Using the robust design framework and relative abundance to predict the population size of pallid sturgeon Scaphirhynchus albus in the lower Missouri River

Introduction

The size of animal populations can be difficult to estimate accurately (Manning & Goldberg, 2010) and estimating the population size of rare species is generally more challenging as capture–recapture data are limited. Robust estimates of population size are, however, critical to understanding how management actions affect the populations, especially for threatened and endangered species. Biologists also need population estimates to determine levels of imperilment to species of conservation concern as recovery efforts are planned. In the context of pallid sturgeon Scaphirhynchus albus (Forbes & Richardson 1905) recovery, Braaten et al. (2009) described two critical components aimed at re-establishing a self-sustaining population. First, managers must determine the factors suppressing population stability or recovery. Second, augmentation programmes (i.e. stocking) must be used to increase the population size. Thus, recovery of the S. albus population requires knowledge of fundamental parameters that are needed for population viability models (White & Burnham, 1999; Steffensen et al., 2013a) used to predict and determine the necessary augmentation rates and aid recovery decisions (Bajer & Wildhaber, 2007). The uncertainty of model predictions can be reduced with greater precision of the empirical parameter estimates used in the model.

Species of sturgeon, globally, are considered the most threatened fish species (Birstein, 1993; Pikitch et al., 2005; Lenhardt et al., 2006). Eight species and one subspecies reside in North American waters, of which five are listed as endangered: S. albus, Alabama sturgeon Scaphirhynchus suttkusi Williams & Clemmer 1991, shortnose sturgeon Acipenser brevirostrum LeSueur 1818, white sturgeon Acipenser transmontanus Richardson 1836 and Atlantic sturgeon Acipenser oxyrinchus oxyrinchus Mitchill 1815 (currently listed endangered in four of five distinct population segments). Green sturgeon Acipenser medirostris Ayres 1854 and the Gulf sturgeon sub-species Acipenser oxyrinchus desotoi Mitchill 1815 are listed threatened due to population concerns and shovelnose sturgeon Scaphirhynchus platorynchus (Rafinesque 1820) is listed threatened under the similarity of appearance clause of the Endangered Species Act (Gov., 1973). Scaphirhynchus albus which are sensitive to anthropogenic modifications to large river systems and overharvest, were recognized as declining and subsequently listed as an endangered species in 1990 (55 FR 36 641–36 647; USFWS, 1990). Historic migration and drift pathways have been fragmented by river modifications, which have negatively affected spawning areas and critical food resources and altered the natural temperature, turbidity and flow regime (Keenlyne & Evenson, 1989; Dryer & Sandvol, 1993; Secor et al., 2000; Pegg et al., 2003, Braaten et al., 2008; Braaten et al., 2010).

Scaphirhynchus albus currently inhabit the riverine reaches of the Missouri River and the Mississippi River below the confluence with the Missouri River, but will also select large tributaries (Bailey & Cross, 1954). Similar to other sturgeon species, S. albus are long-lived and mature late with intermittent spawning cycles (Keenlyne et al., 1992; Billard & Lecointre, 2001; Secor et al., 2000; Steffensen et al., 2013b). The factors affecting natural reproduction and recruitment of S. albus in the lower Missouri River [Gavins Point Dam, river km (rkm) 1305·2 downstream to the confluence of the Missouri and Mississippi Rivers (rkm 0·0)] are not completely understood, but are probably affected by the low level of reproductively-ready fish, poor fitness to achieve reproductive readiness and produce healthy gametes, hybridization with the sympatric S. platorynchus, lack of spawning cues or aggregations, insufficient spawning habitats and water quality.

Owing to the lack of natural reproduction and recruitment, an S. albus conservation augmentation programme was implemented throughout most of its current range to aid in population recovery and prevent local extirpation (USFWS, 2008). Since stocking was initiated in the lower Missouri and middle Mississippi Rivers in 1992 (T. W. Huenemann, unpublished data), >175 000 hatchery-reared S. albus have been stocked. Stocking in the middle Mississippi River has not been as intensive compared with the Missouri River and was limited to the 1992 and 1997 year class (T. W. Huenemann, unpublished data). Hatchery-reared S. albus are surviving (Steffensen et al., 2010; Rotella, 2015; Steffensen et al., 2016) and growing (Shuman et al., 2011) and along with a remnant wild population, comprise the current S. albus population.

