Long-term changes in unionid community in Kentucky Lake: Implications for understanding the effects of impoundment on river systems

Abstract Freshwater mussels are both critically important in their ecosystems and rapidly declining around the world. Damming is a key reason for this decline in many locations because it affects the flow and turbidity of river systems, leading to numerous detrimental effects on benthic communities. Although the ecological effects of impoundment have been well studied on timescales ranging from years to decades, the ecological effects of impoundment on longer (50–100 years) timescales are less well understood, with a key question being: how long after the building of dams and impoundments do we expect community structure to continue changing? In this study, we explore historical changes in the freshwater mussel assemblages in Kentucky Lake (dammed in 1944) using decades-long collections housed at Murray State University in combination with other historical records. After digitizing these collections and applying a robust rarefaction protocol to account for uneven sampling, we quantify changes in unionid assemblage structure alongside coeval water quality data collected through the Kentucky Lake Long-Term Monitoring Program. We find that subsampled richness exhibited declines after dam construction with losses among opportunistic taxa, channelization-tolerant taxa, and impoundment-intolerant taxa. We also find increases in the proportions of equilibrium taxa throughout the dataset. Overall, the assemblage composition reached an equilibrium by the year 2000 (50 years after impoundment). In concert, river water quality data show a decline in turbidity and increase in light penetration in the period 1988–2020. Although the geohistorical records treated in this study are patchy in time, we argue that they are nonetheless valuable and illustrate that freshwater ecosystems may serve as potential sites of restoration decades after anthropogenic disturbance. In turn, this emphasizes the importance of geohistorical collections to studying long-term changes in community structure and developing strategies for conservation and environmental remediation.


Introduction
Freshwater mussels belonging to the family Unionidae are both diverse (Lydeard and Mayden 1995) and critically endangered (Vaughn and Taylor 1999;Haag and Williams 2014).Like most bivalved mollusks, unionids provide numerous ecosystem services to their communities, including water filtration, nutrient cycling, and serving as a food source for larger predators (Kryger and Riisgård 1988;Wotton et al. 2003;McMahon et al. 1991;Barbour et al. 1999;Vaughn and Hakenkamp 2001;Prins et al. 1996;Welker and Walz 1998;Hakenkamp and Palmer 1999;Haag 2012).Additionally, unionids are sensitive to a wide range of abiotic changes, most notably changes in flow, turbidity, depth, and siltation (McMahon et al. 1991).Consequently, unionids are subject to numerous (and often anthropogenic) sources of ecological stress, including eutrophication, impoundments, sediment input alteration, temperature changes, and diversions (Jackson et al. 2001;Giller 2005;Søndergaard and Jeppesen 2007;Haag 2012).In particular, anthropogenic impoundments, of which there are over 80,000 in the United States alone (Ho et al. 2017), have led to habitat fragmentation, silt accumulations, declines in the species richness of unionid mussels, and changes in water quality (Watters 1999;Søndergaard and Jeppesen 2007;Pilger and Gido 2012).Damming has also hindered the ability of fish to move, reproduce, and thrive within rivers, which -given that most unionids have a larval stage that relies on parasitism of specific fish species (Kat 1984;McMahon et al. 1991;Keller and Ruessler 1997;World Commission on Dams 2000;Lydeard et al. 2004;Haag 2012) -can be detrimental to mussel reproduction (Søndergaard and Jeppesen 2007).
Consequently, unionids are experiencing some of the highest current extinction rates of any group of freshwater organisms (Vaughn and Taylor 1999;Haag and Williams 2014).Both as a consequence of their sensitivity to environmental changes and the crucial ecological roles they play in freshwater communities, unionid mussels are an effective biological group for monitoring the health of freshwater ecosystems, many of which are currently undergoing dramatic changes across the United States (Lydeard and Mayden 1995;Quinlan et al. 2015).Despite this, more work needs to be done to completely understand which factors are most detrimental to various mussel species and what steps can be taken to preserve these communities (Lydeard and Mayden 1995;Haag and Williams 2014;Quinlan et al. 2015;Lopes-Lima et al. 2018).Moreover, given the near ubiquity of the anthropogenic impacts listed above, there is a premium on identifying sites that are suitable for restoration efforts.
In this context, a key question with wide relevance to the monitoring and restoration of freshwater unionid ecosystems is: how long after the building of dams and impoundments do we expect assemblage structure to reach an equilibrium?Although impoundments have been shown to have dramatic effects on benthic community structure on the scale of months to years (Armitage 1978;Voshell and Simmons 1984;World Commission on Dams 2000), there is a relative paucity of high-resolution data surrounding the responses of freshwater communities to environmental change on the scale of decades to centuries (Strayer and Dudgeon 2010;White 2014;White et al. 2020).Previous work has assumed that the unionid communities equilibrate in the first few years after the river is impounded, but the lack of decades-long mussel assemblage data has made it difficult to test that assumption (Rickett and Watson 1994;White 2014).Understanding ecological responses over longer timescales is critical to informing conservation efforts, both in helping to identify the signatures of perturbed communities and understanding the long-term ecological impacts of damming (Peacock and Mistak 2008;White et al. 2020).
In this study, we combine historical collections of unionid mussels (spanning the years 1977-2001) collected from Kentucky Lake with historical assemblage data collected by Sickel et al. (2007Sickel et al. ( ) (spanning the years 1930Sickel et al. ( -2001)), which record the long-term response of unionid communities to the construction of the dam in 1944.We analyze community data alongside long-term physiochemical water quality records in order to understand what the long-term environmental effects of impoundment are and how they may have impacted (or are still impacting) benthic ecosystems.These datasets allow us to address specific questions surrounding the long-term effects of dam construction in Kentucky Lake, principally: 1) on what timescales do unionid mussel communities continue to change post-impoundment?And 2) what are the notable temporal trends in richness, assemblage structure, and water quality in Kentucky Lake?Quantifying changes in richness and assemblage structure will reveal whether unionid communities are still changing 40-60 years after dam construction, or, alternatively, whether communities have reached a new ecological equilibrium.Alongside this, our long-term dataset recording water quality will aid in determining if any observed community changes are the result of impoundment or other anthropogenic factors, and potentially whether Kentucky Lake is suitable for further restoration efforts.

