Freshwater mussels are particularly susceptible to injury from releases of hazardous substances. Natural recolonization of injured mussel populations can take decades because of their complex life history. Hence, hatchery propagation and stocking of mussels is commonly used for recovering injured populations. In recent decades, several Natural Resource Damage Assessment and Restoration (NRDAR) cases have involved freshwater mussels, but none have analyzed whether restoration was successful. Our study represents the first evaluation of restoration success of freshwater mussels in an NRDAR context. Its purpose was to determine whether mussel restoration was successful for two large-scale, multiyear (>10 years) NRDAR cases in the Clinch and Powell rivers of Virginia and Tennessee. We used mussel release data compiled from 2004–2017 and a Leslie matrix model to estimate the expected abundance of mussels at nine restoration and monitoring sites. We then compared expected abundances to abundance values estimated from quadrat surveys conducted from 2015–2021 at these same nine sites. Estimated abundances were 57–85% lower than expected. We conducted mark-recapture surveys at two sites and the data from this independent method supported our quadrat survey results; i.e., abundance estimates were much lower than the expected abundance values. However, we observed evidence of successful restoration, such as released mussels reaching breeding ages and presence of mussels at low- to medium densities (0.02–0.48 m–2) at restoration sites, and we confirmed limited recruitment of two species. Nonetheless, lower-than-expected abundance suggests that either mussels are settling or recruiting outside of restoration sites and/or that survival and recruitment of released mussels are lower than expected. Further study is needed to determine to what extent each of these factors explain lower-than-expected abundance to better estimate the scope of restoration required in future NRDAR cases. Finally, we developed a set of metrics to determine whether restoration was successful in this study and for application to future cases involving freshwater mussels.
INTRODUCTION
Freshwater mussels provide many ecosystem services, including regulating services (biofiltration of water), supporting services (nutrient cycling and storage, habitat modification, and environmental monitoring), provisioning services (food and products made from shell), and cultural services (providing cultural and existence values) (Vaughn 2018). Unfortunately, freshwater mussels (Unionida) are among the most imperiled groups of freshwater organisms in North America (Vaughn and Taylor 1999; Lopes-Lima et al. 2018). Of the approximately 300 recognized species, 88 are listed as federally endangered and 15 are listed as federally threatened under the Endangered Species Act (U.S. Fish and Wildlife Service 2018). Water pollution and water quality degradation are among the most frequently cited causes of mussel decline (Downing et al. 2010). Due to their sessile nature, mussels are highly susceptible to injury from releases of hazardous substances into aquatic ecosystems. Releases of contaminants into rivers can drastically reduce the diversity and abundance of local mussel populations (Sheehan et al. 1989). Further, the limited dispersal capabilities of mussels and their complex life history (which involves fish hosts for dispersal) make natural recolonization difficult and unlikely in the short term (∼10–20 years) (Patterson et al. 2018; Irmscher and Vaughn 2018).
Natural Resource Damage Assessment and Restoration (NRDAR) regulations allow the federal government to assess injury to natural resources resulting from the release of a hazardous substance and to recover damages from responsible parties (43 CFR § 11). There have been numerous NRDAR cases involving injury to freshwater mussels in recent decades. Two are particularly relevant to our study. In 1996, a release of coal slurry from the Lone Mountain Processing Inc. (LMPI) facility near St. Charles, Virginia, impacted a large section of the Powell River, affecting 15 species of federally listed endangered mussels present in the river (U.S. Fish and Wildlife Service 2003). Two years later, a tanker truck operated by Certus Inc. overturned in Cedar Bluff, Virginia (U.S. Fish and Wildlife Service 2004); the resulting spill of a hazardous chemical killed an estimated 18,621 mussels of 14 species, three of which were listed as federally endangered at the time and one other that was listed as endangered after the spill (Hyde and Jones 2021). Additional examples include a decades-long mercury release from the DuPont-Waynesboro facility that affected mussel populations in the South River from 1929 to 1950, resulting in a $4 million settlement for mussel restoration in the South River and South Fork Shenandoah River (U.S. Fish and Wildlife Service 2017); the 1999 release of hazardous substances from a ferro-alloy manufacturing facility that killed over 990,000 mussels in the Ohio River, including individuals of two federally listed endangered species (U.S. Fish and Wildlife Service 2007); and a 2014 coal ash spill in the Dan River (Dan River Natural Resource Trustee Council 2020) that adversely affected the federally endangered James Spinymussel (Parvaspina collina) and other mussel species. Together, these cases show that injury to freshwater mussel populations is an ongoing concern.
In these cases, injury assessment and restoration goals varied considerably. For the DuPont-Waynesboro case, an injury of 650,000 mussels was estimated by using reference sites, historical species composition, and impacted habitat to determine expected density and applying the estimated density to the area injured (U.S. Fish and Wildlife Service 2017). For the Ohio River NRDAR case (U.S. Fish and Wildlife 2007), injury was established by documenting the presence of dead mussels immediately downstream from the discharge area of the facility. Surveys at one downstream site found mussel mortality to be 100 percent, and additional surveys estimated 990,000 mussels killed. The selected restoration alternative included translocation of adult mussels, release of infested host fishes, and propagation and release of juvenile mussels. The goal was to restore density in the affected areas to one mussel per square meter, requiring long-term survival of 195,000 individuals to age five. In the Dan River NRDAR case, a Habitat Equivalency Analysis (HEA) was used to determine the level of restoration needed for gains in services to equal the loss of services from the injury. In this case, mussel propagation was not among the selected restoration activities, although three species (Yellow Lampmussel (Lampsilis cariosa), Green Floater (Lasmigona subviridis), and Notched Rainbow (Venustaconcha constricta)) were successfully propagated by the Virginia Fisheries and Aquatic Wildlife Center as part of the pre-NRDAR restoration process (Brian Watson, Virginia Department of Wildlife Resources, personal communication). Rather, several habitat restoration and conservation alternatives were used, including the transfer of 618 acres of land to North Carolina and Virginia state parks and the removal of the Pigg River Power Dam (Dan River Natural Resource Trustee Council 2020). Injury to mussel populations for the Certus Inc. NRDAR case was quantified by counting the number of fresh-dead mussels in the affected area and multiplying by three to account for mussels buried in the substrate (U.S. Fish and Wildlife Service 2001). The primary restoration goal for the Certus Inc. case was to restore the mussel assemblage to approximate baseline conditions, i.e., “the condition of the natural resources and services that would have existed had the incident not occurred,” which was accomplished by propagating and releasing most species injured in the spill (U.S. Fish and Wildlife Service 2004). Injury for the LMPI NRDAR case was largely sublethal and defined, but not quantified, as acute (time of spill) and chronic (resuspension over time) toxicity from exposure to hazardous substances, indirect mortality of glochidia due to loss of host fishes, and indirect losses due to habitat degradation from silt and sedimentation (U.S. Fish and Wildlife Service 2003). The primary goal was to restore mussel assemblages to approximate baseline conditions, which was accomplished by propagating and releasing mussels from a targeted suite of injured species in the mussel assemblage. Compared to more recent cases, baseline conditions for the Certus Inc. and LMPI NRDAR cases were less explicitly defined, and it was difficult to measure success of restoration based on the resulting restoration goals to restore to pre-injury baseline. These examples clearly demonstrate that the methods for determining the extent of injury and subsequent required restoration vary widely from case to case.