Despite the federal status of S. albus as endangered for over 25 years, their population size is unknown throughout much of their range. The S. albus population in the upper Missouri River (Fort Peck [rkm 2850] to Lake Sac [rkm 2524]) has been studied more intensively compared with populations throughout the middle and lower reaches of the Missouri River (Klungle & Baxter, 2005; Braaten et al., 2009; K. Kapuscinski, unpubl. data; S. Krentz, unpubl. data). Steffensen et al. (2012) completed an initial population estimate in an 80·5 rkm reach of the lower Missouri River below the Platte River confluence (rkm 957·6) and determined that wild S. albus population estimates varied from 5·4 to 8·9 fish rkm⁻¹, whereas known hatchery-reared S. albus varied from 28·6 to 32·3 fish rkm⁻¹. Winders & Steffensen (2014) estimated an additional 43·3 rkm reach of the lower river downstream of Kansas City, MO, U.S.A. and estimated from 0·6 to 0·9 wild fish rkm⁻¹ and from 5·5 to 10·2 hatchery-reared fish rkm⁻¹. Recovery efforts have continued however, via population augmentation (i.e. stocking) for 6 years after the most recent population estimate (Steffensen et al., 2012). Quantifying the S. albus population in the lower Missouri River appears to be a priority in the draft Missouri River Science and Adaptive Management Plan version 5 (Welker et al., 2016) because mark–recapture data have been used to estimate population size in only 9·5% of the length of the lower Missouri River.

Currently, the S. albus population is being monitored by the pallid sturgeon population assessment programme (PSPAP; T. L. Welker & M. R. Drobish, unpublished data; T. L. Welker, M. R. Drobish & G. A. Williams, unpublished data), which operates under a multi-gear catch per unit effort (CPUE) approach to detect long-term trends in abundance and the population structure. CPUE indices are often criticized because the indices assume capture probability is constant over time and space, which may be unlikely as sampling conditions and species behaviours change. Therefore, the relationship between CPUE and the actual population may not be a simple linear relationship (Pine et al., 2001; Williams et al., 2002) and this relationship must be assessed if a CPUE index is to be used to make important decisions.

The objectives of this study were to: re-evaluate the previously reported population size and annual survival estimates for naturally produced (hereafter, wild) and hatchery-reared S. albus in a small reach of the lower Missouri River; test the relationship between abundance and CPUE indices to develop a predictive model for abundance using CPUE indices; use the relationship, if possible, to predict the S. albus population size in the entirety of the lower Missouri River from CPUE indices. The goal was to use available data from the PSPAP to update demographic population estimates to provide information to support the recovery efforts of S. albus.

Materials and methods

Sampling occurred in an 80·5 rkm reach of the lower Missouri River from the confluence of the Missouri and Platte Rivers (rkm 957·6) to rkm 877·1 (Fig. 1). This reach has been highly engineered through channelization and is characterized by a uniform 200 m wide incised channel. The outside bends are reveted by large limestone rock and have swift flow that precludes sampling efforts, while the inside bends have a series of dyke structures. Sampling was limited to inside bend habitats (see Fig. 2 in Steffensen et al., 2012).

Data for the mark–recapture population estimate were acquired from the Nebraska Game and Parks Commission (NGPC) intensive broodstock collection effort. This annual 11 day effort was directed at collecting reproductively ready S. albus for the PSPAP which began in 2008 with yearly efforts continuing through to 2016. Sampling occurred in early April when crews randomly deployed trotlines daily and covered the 80·5 rkm study reach of the lower Missouri River.

Baited trotlines were used following protocols outlined for the PSPAP (T. L. Welker & M. R. Drobish, unpublished data; T. L. Welker, M. R. Drobish & G. A. Williams, unpublished data). Trotlines were set parallel to the river's current on the inside bend in channel border and pool habitats. Pool habitats are formed immediately below wing dykes where water velocities are static or in eddy currents, whereas channel-border habitat is defined as the area downstream of the pool habitats when flows are laminar. Also, channel border habitats are located between the terrestrial bank interface and thalweg when depths are greater than 1·2 m. Trotlines were fished overnight for a maximum of 2400 hours (USFWS, 2008). The effective sampling area of a static trotline was ambiguous, but hypothesized that a trotline placed between wing dykes would create the effective sampling area of some unknown distance (Steffensen et al., 2012) due to bait attractant. The subsequent approach was based on the concern that the outer portion of the river bend was outside the effective sampling area.