Locality and historical context
Located in southwest Kentucky and northwest Tennessee (Figure 1), Kentucky Lake is an ideal locality to study unionid bivalves because of the Southeast's high endemism and diversity of the group (Benz and Collins 1997).With over 300 freshwater bivalve mollusk species, the Southeastern United States is the richest locality for Unionoida in the world (Lydeard and Mayden 1995;Master et al. 1998;Haag and Williams 2014).Kentucky Lake is a mainstem reservoir that was created when the Tennessee River was dammed in 1944 (Tennessee Valley Authority (TVA)), 1985; Yurista et al. 2004;Foresta 2013;White 2014).Kentucky Lake has a surface area of 651 km 2 , a mean discharge of 1,812 m 3 sec −1 , and a mean retention time of 16.7 days (White 2014).Water velocities, discharge, and retention time are highly variable, changing from lake-like to river-like depending on the location within the reservoir and anthropogenic manipulations in response to flood or drought conditions (Yurista et al. 2004;White 2014).The substrate of this river was gravel before damming, and this area of Kentucky Lake now has fine silt coverage with areas of oxygen depletion during low flow (Sickel et al. 2007).Most macroinvertebrates are restricted to the shallow, marginal areas of the lake, which represent the now submerged former floodplain.The former floodplain has a mean water depth of 6 m in the summer with patches ranging between erosional and depositional settings (White 2014).Along with water quality data monitored since 1988, freshwater mussels have been collected at the Hancock Biological Station (HBS) and stored at the Murray State University ('MSU') collections (located in nearby Murray, Kentucky) since 1950.Historical accounts of freshwater mussel species and counts in the Kentucky Lake area have been recorded since 1930 (Sickel et al. 2007).