Natural recolonization of injured mussel assemblages may take many years, during which time the services they provide would be lost. Given this lag time, release of propagated mussels to restoration sites is a common action in NRDAR cases. Stocking mussels satisfies the “restoration or rehabilitation of injured natural resources to a condition where they can provide the level of services available at baseline” criteria of NRDAR regulations (43 CFR § 11). Further, NRDAR regulations also allow for “the replacement and/or acquisition of equivalent natural resources capable of providing such services. . .” in lieu of, or in addition to, restoration/rehabilitation (43 CFR § 11). In cases involving injury to freshwater mussels, replacement of equivalent services equates to releasing mussels outside of the impacted area or the use of species capable of providing services similar to those afforded by the injured species (replacement/acquisition). It also may be necessary to correct for differences in services provided by juvenile vs. adult mussels because most propagation involves releases of mussels <5 years old, which might not provide the full suite of services afforded by older individuals (Patterson et al. 2018). The extent to which mussels successfully establish at a site poses implications for the number of mussels that need to be released for successful restoration. A lower rate of establishment and survival at a site would require releasing a higher number of mussels over a longer period compared to a site with higher rates of mussel establishment and survival. Further, if mussels are experiencing higher mortality after release than what they would experience naturally, there would be a corresponding decrease in the expected services provided over their lifetime.
Few published studies have analyzed the success of mussel restoration, particularly in a NRDAR context. Although long-term monitoring is conducted in some cases (e.g., restored populations of P. collina have been monitored for over 10 years), results of such monitoring are seldom published in the primary literature (Lavictoire and West 2024). Thus, the purpose of our study was to monitor restored mussel populations in the Clinch and Powell rivers for the Certus Inc. and LMPI NRDAR cases to determine whether, and to what extent, restoration efforts for these cases were successful. The successful restoration of injured resources assumes that released mussels are establishing and reproducing at restoration sites. We developed the following metrics for assessing successful mussel restoration: (1) settling into suitable habitat and surviving after release, (2) surviving at rates high enough to reach breeding age, (3) being fertilized, resulting in gravid females, (4) producing recruits that successfully establish, and finally (5) that released mussels and their recruits continue successfully breeding to the point that the mussel assemblage is self-sustaining and stable in the long term (Table 1). These criteria are predicated on the successful establishment of released mussels in sufficient numbers. We focused only on abundance, density, and growth for the Certus Inc. and LMPI cases, because these metrics were developed after the design and implementation of the monitoring program for each case. Future monitoring designs should attempt to measure as many of these metrics as feasible. The objectives of this study were to (1) estimate the expected number of mussels surviving at restoration sites, (2) estimate abundance and density of mussel species at restoration sites, (3) determine whether estimated abundance and density differed from expected abundance and density at restoration sites, (4) determine the shell length growth rate of released mussels, and (5) determine whether, and to what extent, restoration goals were achieved for the Certus Inc. and LMPI NRDAR cases.
Detailed Case Background
The Certus Inc. and LMPI Natural Resource Damage Assessment and Restoration (NRDAR) cases in the upper Tennessee River basin of Virginia are among the first and largest cases in the United States involving injury to freshwater mussels due to release of hazardous substances (Hyde and Jones 2021). The Certus Inc. chemical spill released 5,110 liters of Octocure-554 revised, a rubber accelerant, into a tributary of the Clinch River when a tanker truck overturned in Tazewell County, Virginia, on August 27, 1998. An estimated 18,621 mussels, including 750 individuals of three endangered species (Golden Riffleshell (Epioblasma aureola), Purple Bean (Venustaconcha trabalis), and Rough Rabbitsfoot (Theliderma strigillata)), were killed along an 11-kilometer section of stream (U.S. Fish and Wildlife Service 2004). Further, since the spill, both Fluted Kidneyshell (Ptychobranchus subtentus) and Slabside Pearlymussel (Pleuronaia dolabelloides) have been listed as endangered. The loss of their local Clinch River populations likely contributed to the listing of these species.
The LMPI coal slurry spill occurred when a holding pond failed at a processing plant in Lee County, Virginia, on October 24, 1996. The spill released 22.7 million liters of coal slurry into a series of tributaries of the Powell River. The resulting “blackwater” impacted a large section of the Powell River, and coal fines and sediment ultimately were deposited in Norris Reservoir, Tennessee, 105 kilometers downstream from the release. Although no dead mussels were found, coal fines later were detected in mussel gut tissues (U.S. Fish and Wildlife Service 2003). Additionally, at least 11,240 fish of various species were killed, some of which are host fishes for the 15 federally endangered mussel species found in the impacted river reach (U.S. Fish and Wildlife Service 2003). Coal fines and sediment also were deposited in the substrate throughout the affected length of the Powell River and likely continued to impose chronic, sub-lethal impacts due to resuspension during high flow events in 1996 and 1997 (U.S. Fish and Wildlife Service 2003). In contrast to the acute, lethal effects of the Certus Inc. spill, the LMPI spill represented a chronic, sub-lethal effect on the mussel fauna in the impacted river reach (Michalak et al. 2017).
Table 1.
Monitoring criteria and evidence required for documenting successful establishment and reproductive success of restoring freshwater mussel populations.
The principal goal for each case was “to restore the mussel assemblage and its supporting habitats to approximate baseline conditions” (U.S. Fish and Wildlife Service 2003; U.S. Fish and Wildlife Service 2004). Baseline condition for the Certus Inc. NRDAR case was the estimated number of mussels (18,621) and respective species composition present in the impact zone before the spill (Hyde and Jones 2021). Consequently, many mussels released as part of restoration were at sites in the immediate impact zone of the Clinch River between Cedar Bluff, Virginia (River Mile (RM) 324) and Richlands, Virginia (RM 318). However, mussels also were released downstream at other restoration sites in the Clinch River in Russell County, Virginia (RM 270–277.5), to reduce the risk that released mussels would all be impacted by another single, catastrophic event or degradation of habitat in the areas of Cedar Bluff and Richlands. Further, the ability to propagate some affected species was limited. Notably, Oyster Mussel (Epioblasma capsaeformis) and Cumberlandian Comb-shell (E. brevidens) were used as surrogates for the critically endangered E. aureola due to the greater availability of brood-stock and ease of propagating these two species at mussel hatcheries. Specifically, these two Epioblasma species were used as surrogates to develop propagation, culture, and monitoring techniques for E. aureola. Baseline condition of the mussel assemblage was not quantified for the LMPI NRDAR case; however, the goal was to propagate a selected suite of the federally listed and non-listed mussel species affected by the spill in the Powell River. Not all federally listed mussel species impacted by the spill could be propagated and restored due to technological limitations (e.g., undeveloped propagation techniques including unknown host fishes); thus, restoration efforts in this case mainly focused on releasing E. capsaeformis and E. brevidens, as well as numerous non-endangered species, at sites in the Powell River to establish robust populations of these species and to restore their local populations and respective ecosystem services.
METHODS
Study Area
Mussels were sampled at six release and monitoring sites in the Clinch River (Fig. 1) and three of six release and monitoring sites in the Powell River (Fig. 1). These nine sites were sampled from 2015 to 2017 and again in either 2020 or 2021 (Table 2). This period was chosen because, by 2015, most of the restoration had been conducted for the Certus and LMPI NRDAR cases. However, monitoring at the Oakley Property in the Powell River, Tennessee, only began in 2016 and therefore it was not sampled in 2015.
Local mussel populations at two of the sites in the Clinch River, the Sycamore Lane site near Richlands, Virginia (RM 320), and the Payne Property site near Cedar Bluff, Virginia (RM 322.1), in Tazewell County, Virginia, were impacted directly in 1998 by the Certus Inc. chemical spill. These sites were selected for restoration because they represented some of the best and largest available mussel habitat patches in the impact zone. Four additional sites located further downstream in Russell County—Bennett Property (RM 277.5), Artrip (RM 274.5), Whited Property (RM 272.7), and Cleveland Island (RM 270)—were not directly impacted by the spill. However, they were chosen as additional restoration sites for the Certus Inc. NRDAR project to reduce risk for potential future impacts to the two sites located in the impact zone between Cedar Bluff and Richlands, Virginia, as well as to replace/acquire natural resources and services equivalent to those lost during the spill. Additionally, these four sites had suitable habitat for mussels, a viable mussel assemblage, and suitable host fishes. All three Powell River monitoring sites were within the area affected by the LMPI coal slurry spill and are in Claiborne County, Tennessee—upper Brooks Bridge (RM 95.3), lower Brooks Bridge (RM 94.7), and Oakley Property (RM 89.7).