Upon collection, S. albus were thoroughly examined for hatchery marks or tags (i.e. passive integrated transponder tags, coded wire tags, elastomer marks or scute removal) to determine the origin of each individual (i.e. wild or hatchery-reared). In the absences of tags, a genetic sample was collected to confirm origin (Schrey & Heist, 2007; Schrey et al., 2007; DeHaan et al., 2008). Fish were classified wild origin if their genetic results did not match any of the known hatchery-reared parental crosses that historically occurred at hatchery facilities. Fish were then categorized by origin for this analysis.

These mark–recapture analyses followed the methods of Steffensen et al. (2012). The robust design (Kendall et al., 1997) used multiple primary periods (years of sampling: 2008–2016) with multiple secondary periods (days of sampling each year) was used as a traditional sampling approach for this monitoring effort. The robust design analysis was then used to estimate annual population size (N) and the probability of annual survival (Ø), capture (p), recapture (c) and temporary emigration (γ′ and γ′′). The robust design method combines open population (death and movement possible between years) and closed population (no deaths or movement between days within a year) assumptions to estimate critical demographic parameters to allow estimation of population dynamics parameters. Many long-term fisheries monitoring programmes fit the sampling design required for robust design analyses (Steffensen et al. (2012).

This study asked four specific questions relevant to S. albus population recovery to determine if: hatchery and wild fish have different population sizes; annual survival varies by the origin of fish (hatchery or wild); hatchery and wild fish have different emigration patterns; hatchery and wild fish have different capture probabilities. A set of 40 models believed to be biologically reasonable to investigate annual population changes between hatchery and wild groups were used. Akaike's Information Criterion (AIC) was used to compare models (Burnham & Anderson, 2002) and conducted analyses with the robust design module of programme MARK (White & Burnham, 1999; Cooch & White, 2014). AIC_c scores, adjusted for small samples and 95% confidence intervals were used to assess differences between demographic rate estimates for groups. Because annual sample sizes followed the same temporal constraints on parameter variability as in Steffensen et al. (2012); time-constant models for capture rate (p), recapture rate (c), annual survival (Ø) and emigration parameters (γ′ and γ′′) were used in the current models. A time-specific (year) and origin-specific structure for estimates of population size (N) were used in each model however, to meet the study objectives. The data from this rare species did not allow structuring of models with p and c as varying annually. Effort tended to be constant, which supports the assumption of constant capture and recapture probability that is inherent in the analysis. Models in which capture (p) and recapture (c) probabilities were equal or unequal were included to assess whether capture and recapture probability were different.

Dynamics of temporary emigration via two parameters: γ′ and γ′′ were also evaluated. The first, γ′, represents the probability that an individual that is away from the study area remains away from the study area in the next time period (year), given that the individual survives to the next time period. The second, γ′′, represents the probability that an animal within the study area emigrates from the study area in the next time period, given that it survives. Temporary emigration (γ′ and γ′′) was allowed to vary in three specific ways (Kendall et al., 1997; Steffensen et al., 2012) within the set of 40 models: γ′ = γ′′ = 0, no emigration; independent estimates of γ′ and γ′′, Markovian emigration; γ′ = γ′′, random emigration. Models that showed signs of a lack of convergence or those that produced erroneous confidence intervals around the estimate were removed. The methods described in Steffensen et al. (2012) were followed to estimate the super-population size (N_SP; in this study, the 80·5 rkm stretch of the Missouri River) for the study area, dependent on the structure of emigration in the top-ranked models.

Estimates were reported using subscripts and superscripts on parameters. Superscript^origin indicates that the model provided origin-specific estimates for wild and hatchery-reared individuals and subscript _• indicates an estimate that was pooled by origin. The subscript t indicates a parameter allowed to vary by time (year) and a subscript _• indicates the parameter was constant across time (year).

CPUE was calculated as the number of S. albus captured 20-hook-night⁻¹ for each sub-sample, then averaged within each year to calculate the annual CPUE and associated variance. The annual population estimates were then regressed against the annual CPUE to determine the linear relationship between these variables. Annual trotline catch rates were acquired from PSPAP annual reports from the entire lower Missouri River (Herman & Wrasse, 2016; Wrasse, 2016; T. W. Huenemann, unpublished data; T. W. Huenemann & K. D. Steffensen, unpublished data; N. Loecker, J. Kral & S. Stukel, unpublished data; D. Niswonger, T. Huffmon, C. Wilson, & K. Whiteman, unpublished data) and were extrapolated using the linear regression equation to estimate the entire S. albus population in the lower Missouri River.

