Conservation efforts under regulatory programs such as the U.S. Endangered Species Act and Magnuson-Stevens Fisheries Conservation and Management Act seek to restore species in decline to a level where the populations are no longer in jeopardy. Key to designing effective conservation plans for the more than 2,500 species currently managed by these two U.S. programs is the establishment of realistic restoration targets given changes in habitats and ecosystems. Accomplishing this is difficult, particularly for populations that have experienced significant habitat alterations. We developed and assessed an approach for establishing quantifiable goals for the recovery of a threatened species by (1) estimating historical population biomass prior to directed fishing, (2) identifying large-scale habitat modifications that limit available critical habitat, (3) assessing the relationship between historical baselines, habitat characteristics, and extant available habitat, and (4) using these relationships to inform recovery efforts in rivers where traditional stock assessment methods are not possible. In our case study of Gulf Sturgeon Acipenser oxyrinchus desotoi, we found that the current population levels in four of the seven river systems in the recovery plan are likely at or exceeding the mean carrying capacity, given the current levels of available habitat. In the remaining three rivers, extant Gulf Sturgeon populations are likely below their estimated carrying capacity levels. Our approach is of management importance because it establishes realistic recovery criteria through the assessment of changes in habitat from historic to present levels and is widely applicable across a species' range. An important application of this approach is to assess the potential benefits to fish species from management actions, such as dam removal or spawning habitat restoration, that are designed to restore habitat and promote species recovery.
The establishment of quantitative population recovery targets as part of recovery plans for threatened, endangered, or exploited species is often required under legislation such as the U.S. Endangered Species Act (ESA), Canada's Species at Risk Act (SARA), or the Magnuson-Stevens Fishery Conservation and Management Act (MFCMA). These mandates share the fundamental goal of rebuilding and recovering species identified by management agencies. In these cases managers and policymakers develop guidelines for rebuilding targeted populations (recovery plans under the ESA and SARA or established fishery management plans under MFCMA). Many of these plans involve research and modeling efforts to assess population status and trends under various life history or harvest policy scenarios (Morris et al. 2002; Walters and Martell 2004; Babcock et al. 2007).
Goals common to the recovery plans developed by the U.S. Fish and Wildlife Service (USFWS) include maintaining stable or nondeclining populations, minimizing take (incidental or directed), protecting critical habitat, establishing subpopulations, maintaining genetic diversity, and recovering minimum population sizes. Some of these criteria are often poorly defined or qualitative in nature. Tear et al. (1993) found that recovery goals were sometimes set below the known population size for the species the plan is written for. Gerber and Hatch (2002) found that since the ESA was amended in 1988 to require objective criteria for delisting that species whose status was considered to be improving more quickly than expected included biologically based recovery goals for that specific species. In the absence of recovery goals it is difficult for management agencies to assess the effectiveness of recovery actions, which is why the use of quantitative metrics is widely advocated in all recovery plans (NRC 1995; Gerber and Hatch 2002; McGowan and Ryan 2009).
A key challenge in establishing realistic recovery goals is determining what the recovered population may look like in terms of abundance or spatial distribution. For many listed species this problem is particularly vexing because habitat loss may be one of the main reasons for endangerment in the first place (Wilcove et al. 1998). Habitat restoration may be a particularly challenging obstacle to overcome because habitat critical for a species may be modified for other uses such as agriculture (Kerr and Deguise 2004), human settlement, or in river systems for hydropower, navigation, or flood control. These habitat modifications are likely not only a contributing factor to population declines, but also a limiting factor in species recovery (Kerr and Deguise 2004).
The Gulf Sturgeon Acipenser oxyrinchus desotoi is a geographically distinct subspecies of the Atlantic Sturgeon A. oxyrinchus and historically is found throughout much of the northern Gulf of Mexico (Figure 1) from Tampa Bay to the Mississippi River drainage (Wooley and Crateau 1985). Gulf Sturgeon were listed as threatened under the ESA in 1991 (USFWS 1995). This species experienced a brief but intense period of commercial fishing during the late 19th and early 20th centuries (Figure 2) similar to that for many other diadromous species (Limburg and Waldman 2009). In the Gulf of Mexico, Gulf Sturgeon fisheries began in Tampa Bay in the late 1800s and spread west through the Florida panhandle, reaching the Suwannee River in 1895, Ochlockonee River in 1898, and the Apalachicola River and several smaller Florida panhandle rivers by 1900. During the first half of the 20th century, the Apalachicola River provided approximately 40% of the total annual commercial Gulf Sturgeon harvest in Florida (Flowers 2008). These fisheries were relatively short lived; the pattern of exploitation by the fishers clearly demonstrated how effective they were in reducing population levels. A similar pattern of high fishery harvests followed by a rapid decline in landings was also observed in other sturgeon fisheries including conspecific Atlantic Sturgeon during this same time period (Spear 2007). By the 1970s and 1980s the Gulf Sturgeon fishery was abandoned due to low catches, which led to regulatory closure of the fishery in Gulf of Mexico drainage states including Florida in 1984 (USFWS 1995).