In the Clinch River, the Bennett Property had the highest number of released mussels >6 months old (28,538), most of which were E. capsaeformis and E. brevidens (see Table 1.18 in Hyde and Jones 2021). The Sycamore Lane (21,417) and Payne Property (15,314) sites in Tazewell County had the next highest numbers of released mussels, followed by Artrip (11,066), Cleveland Islands (7,344), and Whited (1,297).
In the Powell River, most of the mussels released were at the Lower (4,583) and Upper Brooks Bridge (4,211) sites, although an additional 1,301 mussels were released at the Oakley Property. Many of the released mussels at the Powell River sites were either E. capsaeformis or E. brevidens. The number and timing of mussel releases varied widely based upon the species released at restoration sites, how difficult those species were to propagate (e.g., how available gravid females were from year to year), and year-to-year differences in culture success.
Quadrat Sampling
We used a systematic quadrat sampling design at all nine restoration sites (Strayer and Smith 2003). We used multiple random starts for the location of the first quadrat of each systematic sample and all subsequent. We used a quadrat size of 0.25 m2 because it is generally more accurate and precise than 1.0 m2 quadrats when used to estimate abundance (Pooler and Smith 2005). We used three to four random starts at each site. The number of quadrats sampled at each site in 2015 depended on the expected density of mussel species and the desired level of precision. We determined expected densities using 2004–2014 mussel release data from the Freshwater Mollusk Conservation Center (FMCC) and the Aquatic Wildlife Conservation Center (AWCC). A 95% annual survival rate was applied to each cohort to estimate the population density of each species released at each site (Jones et al. 2012). Survival was assumed to be high because mussels are generally long-lived species (Hart et al. 2001; Villella et al. 2004; Hua 2015). Recruitment from released mussels in the wild was assumed to be zero because released mussels were sub-adults. Assuming no recruitment and high survival also ensured that sufficient quadrats were sampled the first year because the density estimate was lower than if we had assumed recruitment, i.e., lower densities require more quadrats. We used a power analysis (Strayer et al. 1997) to determine the number of quadrats needed to achieve a given level of precision:
where n is the number of quadrats, m is the mean number of mussels expected per quadrat, and CV is the desired coefficient of variation (standard error/mean), i.e., a 15% target level of precision was used in this study. We calculated n starting with the most common species at each site and added fewer common species until the number of quadrats became too high (e.g., >400 per site) to reasonably sample. These data were used to determine the target number of quadrats at each site in 2015. For 2016 and 2017, we used actual density estimates from the 2015 quadrat sampling, rather than estimates based on past releases, to determine the target number of quadrats.
Figure 1.
Locations of mussel population restoration and monitoring sites (red dots) in the upper Clinch River, Russell and Tazewell counties, Virginia, for the Certus Inc. NRDAR case and the Powell River, Claiborne County, Tennessee, and Lee County, Virginia, for the Lone Mountain Processing Inc. NRDAR case. Only monitoring data from lower three sites in the Powell were assessed and included in his study. Blue area in inset is the Clinch (a) and Powell (b) river watersheds. Mussel releases occurred from 2003–2019.
Table 2.
Nine sites quantitatively sampled for the Certus Inc. and Lone Mountain Processing Inc. NRDAR mussel-restoration cases in the Clinch and Powell rivers, Tennessee and Virginia. Site length, mean width, and area are all rounded to nearest whole number (area is calculated using unrounded length and width). Sample sizes for quadrat surveys are number of quadrats sampled and sample sizes for mark-recapture surveys are number of individuals sampled. Dash (-) indicates the site was not sampled that year. We used river mile because it corresponds to the unit used in USGS topographic maps.
The distance between quadrats varied among sites and was determined using the formula:
where L is the total length of a site, W is the mean width of a site, n is the target number of quadrats to be sampled, and k is the number of random starts (Strayer and Smith 2003). The distance between quadrats determined the size of the start area where the first quadrat for each systematic sample was placed. For example, a distance of 8 m resulted in an 8 × 8 m start area, and each random start was randomly placed in this box. Random starts at each site were determined using the RAND() function in Microsoft Excel 2015.
The upper and lower boundaries of each site were determined based on the location of past mussel releases and location of suitable habitat. River width was measured at 10-m intervals along the length of each site using a laser rangefinder with 0.5-m precision. Area in each segment was calculated and used to convert population size estimates to densities per m2 (see Hyde and Jones 2021, Appendix C for Google Earth images of sites). We also used these measurements to calculate distance between quadrats using the above formula.
The initial quadrat for each random start was placed, and then all subsequent quadrats were spaced at even intervals along a transect perpendicular to stream flow. The distance between each transect along the stream was the same as the interval between quadrats. Any distance between the last quadrat on a transect and the stream bank was subtracted from the distance between the bank and the first quadrat on the next transect. For example, an interval of 8 m would result in a distance of 8 m between each quadrat within a transect and a distance of 8 m between each transect. If there were 5 m between the last quadrat of one transect and the stream bank, the first quadrat on the next transect would be 3 m from the bank.
Quadrats were excavated to an approximate depth of 20 cm or until bedrock or hardpan was reached. For mussels found in each quadrat, we identified them to species, sexed them as male/female (for dimorphic species), and measured them to the nearest tenth millimeter using dial calipers (length only). We recorded any mussels visible on the surface as “surface,” whereas those not visible were recorded as “subsurface.” For mussels previously tagged at AWCC and FMCC prior to release, we also recorded the tag color and number.
We used the data from the quadrat surveys to estimate abundance of each species by multiplying the mean number of individuals found in a systematic sample by the total number of possible systematic samples in the area surveyed. Density was determined by dividing abundance by the area of the site sampled. We calculated 95% confidence intervals for abundance using the formula
where 
is the estimate of abundance and var() is the estimate of the variance of the abundance estimate (Smith et al. 2001). The variance of the abundance estimate was calculated using the formula
where M is the number of possible systematic samples, m is the number of random starts, x̄ is the mean number of mussels per systematic sample, and xi is the number of mussels in random start (Smith et al. 2001). Variance for density can be calculated by dividing var() by the squared area. We performed the same calculations on the subset of mussels found on the surface of the substrate for comparison to mark-recapture estimates.
Mark-Recapture Sampling
Because Sycamore Lane and Payne Property were in the impact zone of the Certus Inc. chemical spill, we decided to use mark-recapture sampling to independently estimate abundance and density at those two sites. We used a robust design, mark-recapture framework (Pollock 1982) to sample these Clinch River sites during the late summer/early fall from 2015 to 2017. Each year's sampling represented a single primary period under the robust design framework. The population is assumed to be open to changes due to births, deaths, immigration, or emigration between primary periods, i.e., years. Each primary period consisted of two secondary sampling days as close to each other as possible, usually consecutive, when the population is assumed to be closed to changes due to births, deaths, immigration, or emigration. Each site was divided into 20-m wide transects oriented perpendicular to stream flow. Transects were divided into 1-m wide lanes oriented parallel to flow to ensure full spatial coverage of the site. Each lane was sampled visually by snorkeling from the downstream to upstream end. In areas too shallow to snorkel, we used view-scopes or slowly walked through transect areas and visually inspected for mussels. Substrate was not excavated during sampling. Each individual mussel was identified to species, sexed for dimorphic species, and measured for length to the nearest tenth millimeter using dial calipers. We also noted the collector of each mussel. Mussels already tagged had their tag number and tag color recorded. Any untagged mussels were tagged using Hallprint® glue-on shellfish tags and cyanoacrylate glue. After processing, mussels were returned to the location from which they were sampled.