Historical mussel collections
Local-scale collections, such as those hosted at MSU, provide an excellent opportunity to study historical changes in assemblage composition because of the amount, detail, and quality of information typically available within them that is frequently not duplicated in larger, regional collections (Snow 2005).The MSU collections were largely collected and organized by Dr. James Sickel from several localities around the United States, although collections mainly occurred in Kentucky and Tennessee.Almost all specimens were collected through brailing, meaning that mussels were overwhelmingly likely to have been collected as live samples, rather than representing dead (and thus potentially older) material (Miller and Nelson 1983).The database resulting from the digitization of the collections comprises 3471 specimens with information inclusive of specimen collection date, location, collector, and species identification.Species identifications have been previously verified by the state malacologist from the Kentucky Department of Fish and Wildlife and are updated here following the Integrated Taxonomic Information System (ITIS) (http://www.itis.gov).
For this study, we restricted records to only those collected from Tennessee River Mile 22.4 to 66, which is the location of Kentucky Lake immediately upstream of the Kentucky dam; this geographically restricted dataset comprised 270 specimens.We integrate this new dataset with that of Sickel et al. (2007), which provides a pre-dam ecological baseline from 1930 and additional Kentucky Lake samples .The resulting dataset comprises 5307 individual records including the years 1930 and 1977 to 2001, comprising 36 unique species (Appendix A).
Geohistorical (similar to paleontological) community data frequently suffer from biases resulting from sampling effort -time intervals with more intense sampling (i.e. more specimens) will tend to have artificially inflated species richness compared to those with less intense sampling (e.g.Foote 2001).As the sampling intensity in any year increases (producing greater numbers of specimens), the observed species richness of that time bin will tend to increase (e.g.Alroy 2010).As an illustration of this, rarefaction curves constructed for specimens from individual years in our dataset clearly show a high proportion of under-sampled years (Appendix B).To combat this issue, a process of rarefaction (also known as subsampling) is commonly applied, where consistently sized subsamples are iteratively drawn from species pools in order to generate a vector of sample-standardized community data (e.g.Darroch and Wagner 2015).In our dataset, sampling intensity varied greatly among the collection years with a maximum of 1645 samples in 2001 and several years during which no specimens were collected at all.First, we applied a 'cutoff ' , whereby we only analyze years during which a minimum number of specimens was collected.Many of the years with intermediate to better-sampling possess between 10 and 30 specimens; we therefore randomly subsample five specimens from years during which a minimum of 10 specimens was collected.There are no robust statistical guidelines for choosing the size of the subsample relative to the size of the cutoff value, although a range of between 50-75% is typical (e.g.Darroch and Wagner 2015).As an additional sensitivity test, therefore, we also experimented with different cutoff values and subsampling numbers, allowing us to examine whether rarefied species richness trends are sensitive to the cutoff threshold or subsampling intensity (Appendix C).We iterate (×100,000) this subsampling routine, counting the number of unique species in each iteration.This analysis produced a vector of sample-standardized richness estimates for each year sampled, and which reflects sample-standardized temporal trends in overall richness (hereafter referred to as, 'rarefied richness') within Kentucky Lake over the studied interval (Figure 2).
To better understand changes in the ecological composition of unionid assemblages, all species in the dataset were classified using information provided in Haag (2012) in terms of their channelization tolerance, impoundment tolerance, and life histories (Figure 3).Only years during which at least 15 specimens were collected were analyzed and plotted for this analysis to reduce skewing by low collection years.Life history categories are defined by Haag (2012) and include equilibrium strategists, opportunist strategists, and periodic strategists.Equilibrium strategist species recruit constantly at low rates and have long life spans with late maturity.Opportunist strategist species recruit at high and variable rates with early maturity and short life spans.Periodic strategist species also have high and variable recruitment rates with intermediate life spans (Haag 2012).Proportional abundance, a commonly used metric in paleontological studies, is a robust metric of community structure regardless of sample size and preserves with high fidelity (Kidwell 2001;Kidwell 2002a;Kidwell 2002b;Kidwell 2008;Olszewski and Kidwell 2007;Kidwell 2013).Evaluation of relative mussel abundance in this study allows the understanding of which types of species are dominating the freshwater mussel communities of Kentucky Lake post-impoundment.