In this study we developed a method that incorporates habitat-based metrics of juvenile capacity into a stochastic stockreduction analysis (sSRA; Kimura and Tagart 1982; Kimura et al. 1984; Walters et al. 2006) with the goal of estimating realistic recovery targets for an endangered species by (1) estimating a historical population biomass baseline, (2) assessing the relationship between historical baselines and habitat characteristics, and (3) applying these relationships to provide reference points to inform recovery efforts. We assessed the historical population size of Gulf Sturgeon in Florida prior to directed fishing by fitting a model to fisheries-dependent landings information and abundance estimates made from fisheries-independent sampling in the Apalachicola and Suwannee rivers. We then assessed the relationship between river morphometrics and carrying capacity for age-1 fish and used this relationship to predict age-4 carrying capacity in rivers that have sparse data (e.g.. Bogue Chitto-Pearl, Pascagoula, Escambia Bay system, Choctawhatchee). Our overall intent was to highlight an approach to setting recovery targets for species listed by the ESA, using Gulf Sturgeon as a case study. These insights will inform recovery plans for threatened species emphasizing, for instance, the need to set realistic goals and establish target population levels in situations where habitat losses over portions of a species' range have been large.
Life history.—Information on Gulf Sturgeon life history (particularly the location, timing, and characteristics of spawning habitats) has been collected for several Gulf Sturgeon populations (Marchant and Shutters 1996; Sulak and Clugston 1998; Fox et al. 2000). Some life history information such as age at maturity was based on previous Gulf Sturgeon studies (Huff 1975), while other information such as maximum age or growth rates was updated based on estimates derived from long-term capture-recapture studies (Pine and Martell 2009). Growth curves, age-length, and length-weight relationships were developed by Flowers et al. (2009) and Flowers (2008) following standard allometric relationships and von Bertalanffy growth equations. The parameter values and form of relationships used to represent life history traits are described in Table 1, equations (M.1–M.4) and (M.6).
Notation, definitions, and equations of the main parameters and relationships used in the stochastic stock-reduction assessment model.
We developed a prior for adult natural mortality based on life history characteristics including longevity (Hewitt and Hoenig 2005) and individual metabolic demand (von Bertalanffy k parameter: Jensen 1996). Based on these two approaches, we set a lognormal prior distribution on adult mortality that had a mean of 0.095 and an SE of 0.05 (Table 2, equation P.1). Natural survival depended on mean length at age (Table 1, equation M.3) similar to the pattern described in Lorenzen (2000).
Recruitment compensation.—The recruitment compensation ratio (κ;) describes the relative improvement in juvenile survival at low stock abundance compared with the “unfished” stock (Goodyear 1980). Higher κ; values imply that populations have relatively greater increases in juvenile survival at low population sizes relative to unexploited population sizes. In general, populations with higher κ; values are more resilient to exploitation than populations with low κ; values because they have a stronger compensatory response (Walters and Martell 2004). We constructed a weakly informative prior distribution for κ; following the approach in Martell et al. (2008) that used the management parameters of maximum sustained yield (MSY) and the exploitation rate needed to achieve this yield (Pine and Martell 2009). This approach provides probability density estimates for a range of κ; parameter values (Table 2, equation P.2). Estimated κ; for Gulf Sturgeon was 5 (a steepness of ∼0.56), similar to the estimated value for White Sturgeon A. transmontanus (Walters et al. 2006). This value is relatively low across species (Goodwin et al. 2006) but would be expected given sturgeon life history characteristics of long life and late sexual maturation (Pine et al. 2001). For comparison, Myers et al. (1999) meta-analysis for 246 stocks found a mean steepness of 0.78, though anadromous species, such as Atlantic Salmon Salmo salar and Pacific salmon Oncorhynchus spp., had steepness ranging from 0.46 to 0.65, similar to that used in this analysis.
Notation, definitions, and equations used in the objective function for the stochastic stock-reduction analysis.
Catch.—Catch statistics (Figure 2) were collected from the U.S. Bureau of Fisheries, the Report of the U.S. Commissioner of Fisheries from 1896 to 1936, and from the Florida State Board of Conservation from 1938 to 1985 (see Pine and Martell 2009). Recorded statistics were missing from 26 years during the 89 years of harvest. The largest data gap occurred from 1903 to 1917; catches were imputed for this time period assuming an exponential decay rate to match the patterns observed for Atlantic Sturgeon populations (Spear 2007). For shorter gaps of 2–4 years, we averaged the landings for the preceding and following years. Catches were assumed to have originated from the river closest to the county in which they were reported. Countyspecific records were available for 43 out of the 89 years; these were used to allocate harvest removal between the Suwannee and Apalachicola rivers. Linear interpolation was used to impute the proportional allocation for missing years. This backfilling of catches was checked by comparing predicted landings with reports by U.S. Bureau of Fisheries agents. Our interpolated catches were within the reported ranges (Flowers 2008). Uncertainty in the missing landings values can have an impact on the estimated scale parameters. We checked the sensitivity of these parameters by varying the imputed landings values by ±20%.