A set of eight candidate models was developed for estimating abundance. These models contained the following parameters:
Si = Apparent survival during primary period i
γ' = probability of not being available for capture during primary period i, given that an individual was not available for capture during primary period
i – 1 (i.e., the probability of not immigrating back into study area)
γ ” = probability of not being available for capture during primary period i, given that an individual was available for capture during period i – 1 (i.e., the probability of temporarily emigrating)
pij = probability of being captured during secondary sampling occasion j of primary period i
cij = probability of being recaptured during secondary sampling occasion j of primary period i
All models assumed that capture probability was constant within a primary period (i.e., across the two secondary surveys) but could vary from one primary period to another {i.e., (p11 = p12) 6 = (p21 = p22)}. Temporary emigration was assumed to be constant and random {i.e., γ'(.) = γ”(.)}.
We created various a priori models as follows: Model 1 was the most general model, allowing both initial capture (p) and recapture (c) probabilities to vary with time between primary periods (interval between primary sampling period) and not equal each other between secondary sampling occasions within each primary period i.e., a behavior response to being captured initially. Model 2 still allowed capture and recapture probabilities to vary with time (interval between primary sampling periods) but they were equal for secondary sampling occasions within each primary period (i.e., no behavior response). Capture and recapture were constant between primary sampling periods in models 3 and 4, but model 3 had no behavior response, whereas model 4 had a behavior response. Survival varied with time between primary periods for all four models. Models 5–8 were equivalent to models 1–4 except that survival was constant.
We analyzed our candidate model set using Program MARK (White and Burnham 1999) to determine the model with the highest likelihood (Villella et al. 2004; Meador et al. 2011). Likelihood estimates were based on Akaike's Information Criterion (AIC) (Akaike 1973) modified for small sample sizes (AICc) (Sugiura 1978):
where L(
) is the likelihood of the parameter estimates, given the data, K is the number of parameters, and n is the sample size. We considered the best model as the one with the lowest AIC score and models were considered competing if ΔAIC <2.0. To estimate the abundance of both the total mussel assemblage and the population of Rainbow Mussel (Cambarunio iris) at the Payne Property, we used the top model in each case.
Table 3.
Mussel species that were assessed for expected abundance and density from 2015–2017 and again in either 2020 or 2021 at sites outside of the impact zone of the Certus Inc. chemical spill in the Clinch River, Virginia, and Powell River, Tennessee. These six species did not occur at restoration and monitoring sites before being released or occurred at very low densities. X indicates species was released at and assessed for expected abundance at that site.
Due to low recapture rates, we could not use the robust design model to estimate abundance at Sycamore Lane, although it was used to estimate abundance for both C. iris and the total mussel assemblage at the Payne Property. Therefore, at the Sycamore Lane site, we used the modified Lincoln-Petersen estimator (also known as the Chapman Estimator) to estimate mussel abundance. The formula used was
where is the estimated abundance, n1 is the number of individuals caught on the first occasion, n2 is the number caught on the second, and m2 is the number of marked individuals caught on the second occasion (Chapman 1951). Standard error was calculated using the formula
from Pollock et al. (1990). Abundance estimates from the mark-recapture estimators (Lincoln-Petersen and the robust design model) were compared to estimates of mussels from quadrats, i.e., the combined surface and subsurface mussels and the surface-only mussels.
Expected vs. Estimated Mussel Abundance
We used a Leslie matrix model developed in collaboration with U.S. Department of the Interior economist Kristin Skrabis to estimate the expected number of total mussels at all nine restoration and monitoring sites in 2017, and in either 2020 or 2021. For the model, we assumed all mussels released at these sites could achieve a maximum age of 40 years, began breeding at 5 years old, and had an annual recruitment rate of 7.6% per year. Annual survival was set as 95% until age class 30, when survival began to decrease annually to a survival rate of 60% to the final age class. Maximum age and breeding age were chosen to represent a typical mussel species. The recruitment rate was set so that the population growth rate would be stable over the long-term. Survival rates were based on Jones et al. (2012), who based their survival rates on an empirical study of dead shells and a catch-curve analysis of shell-length at age and unpublished survival rates from field studies by the Virginia Department of Wildlife Resources, and survival rates reported for other long-lived mussel species (Musick 1999; Akçakaya et al. 2004). We assumed all mussels died after reaching 40 years of age.
We used the mussel release data compiled in Hyde and Jones (2021) as input for the model. Only mussels >6 months old at time of release were included in the analysis. We set mussels at 1 year-old at time of release (i.e., in the 1- and 2-year age-class). We included all mussel species released at the Payne Property and Sycamore Lane sites in the model. At the remaining monitoring sites, we included only those species released at the site that did not occur at those sites prior to restoration (Table 3.). Hence, the natural mussel assemblage at sites in the Clinch River in Russell County, Virginia, and in the Powell River, Tennessee, was not included in our analysis of expected versus estimated mussel abundance. We compared the expected number of mussels at all sites in 2015, 2016, 2017, and either 2020 or 2021 with actual abundance estimates based on quadrat and mark-recapture estimates and calculated the percentage of expected mussels not found during monitoring.
Mussel Length and Growth Rates
The shell growth rate of each tagged mussel sampled more than once was calculated using the following formula:
where Mf is the final measurement and Mi is the initial measurement. When an individual was sampled more than twice, G was calculated for each interval. In cases where the later measurement was less than the first measurement, we set the growth rate to zero rather than negative, as this was likely due to measurement error (since shell length, unlike mass, typically cannot decrease), and included the zero in the calculation of the mean and standard deviation.
Figure 2.
Estimated abundances (a) and densities (b) of freshwater mussels at population restoration and monitoring sites in the Clinch and Powell rivers, Virginia and Tennessee, based on quadrat sampling conducted from 2015–2017, and again in 2020 and 2021. Sites are ordered from upstream to downstream within each river, and error bars represent 95% confidence intervals.
We calculated mean lengths of tagged mussels released in 2013 at the Payne and Sycamore Lane sites for Wavyrayed lampmussel (Lampsilis fasciola), Kidneyshell (Ptychobranchus fasciolaris), and Mountain Creekshell (Leaunio vanuxemensis). Individuals from the 2013 cohort that were sampled from 2015–2017 during our quadrat and mark-recapture sampling were measured and mean lengths calculated for each year. We also calculated mean lengths of C. iris initially tagged during 2015 mark-recapture sampling and tracked the mean lengths of this cohort in 2016 and 2017. Finally, we compared mean lengths of P. fasciolaris among all sites using 95% confidence intervals.
RESULTS
Quadrat Monitoring Data
Across all nine monitoring sites, mussel abundances and densities were generally higher in 2017 compared to 2016, but lower than the first year of monitoring in 2015 (Fig. 2). Sampling in 2020 and 2021 found similar estimates as prior years. In the Clinch River, the Bennett Property had the highest abundances and densities of all sites in all four monitoring years. In the Powell River, Lower Brooks Bridge generally had the highest abundances across all sites and years and the highest densities in 2015 (2.03 m2) and 2016 (1.25/m2), but the highest density observed during the study was at the Oakley Property in 2021 (2.10/m2).
Clinch River.—Total mussel assemblage abundance and density at the Payne Property ranged from 1,257 individuals (0.36/m2) in 2016 to 2,537 individuals (0.72/m2) in 2015 (Tables 4 and 5; Fig. 2). Cambarunio iris was the most abundant species, followed by Lampsilis fasciola and Ptychobranchus fasciolaris. All three of these species were released at the site in relatively high numbers for restoration. Pocketbook (Lampsilis ovata), which also was released in high numbers at the site, was not found during any of the four monitoring years. However, this species has been found at the Payne Property more recently (Tim Lane, Virginia Department of Wildlife Resources, personal observation).
Table 4.