Results

Nine years of intensive sampling efforts resulted in 3878 trotline deployments or 154 920 hook nights in the local sampling reach of the Missouri River (Fig. 1). Overall, there were 1517 S. albus capture observations from 1223 unique S. albus that included 201 wild S. albus and 1022 known hatchery-reared S. albus. Only 18·3% (n = 224) of the capture observations were recaptured. One fish was recaptured five times, 12 fish were recaptured four times, 43 fish were recaptured three times and 168 fish were recaptured twice. Wild (14·4%) and hatchery-reared (14·7%) fish were recaptured at similar frequencies. The fork lengths (L_F) for wild S. albus varied from 543 to 1180 mm with a mean of 919 mm, compared with hatchery-reared S. albus where L_F ranged from 270 to 1075 mm with a mean of 613 mm. CPUE within the intensively sampled study area (Fig. 1) ranged from 0·08 to 0·22 for hatchery-reared fish 20-hook-night⁻¹ and 0·00 to 0·07 for wild fish 20-hook-night⁻¹. The PSPAP database for the entire lower Missouri River comprised 143 captures during 2015, with a mean reach-specific CPUE of 0·09 fish 20-hook-night⁻¹ (range 0·01–0·17 fish 20-hook-night⁻¹).

The highest-ranking model was structured to provide survival estimates (Ø), temporary emigration probabilities (γ′′ and γ′), capture probabilities (p) and recapture probabilities (c) that were fixed time-constant and origin-specific: Ø(.^origin) γ″(_•^origin) = γ′(_•^origin) p(_•^origin) c(_•^origin) N(_t^origin); W_AICc = 0·47; Table I). Super-population estimates (N_SP) of wild S. albus within the 80·5 rkm study area were highly variable, varying from 319 ± 301–339 C.I. to 584 ± 553–616 C.I. Estimates of N_SP for known hatchery-reared S. albus population were also variable, ranging from 673 ± 600–745 C.I. to 1484 ± 1338–1639 C.I. The population estimates ± s.e. varied from 4·0 ± 0·2 to 7·3 ± 0·3 fish rkm⁻¹ for wild S. albus and 8·4 ± 0·7 to 18·4 ± 1·4 fish rkm⁻¹ for hatchery-reared S. albus (Fig. 2). Excluding 2012 when unusually high water temperature greatly affected catch rates, 2016 was the lowest observed population estimates since these efforts began.

The estimate for annual survival of wild S. albus was 0·92 ± 0·06 S.E. while hatchery-reared survival was much lower (0·63 ± 0·04 S.E.; Table I). The temporary emigration probability was also origin specific and equal (γ′′ = γ′) for the highest ranking model. Wild S. albus had a greater estimated emigration probability (0·88 ± 0·04 S.E.) compared with hatchery-reared fish (0·73 ± 0·05 S.E.). The top-ranked model provided evidence that wild S. albus had a higher capture probability (P = 0·07 ± 0·02 S.E.) than hatchery-reared S. albus (P = 0·05 ± 0·01 S.E.); in contrast, recapture rates (c) were higher for hatchery-reared fish (wild, c = 0·002 ± 0·001 S.E.; hatchery-reared, c = 0·010 ± 0·002 S.E.). The average probability of eventually recapturing a wild fish during the 11 day capture period each year ranged from 4 to 29%, assuming that actual recapture probabilities (c) for wild fish were between their wild initial capture probability (P = 0·07) or the recapture probability of hatchery-reared fish (c = 0·01).

The population estimate within the study area effectively predicted by the annual trotline CPUE (x = 70.726y + 2·533, R² = 0·91, P < 0·001; Fig. 3). The extrapolation of this relationship to the river-wide standardized PSPAP segment specific trotline CPUE was predicted 13 616 ± 7142 S.E. total S. albus in the lower Missouri River (Table II). Under the assumption that the relationship between the estimates of N_SP and CPUE extended to other reaches, the model predicted that the reach from the Platte River (rkm 957·6) to the Kansas River (rkm 591·4) and the reach from Ponca State Park (rkm 1212·6) to the Platte River confluence has the highest local level of S. albus, while the reach immediately below Gavins Point Dam (rkm 1305·1) to Ponca State Park has the lowest local level of individuals (Table II).