Water quality data
The Kentucky Lake Long-Term Monitoring Program, maintained by the MSU Watershed Studies Institute and HBS, has been collecting physiochemical water quality data every 16 days from 12 sites in Kentucky Lake since July 1988 (Figure 1; Kentucky Lake Long-Term Monitoring Program Database, Hancock Biological Station, Murray State University, Murray, KY 42071).All water parameters were measured from the lake's bottom, which reflects the water conditions most affecting the mussels.The main water quality variable of interest, turbidity, quantifies the amount of light scattering (Davies-Colley and Nagels 2008) and is inversely related to water clarity, which impacts light penetration (Davies-Colley and Smith 2001).Moreover, turbidity here is taken to reflect how much fine sediment is suspended in the water column (Davies-Colley and Smith 2001).Additional water quality factors of interest are chlorophyll a (a proxy for primary productivity), total dissolved phosphorous and dissolved organic nitrogen (measures of eutrophication), pH, dissolved oxygen, chlorine, temperature, 1-m light, and 5-m light.Turbidity measurements are shown in Appendix D, along with other water quality measurements.

Data analysis
Mean rarefied species richness values for each year were plotted and a linear regression applied to illustrate richness trends through time (Figure 2).Rarefied species richness trends were analyzed using Spearman's rank correlation coefficients ('Spearman's rho'; Figure 2; Appendix C).Spearman's rho is a non-parametric rank-based test that is commonly used with historical and paleontological datasets, and it assesses how the relationship between two variables can be described using a monotonic function that does not assume normality of the data.Environmental water quality and chemistry data from the Kentucky Lake Long-Term Monitoring Program were plotted against time and temporal trends were illustrated using locally estimated scatterplot smoothing ('LOESS') lines (Appendix D).Analysis of community structure was performed by first filtering out all years during which fewer than 15 specimens were collected.Proportions of species belonging to different categories (Figure 3; impoundment tolerance, channelization tolerance, and life history) were calculated for each of the years and plotted.A Generalized Additive Model (GAM) was fitted for each group identity using the 'mgcv' package in R (Wood 2023).GAMs are nonparametric extensions of linear regression modeling that allow for nonlinear predictors, which is ideal for this dataset which contains noticeable trends and fluctuations throughout which would not be well-represented by linear regression modeling.All analyses were performed in R version 4.2.2 (R Core Team 2022) in RStudio version 2023.03.0 + 386.

Species richness and community structure
Using the 1930 data from Sickel et al. (2007), we can compare the species found in Kentucky Lake before and after dam construction in 1944 (Appendix A).21 distinct species were collected in 1930 before the dam was built, 9 of which were not present in the collections in Kentucky Lake after dam construction and seem to have been extirpated from the lake.The remaining 12 species that were recorded in 1930 were present throughout the dataset, with only one species (Cyclonaias tuberculata) being extirpated before the end of the dataset in 2001.14 new species were recorded in 1977-1981, and only one of those species, Tritogonia verrucosa, was only collected during one year.Five species were last recorded between 1984 and 1987, and one species was last recorded in 1993.The remaining species in the dataset were found in the Kentucky Lake until the end of the dataset in 2001.
Linear regression reveals that the rarefied richness counts decreased between 1930 and 2001 (Figure 2; corr = −0.355,p < 2.2 e-16).This trend was robust to changes in the intensity of subsampling in our rarefaction analyses, maintaining significant decrease in rarefied richness towards the modern when anywhere between 2-10 specimens are iteratively subsampled (Appendix C).The trend was also robust to changing the threshold of specimens required to be collected during a year in order for that year of data to be included in the rarefaction protocol (and hence inclusion in the final dataset; see Appendix C).Spearman-rho statistics illustrate the consistency of decreasing richness trends (consistently negative correlation values, Score range = −0.08 to −0.31) with all p-values being less than 2.2 e −16 .
In terms of community compositions, the generalized additive model results indicated the proportion of channelization tolerant species significantly decreased over the study interval (Figure 3(a); p = 0.003).In Figure 3(b), it is apparent that impoundment intolerant species were extirpated from the dataset after 1930.One impoundment intolerant specimen was collected in 1987, but because there was only one specimen collected in the dataset during that year and the minimum number of specimens collected in this analysis is 15, that year of collections was excluded from this analysis and is not represented in Figure 3(b).The proportion of highly impoundment tolerant species increased over the study interval (p = 0.090) and the proportion of mildly impoundment tolerant species decreased over the study interval, although this is a weaker relationship (p = 0.136).Figure 3(c) shows that the proportion of opportunistic species decreased over the study interval (p = 0.037), the proportion of periodic species remained constant (p = 0.098), and the proportion of equilibrium species increased towards the present (p = 0.057).