Estimated abundances of freshwater mussels at population restoration and monitoring sites based on quadrat sampling in the Clinch River, Virginia from 2015–2017, and again in either 2020 or 2021. Site at Cleveland Islands was located in the lower-half of the right-descending channel.
Table 5.
Estimated densities of freshwater mussels at population restoration and monitoring sites based on quadrat sampling in the Clinch River, Virginia from 2015 to 2017. Density is reported as individuals per m2. Site at Cleveland Islands was located in the lower-half of the right-descending channel.
Total mussel assemblage abundance and density at Sycamore Lane ranged from 1,590 individuals (0.32/m2) in 2016 to 2,836 individuals (0.67/m2) in 2017 (Tables 4 and 5; Fig. 2). Cambarunio iris was the most common species, followed by P. fasciolaris and Ptychobranchus subtentus. Cambarunio iris was released in the highest numbers during restoration at this site, followed by L. fasciola and Cumberland Moccasinshell (Medionidus conradicus).
Total mussel assemblage abundance and density at the Bennett Property ranged from 22,920 individuals (2.73/m2) in 2016 to 42,360 individuals (5.07/m2) in 2020 (Tables 4 and 5; Fig. 2). The most common species was Pheasantshell (Actinonaias pectorosa), which was not released during restoration and was already present at the site. Epioblasma capsaeformis and Epioblasma brevidens, which were not present at the site before being released there, were the third and fourth most abundant species, respectively. In addition, 1,547 individuals of the federally endangered Snuffbox (Epioblasma triquetra) were released from 2017–2019 (most in 2018). This species was found during monitoring in 2020 with an estimated abundance of 652 individuals (0.8/m2).
Total mussel assemblage abundance and density at Artrip ranged from 4,423 individuals (1.02/m2) in 2016 to 11,359 individuals (2.58/m2) in 2015 (Tables 4 and 5; Fig. 2). The most common species in all years at the site was A. pectorosa. Epioblasma brevidens had the highest number of individuals released at the site, followed by releases of E. capsaeformis, which was detected during all monitoring years.
Total mussel assemblage abundance and density at the Whited Property ranged from 5,789 individuals (1.04/m2) in 2016 to 14,454 individuals (2.59/m2) in 2015 (Tables 4 and 5; Fig. 2). A. pectorosa was the most common species at the site in all years. However, mussel releases at the Whited Property were relatively low, and all occurred before 2013. Epioblasma capsaeformis was the species with the most individuals released but was only detected during the first monitoring year in 2015.
Total mussel assemblage abundance and density at Cleveland Islands in the right descending channel ranged from 2,423 individuals (0.58/m2) in 2016 to 8,529 individuals (1.93/m2) in 2015 (Tables 4 and 5; Fig. 2). Actinonaias pectorosa was the most common species, followed by Eurynia dilatata and Pleuronaia spp. Epioblasma capsaeformis had the highest number of released individuals at the site and was found in all monitoring years. Epioblasma brevidens was released in 2013 (N = 789) but was not detected during 2015 and 2016 monitoring. Epioblasma brevidens was released again in both 2017 and 2018 (>1,000 in both years) and was found in both 2017 and 2020.
Powell River.—Total mussel assemblage abundance and density at Upper Brooks Bridge ranged from 3,232 individuals (0.63/m2) in 2016 to 8,392 individuals (1.67/m2) in 2015 (Tables 6 and 7; Fig. 2). The most common species were A. pectorosa, followed by Mucket (Ortmanniana ligamentina), and M. conradicus. E. capsaeformis and E. brevidens had the most released individuals at the site and were last released in 2013, but only E. capsaeformis was found in 2021.
Lower Brooks Bridge had the highest abundance and density of the Powell River sites across all years. Total mussel assemblage abundance and density ranged from 8,860 individuals (1.24/m2) in 2016 to 14,367 individuals (2.04/m2) in 2015 (Tables 6 and 7; Fig. 2). Actinonaias pectorosa and O. ligamentina were the most dominant species. Epioblasma capsaeformis and E. brevidens were released as late as 2017. Epioblasma capsaeformis was not found in 2021, but E. brevidens still occurred at a density of 0.11 mussels/m2.
Because of its small size, the Oakley site had the lowest abundance of the Powell River sites across all years but had the highest density in 2017 and 2021 and the second-highest density in 2016. Total mussel assemblage abundance and density ranged from 906 individuals (0.80/m2) in 2016 to 2,426 individuals (2.10/m2) in 2021 (Tables 6 and 7; Fig. 2). Epioblasma capsaeformis had the most released individuals at the site, almost all of which occurred in 2012. In 2016, abundances ranged from 32 individuals of Purple Wartyback (Cyclonaias tuberculata), L. fasciola, L. ovata, and P. fasciolaris to 259 individuals of E. capsaeformis, with densities ranging from 0.03–0.23/m2 for these species, respectively. In 2017, abundances ranged from 21 individuals of E. brevidens to 418 individuals of E. capsaeformis, with densities ranging from 0.02 to 0.36/m2, respectively. Overall, E. capsaeformis was the most common species at this site.
Mark-Recapture Monitoring Data
During mark-recapture sampling in 2015, we collected and tagged 105 untagged mussels in the Clinch River at the Payne Property. We also collected 26 mussels that were tagged from previous releases. The total number of observations (including mussels collected on both sampling days) was 137 at the Payne Property (Table 8). In 2016, we collected and tagged 92 untagged mussels at the Payne Property. Including mussels that were already tagged (42) and mussels observed on both days, we had a total of 147 observations. Of these observations, only 11 were recaptures from 2015. In 2017, we collected and tagged 99 untagged mussels at the Payne Property. We also sampled 31 previously tagged mussels and had a total of 141 observations, 18 of which were recaptures from 2015 and 2016. An individual E. capsaeformis collected at the Payne Property in 2015 was likely an inadvertent release from a past study or from hatchery-produced sources and the individual was removed from the site.
During mark-recapture sampling in 2015, we sampled and tagged 101 untagged mussels in the Clinch River at the Sycamore Lane site. We also collected 84 mussels that were already tagged from previous releases/studies for a total of 194 observations at Sycamore Lane (Table 8). During mark-recapture sampling in 2016, we sampled and tagged 184 untagged mussels at Sycamore Lane. Including sampled mussels that were already tagged (213), we had a total of 418 observations at Sycamore Lane. Of these mussels, only 13 were recaptures from 2015. During mark-recapture sampling in 2017, we sampled and tagged 253 untagged mussels at the Sycamore Lane site. Including sampled mussels that were already tagged (331), we had a total of 644 observations at Sycamore Lane, 49 of which were recaptures in 2015 and 2016. An individual E. brevidens collected at the Sycamore Lane site in 2016 was likely an inadvertent release from a past study or from hatchery-produced sources and was removed from the site.
Table 6.
Estimated abundances of freshwater mussels at population restoration and monitoring sites based on quadrat sampling in the Powell River, Tennessee from 2015–2017, and again in 2021.
We could not estimate abundance at Sycamore Lane using the robust design model (Pollock 1982), possibly due to lower recapture rates compared to the Payne Property, especially in 2016 when only 3% of observations were of previously observed mussels. Estimates using the Lincoln-Petersen estimator ranged from 976–1,872 individuals comprising the total mussel assemblage at this site. These estimates were generally higher than the quadrat abundance estimates calculated using only mussels found at the substrate surface during quadrat sampling, but not higher than quadrat estimates using combined surface and subsurface mussels (Fig. 3). For C. iris at Sycamore Lane, Lincoln-Petersen estimates ranged from 357–914 and were generally higher than surface quadrat estimates but lower than combined quadrat estimates (Fig. 3). We were unable to estimate apparent survival at Sycamore Lane for either the total assemblage or C. iris, despite having three years of data.