Discussion

The robust design is well-suited to provide an estimated population size for this dataset as the robust design accounts for temporary emigration to estimate N_SP (Kendall et al., 1997), which was the main objective of this study. The correlation between the estimated population size and associated catch rates is however, an issue at the crux of every sampling design that relies on a relative abundance index as a means to track population trends (Pine et al., 2012). Yet, this is an issue that often goes overlooked. Studies that have evaluated the relation between relative and absolute abundance estimates have shown no consistent pattern among populations (Emmrich et al., 2012; Hangsleben et al., 2013; Guzzo et al., 2014) further emphasizing the need to specifically understand the potential bias in using such approaches. In the current study, CPUE did correlate well to abundance estimates suggesting that changes in relative abundance in sampling of S. albus on the Missouri River accurately reflect real changes in the population.

This analysis provides an expanded population estimate for S. albus in this 80·5 rkm reach of the lower Missouri River. Steffensen et al. (2012) were the first to estimate N_SP, but their data were limited to 3 years. Here, the current 9 year dataset provided more insight into the trends and effects of the hatchery supplementation programme. The 2012 broodstock sampling season was an abnormal year as water temperatures greatly exceeded the average seasonal water temperatures. These increased water temperatures probably affected the catch rates as other non-sturgeon species removed the trotline bait and reduce the bait availability for sturgeon species (K. D. Steffensen, unpublished data).

Excluding the 2012 sampling year, the mean ± S.E. wild S. albus population was estimated at 5·5 ± 1·2 (range 4·0–7·3) fish rkm⁻¹. The wild S. albus population was variable, but showed no discernible trend in overall population change. The smallest genetically confirmed wild S. albus collected was 543 mm, which indicates that wild S. albus are fully recruited to trotlines and the estimates includes the entire wild population. The hatchery-reared S. albus population was estimated at 12·9 ± 3·3 S.E. (range 8·4–18·4) fish rkm⁻¹. As expected with the hatchery supplementation programme continuing to stock hatchery-reared S. albus, the estimated hatchery-reared S. albus population appeared to increase from 2008 to 2014, followed by a decline during 2015–2016. It is noteworthy, that the population size estimated for hatchery-reared S. albus is limited to the catchable size group of fish greater than 300 mm L_F or approximately age 2 years. Therefore, fish stocked from the previous 2 year classes are probably not fully recruited to trotlines and the estimate is conservative. The continual monitoring and assessment of the hatchery-reared population are encouraged because of the apparent decline in the annual population estimate in this river section.

Survival estimates (Ø) were origin specific for the highest ranking model, which differs compared with the Steffensen et al. (2012) robust model. Estimates for wild S. albus were much higher (Ø = 0·92, S.E. = 0·06) compared with hatchery-reared fish (Ø = 0·63, S.E. = 0·03) without overlap between confidence intervals. Comparatively, the top model from Steffensen et al. (2012) estimate was 0·78, but the model that averaged survival for wild S. albus was only 8% higher than hatchery-reared fish. These estimates also yielded confidence intervals with considerable overlap. Therefore, the authors suspect there is some mechanism specifically affecting the survival of hatchery-reared fish, which is supported by the weight given to the top model with origin-specific survival estimates. The declining trend has also been observed between the Steffensen et al. (2010) and Steffensen et al. (2016) survival specific analyses, especially for age 1 year fish. Several possible hypotheses to these declining survival estimates include age at stocking, rearing hatchery, disease, or fin abnormalities, but none have shown a definitive correlation.

Wild S. albus had a greater estimated probability of emigration (the probability of being outside the sampling area in a given time period) compared with hatchery-reared fish. The wild S. albus population is limited to mainly large, adult-sized fish. These fish are more tolerant of the increased depths and velocities associated with the infrequently sampled mid-to-outer Missouri River channel, which are not sampled due to safety and gear limitation. When wild S. albus were present in the sampling area, they had a higher capture probability (p) than hatchery-reared S. albus. But, recapture probability (c) of wild fish was lower than hatchery-reared individuals. This is because once a wild fish was captured in a given year, it was usually transferred to a hatchery facility for potential use in the propagation programme, thus making the possibility of recapture close to zero. Wild fish are typically returned to the river prior to the following year's surveys; therefore, the survival and annual capture estimates derived from this assessment are not affected by the short-term removal of these fish.