Water quality
Indicators of water quality revealed declines in turbidity, dissolved total phosphorus, and dissolved organic nitrogen between 1988 and 2020 (Appendix D).Primary productivity, pH, and dissolved oxygen were stable within the same period, while increases are seen in chlorine, temperature, 5-m light, and one-meter light (Appendix D).Seasonal variations corresponded with what previous studies have found in Kentucky Lake (e.g.Yurista et al. 2004).

Long-term temporal trends in rarefied richness and community structure
Quantitative analysis of historical unionid collections reveals changes in species richness and assemblage structure in Kentucky Lake between the pre-dam sampling completed in 1930 and the period 1977-2001, potentially providing an answer to our first question ('on what timescales do unionid mussel communities change post-impoundment?').The data show changes in richness and community structure from 1930 (i.e.pre-dam) to 2001, involving a marked drop in rarefied richness and increases in the proportions of channelization intolerant and impoundment tolerant species.Although we do not have water quality data from this same period, the observed changes make intuitive sense in context of impoundment -reduced current velocities changing habitats from lotic to lentic and resulting in substantial deposition of silt.This also matches with observations from other lakes across the southeastern US that have been affected by dam construction (see e.g. Bates 1962;Blalock and Sickel 1996;Sickel et al. 2007).These records paint a detailed picture of changing community structure in Kentucky Lake in response to impoundment and indicate that many species were unable to adapt to the more lentic, lacustrine conditions of the reservoir or that their fish hosts have become absent (from the lake as a whole or from the depths at which mussels now live).These data thus indicate that mollusk communities reorganized relatively rapidly (<40 years) in response to impoundment.
The question as to whether communities are still changing (and thus the effects of dam construction are ongoing) is harder to answer.When rarefied species richness data are plotted at annual resolution (Figure 2), there is a significant downward trend, which suggests that the dam is still having an impact on communities 40-60 years after construction.When assessing community structure since 1980, it seems that structures have experienced large changes from before damming (1930) to 1980 but have largely stabilized since then in terms of impoundment tolerance (Figure 3).Our dataset combining records from the MSU collections and culled from the literature thus indicate that communities changed extensively after construction of Kentucky Dam, but that the rate of change has likely slowed in the 20 years spanning ~1980-2000.