The top model for the total mussel assemblage at the Payne Property was Model 5, suggesting that detectability varied among years, and recapture rates of individuals marked on the first sampling day of each year were lower the next day (behavior response – likely due to captured mussels burrowing into the substrate after being returned to the stream). Abundance estimates for the total assemblage at the Payne Property ranged from 155–186 individuals and the estimate for apparent survival was 86% (95% CI [5%, 99%]). For C. iris at the Payne property, the top model was Model 8, suggesting that detectability was similar among years and recapture rates were lower on the second day of sampling. Abundance estimates for C. iris at the Payne Property ranged from 113–135 individuals and the estimate for apparent survival was 96% (95% CI [0%, 100%]). The lowest estimates of abundance for C. iris and the total assemblage at the Payne Property were calculated using the robust design model (Fig. 3). For the total assemblage, estimates of abundance based on surface quadrat data and the Lincoln-Petersen estimator of our mark-recapture data were similar, whereas the Lincoln-Petersen estimate of abundance was slightly higher for C. iris. At the Payne Property, estimates of abundance from combined quadrat data (surface and subsurface mussels) were comparatively higher for both the total assemblage and C. iris.
Table 7.
Estimated densities of freshwater mussels at population restoration and monitoring sites based on quadrat sampling in the Powell River, Tennessee from 2015–2017, and again in 2021. Density is reported as individuals per m2.
Expected vs. Estimated Mussel Abundance
Estimated abundance was lower than expected abundance across all years at all sites, although this effect was especially pronounced at sites in the impact zone of the Clinch River (Payne Property and Sycamore Lane) (Fig. 4). Overall, the percentage of expected mussels not found during quadrat monitoring across sites and years ranged from 42.6%–97.6%, with a mean of 75.4% (Table 9). Mean discrepancies were similar for the Clinch and Powell River sites (75.7% and 74.8%, respectively). The Payne Property had the highest mean discrepancy across all years (85.3%), followed by Cleveland Islands RDC (83.8%) and Sycamore Lane (81.8%). The Bennett Property had the lowest discrepancy at 57%.
Mussel Length and Growth Rates
Shell growth rates of mussels sampled during mark-recapture surveys were calculated only for C. iris at the Payne Property and Sycamore Lane sites due to low recapture rates of other species. The mean length of C. iris increased by 1.34 mm, or 3.7%, from 2015–2017, with a mean growth of 0.67 mm (1.85%) per year. No tagged mussels from 2015 were recaptured in 2020.
The mean lengths of L. fasciola, P. fasciolaris, and L. vanuxemensis from the 2013 release cohort all increased substantially from 2013–2015, with a much slower increase from 2015–2017 (Fig. 5). Only two individuals of P. fasciolaris and one individual of L. vanuxemensis from the 2013 cohort were observed in 2020. This 2013 cohort tracks mussels released in 2013 and later sampled during mark-recapture surveys from 2015–2017 (Fig. 5). Growth rates of C. iris from the 2015 mark-recapture cohort were similar to the other three species from 2015–2017 (3.8 mm) (Fig. 5). This cohort represents untagged mussels that were first sampled during mark-recapture surveys in 2015 and later sampled in 2016 and 2017. Based on confidence intervals, mean lengths of P. fasciolaris were significantly lower at the Payne Property and Sycamore Lane sites compared to most other monitored sites (Fig. 6).
Table 8.
Numbers of mussels sampled at two population restoration and monitoring sites in the impact zone for the Certus Inc. NRDAR case in the Clinch River, Tazewell County, Virginia, using transect guided mark-recapture sampling from 2015–2017. An asterisk (*) indicates inadvertent release and individual was removed from site. These values indicate observations during each pass, including observations of the same mussel during both passes.
DISCUSSION
There are several potential causes of lower estimated abundance relative to expected abundance. Estimates of mussel abundance from quadrat surveys were 57% to 85% lower than the expected number of mussels (based on past releases and expected survival and recruitment rates) at all restoration and monitoring sites for both the Certus and LMPI NRDAR cases (Table 9; Fig. 4). First, survival might be lower than we are currently assuming in the Leslie matrix model (e.g., 95% per year). Possibly, the release of propagated individuals into the wild might result in a higher-than-expected mortality. High initial mortality after releases for reintroduction are common for many taxa (Sarrazin and Legendre 2000). At the Payne Property, there were anecdotal reports of Canadian geese possibly feeding on mussels for several days after the mussels were released (Tim Lane, Virginia Department of Wildlife Resources, personal communication). However, estimates of apparent survival from our mark-recapture survey suggest survival is relatively high at the Payne Property (86%–96%), although these data do not cover the timeframe of initial release. Further, freshwater mussels typically have high annual survival rates. A study of naturally occurring Threeridge (Amblema plicata) in the Mississippi and Otter Tail rivers, Minnesota, found that annual survival was greater than 97% in natural habitats (Hart et al. 2001). Meador et al. (2011) found high annual survival of naturally occurring mussels in slack-water and pool habitats (>90%) in the Altamaha River, Georgia, in 2006 and 2007, although mussels in swift-water habitats had somewhat lower survival (75%). Villella et al. (2004) found annual survival was >90% for three species of naturally occurring adult mussels of Eastern Elliptio (Elliptio complanata), Northern Lance (E. fisheriana), and Yellow Lampmussel (Lampsilis cariosa) in the Cacapon River, West Virginia. Carey et al. (2015) found that 65–70% of laboratory-propagated E. capsaeformis released in 2010 and 2011 in the Clinch River at Cleveland Islands survived when the population was sampled in 2011 and 2012. A recovery survival rate of 82% also was observed a year after the release of laboratory-propagated E. brevidens into cages in the Powell River, Tennessee (Hua et al. 2011), although the cages may have contributed to high survival. However, recovery of PIT tagged E. brevidens also found high month-to-month survival (0.98) at this same site over a 2-year period (Hua 2015). Thus, available data suggest lower-than-expected annual survival is not the major contributor to the lower-than-expected abundance found at our sites. Regardless, given the potential for high initial mortality, it would be prudent to bury mussels when released, rather than spreading them on top of the substrate, especially if the presence of predators has been observed.
Figure 3.
A comparison of abundance estimates for the total mussel assemblage and Cambarunio iris at the Sycamore Lane and Payne Property sites, Clinch River, Tazewell County, Virginia, from 2015–2017 based on quadrat and mark-recapture surveys. Surface quadrat abundance was calculated using only mussels found on the surface of the substrate during quadrat surveys. All quadrat abundance includes surface and subsurface mussels. Both the modified Lincoln Petersen estimator and robust design model were used to estimate abundance from data collected from mark-recapture surveys. Error bars represent standard error.
Another possibility is that mussels released at restoration sites are dispersing downstream from the immediate release and monitoring areas. For example, out of 100 mussels relocated in the Kishwaukee River, Illinois, 20 were detected outside of the relocation area over the course of three years, one of which moved approximately 50 m downstream over two months (Tiemann et al. 2016). However, this study only included a buffer zone of 75 m downstream of their immediate sampling area. Other studies have found limited downstream movement. Balfour and Smock (1995) found that the mean net movement downstream of 84 naturally occurring E. complanata (out of 160 initially tagged) in a first-order stream in Virginia over the course of a year was 27 cm, although three mussels (i.e., outliers) moved much further than 27 cm (12.5 m upstream, 25.5 m upstream, and 46.2 m downstream). Another study found the probability of moving downstream among twelve 20-m sections of stream was less than 1% over a period of four years with most movement within 40 m (Villella et al. 2004). Increasing the recapture area compared to the initial sampling area can detect greater movement of mobile organisms such as fish (Albanese et al. 2003). This might also apply to mussels, although the effect would likely be less pronounced. None of the above studies were explicitly examining downstream dispersal. Further, both Balfour and Smock (1995) and Villella et al. (2004) were examining natural populations of mussels. Propagated mussels released into the wild or translocated mussels released at a different site might have higher downstream dispersal than natural populations, possibly due to a failure to burrow sufficiently and thus being more susceptible to high-flow events (Stodola et al. 2017). We also found some evidence of downstream dispersal in our study when qualitatively sampling (visual/snorkel) other potential mussel habitat in the impact zone of the Certus Inc. spill. Two tagged mussels were found at least a kilometer downstream from where they were released at the Sycamore Lane site (one C. iris and one A. pectorosa), and we observed a dead, tagged L. fasciola ∼150 m downstream of its release location at the Payne Property. In 2015, we observed a tagged Flutedshell (Lasmigona costata) in the downstream section of Sycamore Lane, which was released at the Payne Property in 2009, approximately 2.5 km upstream. Several E. capsaeformis and E. brevidens released in the Powell River, Tennessee, in 2012 were observed alive 300 meters downstream of Upper Brooks Bridge in June 2022, and females were observed displaying their mantle lures (Tim Lane, Virginia Department of Wildlife Resources, personal observation). Finally, P. collina have been observed about 2 km downstream from release sites in Rock Island Creek (Brian Watson, Virginia Department of Wildlife Resources, personal communication). Future monitoring should include some form of sampling farther downstream of the immediate monitoring area to account for dispersal.
