations. For example, yellow perch (Perca flavescens) prefer impounded backwater habitats with vegetation and slow water velocities consistent with conditions found in upper reaches and Pool 19 (Pflieger, 1997;Upper Mississippi River Conservation Committee, 2004). For these reasons, it seems likely that changes in environmental conditions are a major contributor to the distance-decay relationship.
By contrast, little evidence supported a strong role of dispersal barriers in driving differences in the overall fish community structure. Soininen et al. (2007) proposed using the similarity halving distance (the distance at which community similarity is reduced to half the initial similarity) as a metric of comparing distance-decay relationships among different ecosystems and taxa. Araújo et al. (2013) found that the halving distance for fish communities over 2 years in the undammed Tocantins River was 702 and 1,387 km, compared with 682 and 972 km in the 2 years of our study. The halving distance in the Tocantins River appeared to be much shorter after a large, impassable dam was installed (51% and 83% decrease). In Soininen et al. (2007), the halving distance of highly mobile taxa with few dispersal barriers (i.e., flying taxa) was about the same as those we observed in the UMR. Chick et al. (2006) considered the strong distance-decay relationship itself to be strong evidence against fragmentation as a major cause of variation in fish community structure in the UMR, instead reasoning environmental and geomorphic settings were more important. In addition, the most likely dispersal barrier (LD19) did not appear to be driving unexpectedly low similarity between Pools 19 and 20.
Several possibilities exist to explain why LD19 and other dams on the UMR appear to have little effect on variation of fish community structure. One possibility is simply that we lack the appropriate data to detect this effect. Abundant evidence suggests dams, even semipermeable navigation dams, can reduce or eliminate movement of certain fish species (Tripp, Brooks, Herzog, & Garvey, 2014;Wilcox et al., 2004;Zigler, Dewey, Knights, Runstrom, & Steingraeber, 2004). For this to influence community structure, these species must be significant contributors to the community. Many species that may have once   (Larson, Knights, & McCalla, 2017;Tripp et al., 2014), management actions within the lock chamber could presumably reduce the likelihood of upstream spread of invasive fish species as well as prevent the re-establishment of migratory species. Despite species having upstream access through the lock chamber, a greater abundance of invasive carps and migratory native species (i.e., skipjack herring and ebony shell mussel) occupy lower reaches (19 and below), presumably because LD19 has slowed the upstream migration of these species (Coker, Shira, Clark, & Howard, 1921;Kelner & Sietman, 2000;Nielsen, Sheehan, & Orth, 1986).
Regardless of LD19's current minimal impact on fish community structure, long-term monitoring in this ecologically important reach of the UMR could be useful to detect changes in fish community structure in future decades. The high-head structure of LD19 has caused deposition of more than 10 m of sediment behind the dam since its completion in 1913 (Bhowmik & Adams, 1989). Deposition of sediment has reduced water depth in the lower half of Pool 19, creating an impounded habitat from immediately above the dam to 24 river km upstream. The shallow depths and still waters in this impounded area provide ideal habitat for macrophyte colonization.
Aerial surveys have shown increased macrophyte expansion since 1966 (Tazik, Anderson, & Day, 1993;Thompson, 1973). Adams (1986, 1989) predicted Pool 19 will reach dynamic equilibrium by the year 2050 when the pool volume will be 20% of its initial post-impounded volume. In addition, this study focused only on catch in main channel and side channel habitats, but obviously the pool-wide population of fish species may change as the proportional contribution of habitat changes.
Our study provides important insight on the impacts of dispersal barriers, specifically high-head dams, to fish communities in large river systems. Although dispersal barriers can limit the range of certain species and create abrupt transitions in community structure, our findings describe a strong relationship between physical distance influenced by habitat structure/availability and environmental conditions and fish community similarities.

ACKNOWLEDGEMENTS
This project was partially funded by the Department of Interior, U.S.  Note. Mean and standard deviation were calculated after 10% resampling. All pools are separated by navigation dams, but the dam separating Pools 19 and 20 is a high-head dam that is suspected to restrict fish passage. The ">1% catch" species list refers to all species that occupy greater than 1% of the total catch in at least one pool. The "highly catchable" species list refers to species that have been found to be particularly vulnerable to the daytime electrofishing used in this study.
Jacques for critical review of the manuscript. We thank Karen Rivera

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.