More recent environmental changes
Some additional context can be gleaned from examining records of water quality collected from the Hancock Biological Station.While our water quality data and mussel data possess limited overlap in time (and thus prevent a complete comparison between the two) the water data shows only minor shifts -principally a decrease in turbidity and increase in 5-m light penetration depth in the period 1990-2020 -providing an intuitive correlation with stable mussel community structure.The observed changes in water quality are, however, unusual in context of an impounded river system where lower water velocities are expected to lead to an increase in find sediment deposition (siltation).
We hypothesize that that the observed decrease in turbidity (and associated increase in light penetration) is the result of reduction in suspended sediment.Although a decrease in the concentration of suspended plant matter could also be a plausible driver of increasing light penetration (Gallegos and Jordan 2002), our data on chlorophyll concentration (a proxy for phytoplankton density) is relatively stable through the studied interval (Appendix D), indicating that reduced sediment load is the most likely culprit.The reasons for declining levels of suspended sediment are, however, unclear.One plausible explanation is that Kentucky Lake is a 'mainstem' reservoir, meaning that human manipulation and rain events can change the flow regime from lotic to lentic (Yurista et al. 2004;White 2014).Because the lake has maintained a strong flow relative to when it was a wild river (Yurista et al. 2004), it is possible that the majority of the silt in the river has remained entrained in the flow and has not settled out as much as expected with the addition of the impoundment.In addition, there has been a relatively recent effort to reduce soil erosion in the Kentucky Lake area (United States Department of Agriculture (USDA) and Forest Service 2004), which may contribute to declining turbidity in the river.While it is unclear to what extent these efforts have been implemented, efforts have included promoting the use of natural vegetation buffers along waterways and installing fencing to prevent livestock from entering waterways, which lowers the potential for cattle to trample the mussels and introduce waste into the water (Windsor 2000).The eastern side of Kentucky Lake is bordered by the Land Between the Lakes National Recreation Area which has been uninhabited since its created in 1963 and is ~95% forested (Yurista et al. 2004).Thus, strategies designed to prevent soil erosion could be especially important for the western edge of the lake, where land use is predominantly agricultural with several small towns and less than 50% forest cover (Yurista et al. 2004).The rural counties on the western side of Kentucky Lake have small populations (Calloway County, KY: 37,103 people, Marshall County, KY: 31,163 people, and Henry County, TN: 32,056 people in 2020) and slow population growth compared to more urban areas (none of these counties has doubled in population since 1940 whereas Nashville's Davidson County has nearly tripled in population during the same interval) (https://www.census.gov/programs-surveys/decennial-census.html).This limited population increase has prevented mass urbanization of the landscape seen near other impoundment and may allowed soil preservation measures to have a substantial impact on sediment input to the river (Tennessee Valley Authority (TVA) 1985; United States Department of Agriculture (USDA) and Forest Service 2004).
Another possible driver of the decline in suspended sediment is damming upstream of this area; Kentucky Dam is the last of nine dams on the Tennessee River (White 2014), which could be responsible for reducing the fine sediment suspended load entering Kentucky Lake.However, a problem with this hypothesis is that most of the dams upstream were built between 1930 and 1950, so it is unclear why turbidity has been steadily declining into the present.Future work will focus on sampling the modern mussel community and assess to what extent the ecosystem has continued to change in the most recent two decades.
In North America, the biggest threats to freshwater mussel biodiversity (in addition to river impoundment) include land-use changes (often with associated eutrophication-induced impacts), pollution, exploitation, and the introduction of invasive species (see e.g.Vaughn and Taylor 2000).In terms of land-use, the only notable change in this area since impoundment in 1944 has been the establishment of the Land Between the Lakes National Recreation Area in the 1960s (Nickell 2007).Since then, the eastern shore of the lake has remained forested, while the western shore has consistently comprised a mixture of small towns and farmland (Yurista et al. 2004).Moreover, there is little evidence for eutrophication-induced impacts such as harmful algal blooms (given that both dissolved total phosphorous and dissolved organic nitrogen have both decreased through time), or hypoxia/anoxia (given that levels of dissolved oxygen have stayed broadly consistent) (Appendix D).There is similarly little evidence for pollution, with no evidence of toxic waste or unusual pollutants found in previous studies (e.g.White 2014) and little reason to believe that unionid populations in this area are being over-harvested.Lastly, in terms of invasive species, zebra mussels (which have posed significant threats to freshwater ecosystems elsewhere in the United States -see e.g.Nalepa and Schloesser 1992;Strayer 2009) have not yet become established or prevalent within Kentucky Lake (White 2014).While zebra mussels were introduced to Kentucky Lake in the early 1990s, the lack of established colonies and low mussel densities within the lake (White 2014;Meystedt 2023) exclude the invasives as likely drivers of the reduced turbidity trends discussed above.