Figure 4.
Comparison of expected versus estimated abundance of released mussels that did not previously occur at nine monitoring sites in the Clinch and Powell rivers, Virginia and Tennessee (see Table 3 for a list of mussels included). Expected abundance was determined using release data inputted to a Leslie matrix model assuming 95% survival, and estimated abundance was determined from quadrat sampling data. Bars represent discrepancy (in percentage) from expected abundance. Abundance was not estimated for the Whited Property in 2016 and 2017 because the species released at the sites were not detected in those years.
It is also possible that a high proportion of newly transformed juvenile mussels are excysting from host fish outside of the monitoring areas, i.e., fish that were infected with glochidia from mussels released at these restoration sites. For example, C. iris, L. ovata, and L. fasciola use mobile Centrarchids such as Rockbass (Ambloplites rupestris), Large-mouth Bass (Micropterus salmoides), and Smallmouth Bass (Micropterus dolomieu) as hosts and their transformed glochidia wouldn't be expected to settle in the immediate areas where fish hosts were initially infected. Hence, setting recruitment to zero in our Leslie matrix model decreases the expected number of mussels in 2017 at the Payne Property to 10,800 individuals and at Sycamore Lane to 10,996 individuals. However, zero recruitment alone cannot account for the large discrepancies between our expected densities and estimated densities from quadrat samples, especially at these two sites.
Table 9.
Percentage of expected mussels not accounted for in quadrat estimates at each restoration and monitoring site in the Clinch and Powell rivers in Tennessee and Virginia. Percentage unaccounted mussels was likely a function of both emigration and additional mortality. Cleveland Islands was the lower-half of the right-descending channel.
Figure 5.
Mean lengths of selected mussel species at the Payne Property and Sycamore Lane sites in the Clinch River, Virginia, from 2013–2020. Subfigures a–c track cohorts of mussels released in 2013 and their mean lengths when recaptured during monitoring. Subfigure d tracks mussels tagged in 2015 during mark-recapture surveys and their mean lengths when captured during subsequent mark-recapture surveys. Numbers above means represent sample size and error bars show standard deviation. Standard deviation was not calculated for Ptychobranchus fasciolaris in 2020 because both mussels were the same length.
Another possible explanation is that we failed to collect 100% of the individuals present in our sampling units (e.g., sampling lanes or quadrats) at the surveyed sites. For example, Balfour and Smock (1995) found that most mussels <3 years old remained buried in the sediment year-round. Amyot and Downing (1991) found that mussels that were buried in mid-summer tended to be smaller and were likely juveniles. The expected age distribution of mussels at the Payne Property and Sycamore Lane sites in 2017 suggests that 35% of the mussels might be <5 years old. This observation might have caused a negative bias in our mark-recapture estimates, given that we were searching only on the surface. However, if detectability was near 100% in the quadrat survey, buried juveniles should have been detected (and were in our study) and thus would not have affected our abundance. Collector experience can also affect detectability (Wisniewski et al. 2014), suggesting that differences in the experience of collectors may have influenced the survey results of both methods. For example, during quadrat surveys there was a consistent decline in estimated abundance across all sites in 2016 compared to 2015 (Fig. 2). While this decrease may be partly due to a real decrease in abundance, it seems unlikely that such a consistent decrease in estimated abundance would be entirely a result of an actual decrease in abundance, given that our sites were in two different watersheds.
Figure 6.
Mean lengths of Ptychobranchus fasciolaris at restoration and monitoring sites in the Clinch and Powell rivers, Virginia and Tennessee, from 2015–2017 and in 2020 and 2021. Error bars represent 95% confidence intervals. The error bar for Oakley property in 2016 was not calculated because only one mussel was sampled.
Taken together, the reasons for the discrepancy between expected and estimated abundance have substantial implications for future planning of mussel restoration via propagation. Based on Leslie matrix analysis and monitoring, up to 85% of the expected number of mussels (based on number released and expected survival) were unaccounted for. Mussels that emigrate downstream from a release site and are alive should still be credited toward restoration, even if they are no longer at the immediate restoration site, because they satisfy NRDAR's criterion of replacement and/or acquisition of equivalent natural resources. However, mussels that have died because of higher-than-expected mortality should not be credited. Knowing what proportion of this discrepancy is due to higher-than-expected mortality rather than emigration is important for planning, as it would allow for a more realistic estimate of the necessary yearly production to result in the targeted abundance. Future studies should examine dispersal rates of live mussels downstream. Survival also could be studied more thoroughly using PIT tags or estimates from the recovery of dead shells, especially in the period immediately following releases. Thus, future sampling designs should include assessment of areas downstream of the immediate release area to determine site-specific emigration and survival rates.
The large mussel assemblage present in the Clinch River at the Bennett Property is mostly due to one species (A. pectorosa) that was already naturally present at the site and was not released there as part of ongoing propagation efforts. However, E. capsaeformis, E. brevidens, E. triquetra, and several other mussel species listed as endangered were not present at the Bennett Property prior to propagation efforts and they are now among the most common species at this site. The only species released in the Clinch River at the Whited Property was E. capsaeformis, which was not detected in quadrat samples in 2016 or 2017, although a few individuals were collected there in 2015. Similarly, E. brevidens was detected only in 2017 in the right descending channel of Cleveland Islands in the Clinch River, although E. capsaeformis was collected in this channel at a higher density and abundance in 2015 compared to 2016 and 2017. In the Powell River, at the Upper and Lower Brooks Bridge sites, as well as at the Oakley Property, a high proportion of E. capsaeformis and E. brevidens were collected relative to the number of mussels released, suggesting survival was higher than expected at these sites.
Compared to quadrat sampling and the Lincoln-Petersen estimator, a robust design model tended to underestimate abundance. This outcome is likely due to the very low recapture rate, both within and among primary periods, making modeling difficult. For example, of the 39 C. iris sampled in 2015 on the first sampling day at Sycamore Lane, only six were recaptured the next day. Of the 88 C. iris that were sampled on both days in 2015, only 8 were sampled again in 2016, whereas 183 were sampled for the first time that year. Thus, one should expect actual abundance to be much higher than the number sampled and likely higher than the estimates from the robust design model. It is also possible that smaller individuals were buried in the sediment and unavailable for capture during our mark-recapture survey, which also would underestimate abundance.
The Lincoln-Petersen estimator provided a better, lower-bound estimate of abundance compared to the robust design model. It generally yielded higher estimates of abundance relative to abundance estimates made using the surface quadrat data because it accounts for <100% detectability. Both estimates accounted only for mussels found on the substrate surface. However, the mark-recapture surveys were typically conducted during the early fall when detectability at the substrate surface was expected to be higher, whereas the quadrat surveys were conducted in mid-to-late summer, when detectability at the substrate surface was likely lower (Carey et al. 2015).