The removal of wild fish during sampling for transport to hatcheries did not bias the estimates of population size because the robust design analysis allowed capture (p) and recapture (c) probabilities to vary. The result showed very low probability of recapture for wild fish (c = 0·002). The CPUE indices did have potential to be biased because of removal, but this suggests the bias is extremely small. Present estimates of eventual recapture of wild fish, in the absence of removal, range from 4 to 29%. Thus, the CPUE estimates for wild fish are only biased by that amount and most likely <10% with the assumption wild fish behave as hatchery-reared fish, which had a 4% decrease in probability of capture after initial capture (hatchery-reared: p = 0·05, c = 0·01). Additionally, the removal of the small number of wild fish is probably absorbed in the variability associated with the annual catch rates.

Scaphirhynchus albus is a rare species that inhabits areas that are inherently difficult to sample and sample sizes in the study were small despite intensive capture efforts. Given that more effort is not logistically possible at present, the use of models with constant capture probabilities represents the best approach to understand dynamics of the species of conservation concerns. The use of constant capture probabilities requires however, the same assumptions as the use of CPUE (Williams et al., 2002). The use of the population estimates for 2012 is not recommended (Fig. 2) as the populations were estimated using the same capture probabilities as other years. Abnormally warm water temperature during 2012 sampling caused capture–recapture rates to be lower, which is probably responsible for the abnormally low estimates of abundance. The authors attempted to modify the best model (Table I) to allow p and c to vary annually, but there were not enough captures during the daily capture periods to estimate c and p. The estimates of population size were biased when estimates of capture probabilities were near zero and this model structure was not further considered.

Thus, the development of a predictive model to estimate population size from associated trotline catch rates provides an additional method for predicting the population size of S. albus throughout the lower Missouri River. Steffensen et al. (2013a) used the population estimate from Steffensen et al. (2012) to extrapolate across the entire lower Missouri River. This method was inherently flawed as it assumed equal distribution, but was the only available dataset at that time. Wu & Halon (2016) used a Bayesian hierarchical multi-population multistate Jolly-Seber model framework to estimate the S. albus population at 5655 (95% C.I. = 4253–7572) using PSPAP capture–recapture data from 2006 through 2010. Similarly, the methods used within this paper predicted the S. albus population at 13616 ± 7142 s.e. in the lower Missouri River. The uncertainty in the predictions are slightly higher than that of Wu & Halon (2016), but the means are similar with the population supplementation programme ongoing for 6 years.

In the lower Missouri River, Winders & Steffensen (2014) estimated the S. albus population for a 43·3 rkm reach just downstream of Kansas City, MO, U.S.A. The current population estimates are five times greater than the Winders & Steffensen (2014) estimate for wild S. albus and approximately double for hatchery-reared S. albus. The similar trend is seen in the PSPAP trotline catch-rate data where the segment 9 prediction is basically double the population prediction for segment 10 (Fig. 1). Overall, the highest S. albus catch rates occur in segments 8 and 9 or in the upper two channelized segments.

The Missouri River recovery programme's science and adaptive management plan will depend on robust estimates of demographic parameters to support recovery efforts, determine recruitment bottlenecks and stocking rates and parameterize population viability models (Fischenich et al., 2016a,b). This approach provides an additional population assessment tool and appears comparable with much more complex Bayesian models (Wu & Halon, 2016) while using the long-term dataset from PSPAP. Monitoring and managing S. albus in a large river system is fundamentally difficult, but the dataset used for this analysis is well-suited for the robust design framework as sampling efforts are an intensive, short duration effort which assumes a closed population (i.e. no immigration, emigration, recruitment or mortality). Integrating traditional relative abundance indices and short-term, intensive sampling periods through a robust design can provide imperative data for population demographics.

We thank the staff members of the Nebraska Game and Parks Commission Missouri River Program and the countless number of volunteers that conducted and participated in our annual broodstock collection efforts. Without their efforts, this analysis would not be possible. Our analyses were supported by Hatch Act funds through the University of Nebraska Agriculture Research Division, Lincoln, Nebraska. Finally, we thank the Nebraska Game and Parks Commission and the U.S. Army Corps of Engineers for funding this effort.