Implications for remediation efforts
Our combined community and environmental data allow us to make some rudimentary predictions about the current state of mussel communities in Kentucky Lake, as well as the potential for ongoing monitoring and remediation efforts.High turbidity, or the prevalence of fine sediment entrained in the water flow, harms freshwater mussels by increasing recruitment failure, inhibiting their ability to effectively filter out nutrients from the water column, and disrupting reproduction (Box and Mossa 1999;Gascho Landis et al. 2013;Gascho Landis and Stoeckel 2016;Goldsmith et al. 2020).The surprising decrease in turbidity and increase in light penetration over the last 30 years has been coupled with an increase in the proportion of equilibrium species between the 1980s and 2000.If the lake ecosystem has reached equilibrium, the combination of improving water conditions, absence of strong indications of eutrophication, pollution, or establishment of invasive species, may identify Kentucky Lake as an area suitable for the reintroduction of historically extirpated species and other remediation efforts.
One major limitation in restoration of freshwater mussels is the availability of fish hosts.Most larval freshwater mussel species parasitize fish, and consequently mussel reproduction is highly dependent on healthy populations of those fish species (Kat 1984;McMahon et al. 1991;Keller and Ruessler 1997;Lydeard et al. 2004).Illustrating this point, both Watters (1992) and Vaughn and Taylor (2000) found that the health of fish populations in freshwater systems acts as an effective indicator for the richness of mussels in the same system.Fish movement and reproduction is often greatly hindered by the damming of rivers (World Commission on Dams 2000), so the negative effects of impoundment on fish populations have a downstream negative effect on mussel populations.A detailed understanding of the preferred fish hosts and their ecology will therefore be crucial to ensure successful remediation attempts.

Conclusions
In summary, a quantitative analysis of geohistorical collections housed at MSU, in combination with more recent surveys, shows that that mussel richness has declined after dam construction with losses among channelization-tolerant and impoundment-intolerant or -mildly tolerant species.However, a combination of subsampling and generalized additive models suggests that community structure has since stabilized, in concert with a decline in turbidity and increase in light penetration over the period ~1980-2020.These findings suggest that Kentucky Lake may be a site that would benefit from dedicated restoration efforts, particularly if unionid-specific fish host species are also present.
Lastly, the results of this work emphasize the utility and importance of local scale, geohistorical collections such as that housed at MSU and highlight how these data can be brought to bear on pressing issues of ecological monitoring and conservation.Given that both digitization and analysis of geohistorical collections represent relatively inexpensive ways of examining long-term trends in community health, studies along these lines may present an affordable way forward to identifying freshwater river systems in need of monitoring and remediation.As part of this ongoing effort, we have made the digitized MSU collection data freely available upon request.

Figure 1 .
Figure 1.Map of land Between the lakes national recreation area in southwestern Kentucky, bordering tennessee.water quality sites (yellow circles and associated site abbreviations) are monitored by hancock Biological station (hBs), and mussel sampling localities (black dots) are sites recorded in the Murray state university (Msu) freshwater mussel collections.the green shaded land represents the land Between the lakes national recreation area.

Figure 2 .
Figure 2. yearly mean values of rarefied species richness in Kentucky lake through time based on 100,000 iteratively subsampled pools of 5 specimens for years during which > = 10 specimens were collected.the blue line represents a linear model line of best fit (correlation estimate = −.355,p < 2.2 e-16) and the grey shaded region indicates a 99% confidence interval.Because the confidence level is so high, the grey shaded region is present but extremely close to the line of best fit. the black points represent mean values for each year, but linear regression was calculated with all values.

Figure 3 .
Figure 3. temporal changes in the proportion of the assemblage belonging to categories provided and described by haag (2012).years during which at least 15 specimens were collected are represented in these plots.lines on all plots represent generalized additive Models (gaM) of best fit for the data.(a) Proportion of channelization tolerant species.gaM results: proportion of channelization tolerant species decreases significantly over the study interval (p = 0.003 with 93.7% of deviance in proportion explained by the time series).(b) Proportion of Impoundment tolerant species.Intolerant species are only present in 1930 (pre-dam) in this analysis.gaM results: the proportion of mildly impoundment tolerant species decreases over the study interval (p = 0.136 with 46.5% of deviance in proportion explained by the time series).the proportion of highly impoundment tolerant species increases over the span of the dataset (p = 0.090 with 55.2% of the deviance in proportion explained by the time series).(c) Proportion of Different life histories.the proportion of opportunistic species significantly decreases over the study interval (p = 0.037 with 81.3% of the deviance explained by the time series).the proportion of periodic species remains constant (p = 0.098 with 53.6% of the deviance explained by the time series).the proportion of equilibrium species increased significantly near the end of the study interval (p = 0.057 with 76.3% of the deviance explained by the time series).