The mean length of P. fasciolaris was significantly lower at the Payne Property and Sycamore Lane sites than at the other sites. Physicochemical factors, such as habitat, temperature and degree of eutrophication, can affect the growth rates and sizes of freshwater mussels via effects on productivity of the habitat and metabolism of mussels (Bauer 1992). However, the majority of P. fasciolaris released before monitoring in 2015 and 2016 was at the Payne Property and Sycamore Lane sites (741 and 608, respectively). Only 196 mussels were released at the Bennett Property, and none were released at the other 6 restoration and monitoring sites. Many of the mussels released at the Payne Property and Sycamore Lane sites also were released before 2013, and because these sites were in the impact zone of the Certus Inc. chemical spill, there was no population of these species present before releases. The smaller size of the P. fasciolaris populations at these sites is likely because the populations there are much younger than populations at other restoration sites. Further, the mean length of P. fasciolaris, L. fasciola, and L. vanuxemensis in the impact zone sites increased 20–28 mm from 2013–2017. Growth of C. iris from 2015–2017 was only 3.7 mm, but this was not much lower than the 4.8–7.4 mm that the other three species grew during the same period. Since three of the four species were released before 2013, it is likely that the much lower growth from 2015–2017 was a result of mussels reaching an age where overall growth rate begins to slow down.
Growth rates of C. iris at the Payne Property and Sycamore Lane sites were similar to comparably sized C. iris sampled at three sites in the Clinch River from 1988–1993 (Scott 1994). The same study found that growth rates of L. fasciola also were similar to comparably sized L. fasciola at four sites in the Clinch River. Hence, our results suggest that growth has not been negatively affected at the Payne Property and Sycamore Lane sites in the impact zone of the Certus Inc. chemical spill.
Overall, there is evidence of successful restoration for the Certus NRDAR case. Although current abundances in the impact zone have not reached baseline conditions (i.e., 18,621 mussels), we observed released mussels that had grown to breeding ages, displaying females, and recruitment of one species (C. iris) in all three monitoring years (length <20 mm). Because the local mussel assemblage at these sites was completely extirpated, released mussels are clearly breeding successfully. Collectively, our observations indicate moderate restoration success based on the criteria presented in Table 1. Abundance at the two monitoring sites in the immediate impact zone is lower than expected based on assumed survival, but there are populations of numerous species that have low-to-medium densities. Moreover, 37,101 mussels >6 months old, representing 14 species, were released in the impact zone, and another 60,486 mussels representing 20 species have been released at restoration and monitoring sites downstream in the Clinch River in Russell County (see Table 1.18 in Hyde and Jones 2021). Together, this total is far greater than the estimated 18,621 mussels killed during the spill, and therefore restoration at these sites satisfies the NRDAR criteria of recovering or acquiring equivalent natural resources as those injured. Further, the estimated kill was calculated by multiplying dead mussels by three to account for mussels buried in the substrate (U.S. Fish and Wildlife Service 2001). If the spill caused a significant number (e.g., 80%) of mussels to migrate to the surface before dying, then that 3x multiplier overestimated the injury. However, no quantitative sampling was conducted after the spill to validate the use of the 3x multiplier. Future spill assessment studies should include some quantitative sampling, such as excavation of quadrats, to more accurately determine the best multiplier for estimating injury. Although it is unknown why estimated abundance is lower than expected in the impact zone, if released mussels are migrating downstream, then they should still be counted toward restoration for the Certus Inc. NRDAR case. In 2016 and 2017, 731 E. aureola were reintroduced in the Clinch River, 300 of which were released in the impact zone at Sycamore Lane. However, no individuals from this release have been observed alive since 2019 (Sarah Colletti, Virginia Department of Wildlife Resources, personal communication). Further, E. aureola surrogates–E. capsaeformis and E. brevidens–have been well established at other augmentation sites in the Clinch River, Virginia, and they are now the second and third most common species at the Bennett Property, despite not occurring there before restoration.
Restoration success for the LMPI NRDAR case was harder to measure as the impacts to mussels were potentially chronic and sub-lethal. Nevertheless, E. capsaeformis, one of the primary species released in the Powell River, Tennessee, and one which did not occur at restoration sites prior to release, is currently found at low-to-moderate densities at all three sites in the Powell River. These mussels include breeding-age individuals, and both gravidity and evidence of recruitment have been confirmed. Quantification of success for LMPI could have been improved by setting clear, explicit goals to define success.
To document full success, i.e., long-term presence of a stable population, in these and future NRDAR cases involving freshwater mussels, requires long-term monitoring well past the point of final restoration activities, i.e., >20+ years. This recommendation is due to many mussel species having periods of low recruitment punctuated by years with exceptionally high recruitment (Jones et al. 2012). More moderate strength of evidence could be obtained over the medium-term by documenting an increase in the number of recruits in the years (5-10) immediately following restoration activity. Ideally, this increase should be documented long enough after restoration activities that the increase could be attributed to successful breeding of 1st-generation recruits of released mussels, i.e., recruitment is not solely due to released mussels. Regardless, determination of restoration success in future NRDAR cases require both clear, concrete metrics for what constitutes baseline conditions, as well as medium- to long-term monitoring of restored populations.
It may not always be feasible to release enough mussels to reach baseline conditions or to recover the value of their lost ecosystem services. A case such as Certus may result in hundreds of thousands of lost mussel-years and associated services because of the services that would have been provided over the lifetime of the injured mussels and their offspring (Jones et al. 2012). Nonetheless, recruitment from released mussels must be sufficient to eventually reach restoration goals. To determine the appropriate amount of restoration, assumptions about survival and recruitment rates must be made. This study used a Leslie matrix with reasonable age-specific survival rates for long-lived species and sufficient recruitment to maintain a stable population to estimate the expected number of mussels at restoration and monitoring sites for two NRDAR cases. Monitoring of these sites suggested that either survival and/or recruitment of released mussels were lower than expected or that mussels are settling and/or recruiting downstream of the monitoring area. Further study is needed to determine the reasons for this discrepancy and to inform the amount of restoration needed for future NRDAR cases involving freshwater mussels.
Finally, we have developed a set of metrics that can be used to assess whether mussel restoration was successful (Table 1). These metrics range from easy, such as survival of mussels to breeding age, to difficult, such as determining the establishment of a self-sustaining population. They will be useful for designing monitoring programs for future restoration mussel-restoration activities, including determining what metrics are feasible given case-specific time and monetary restraints. Although the Certus Inc. and LMPI NRDAR cases did not assess all of these monitoring metrics, our study represents the largest evaluation to date of restoration of freshwater mussels in a NRDAR context, and we urge future monitoring programs to assess as many of these metrics as reasonable.
ACKNOWLEDGMENTS
Financial support for this project was received from the U.S. Department of the Interior's Office of Restoration and Damage Assessment, Washington, D.C., the U.S. Fish and Wildlife Service and the Virginia Department of Wildlife Resources, with whom we have collaborated extensively on this project. We thank economist Dr. Kristin Skrabis from the U.S. Department of the Interior for her invaluable help with developing the Leslie Matrix used for analysis. We also thank students and technicians at the FMCC, Virginia Tech, who helped with the field and laboratory work for the project, including Aaron Adkins, Anna Dellapenta, John Moore and Andrew Phipps, staff from Virginia Department of Wildlife Resources including Sarah Colletti and Tiffany Leach, and Dr. Catherine Gatenby, U.S. Fish and Wildlife Service. We also thank Dr. Paul Angermeier, U.S. Geological Survey, Blacksburg, Virginia, and several anonymous journal referees, all of whom reviewed and helped improve the quality of the manuscript.
© Freshwater Mollusk Conservation Society 2026
English (US) ·