Abundance data.—Ongoing life history research and population assessment efforts include periodic tagging programs for Gulf Sturgeon populations throughout their range by academic, state, and federal cooperators for over 30 years. Abundance data from a number of mark-recapture studies are available for each of the seven critical habitats assessed (summarized in Appendix A in USFWS 2009).
River morphology.—We assessed seven critical habitat units for Gulf Sturgeon in nine rivers in the southeastern United States (Table 3). We compiled data on historic river discharge and anadromous length (which we defined as the distance available for passage of anadromous fishes below nonpassable barriers to the ocean) from USFWS recovery documents and habitat protection efforts (Table 3). River discharge was measured at the most downstream U.S. Geological Survey (USGS) gauging station on each river having at least 50 years of discharge data (Table 3). When a critical habitat unit was comprised of more than one river, anadromous river lengths were summed. The annual metric used to relate total juvenile production to river habitat was the product of mean annual discharge (cm/s) multiplied by total anadromous length (km). A variety of river metrics have been related to the production of juvenile salmonids such as, substrate type, velocity, and woody debris (Cramer and Ackerman 2009). However, watershed area and stream length can be stronger predictors of juvenile production for these anadromous species. Unfortunately, juvenile Gulf Sturgeon abundance is not measured (unlike many salmonid populations) making it difficult to characterize the relationship between specific river attributes and juvenile production. Our intention with the choice of anadromous length and mean annual discharge as metrics of juvenile production was to select metrics that were available for all rivers in our analysis and would capture finer scale attributes that may allow for more accurate prediction of juvenile production.
The USGS gauge number, geographic location, critical habitat length, and reduced habitat length due to obstructions used to develop the juvenile habitat metrics used in the stochastic stock-reduction analysis. River lengths include tributaries.
Access to potential spawning and rearing habitats is significantly impeded by dams (USFWS 1995, 2009) in the Bogue Chitto in the Pearl River watershed and the Apalachicola River. The fragmentation of rivers by dams is considered a major cause of diadromous fish declines across the Atlantic coast of the United States and a significant impediment for population recovery for many historically abundant species including Atlantic Sturgeon (Brown et al. 2013). We assessed how these barriers to spawning and rearing habitat may decrease the potential carrying capacity in each system. Our assessments can inform recovery criteria for these rivers, for example, the potential benefits from habitat restoration and from mitigation efforts through dam removal or innovative fish passage development.
For the stock-reduction analysis, annual discharge data were available from 1931 and 1941 to the present for the Suwannee and Apalachicola rivers, respectively. For dates prior to the period of record, flow discharge was imputed as the average of the first 5 years on record. The Jim Woodruff Dam on the Apalachicola River was completed in 1957; this reduction in available habitat was included in the stock-reduction analysis.
We developed a Bayesian age-structured stochastic stock reduction analysis (sSRA) to simulate population dynamics from a historical reference point to the beginning of fishery exploitation. This model uses age-structured representations of growth, survival, and recruitment. The sSRA can differ from other age-structured, forward-projecting population models in that it is conditioned on catch; i.e., the number of fish removed from the population by fishing is treated as a known quantity and subtracted from the estimated total population size. Details of the sSRA approach are available in Walters et al. (2006) and the general framework is described below. All model equations are presented in Table 1 and components of the objective function are provided in Table 2. Numbers at age of Gulf Sturgeon were assumed to change over time following discrete time equations (Table 1, equation M.8) and individuals 50 years and older were modeled as a plus group. The age structure was initialized in 1897 and assumed that juvenile capacity had been near the capacity calculated for 1897 (Table 1, equations M.10 and M.11). Vulnerability at age (Table 1, equation M.5) was length dependent and modeled using a logistic function. The vulnerability schedule was used to represent individuals within the populations vulnerable to commercial nets and scientific sampling (referred to as vulnerable numbers) and allowed for direct comparison with mark-recapture population estimates that were obtained using similar gear. Parameter values describing this relationship were not estimated and are determined from examination of sizes captured in commercial landings and fishery-independent sampling (Huff 1975). Annual exploitation (harvest rate) on fully vulnerable individuals (Table 1, equation M.7) was calculated conditional on the observed catch.
We assumed a Beverton-Holt stock-recruitment relationship using the compensation ratio parameterization of the equation (Walters and Martell 2004) and linked juvenile capacity to available habitat area (Table 1, equation M.9). Incorporation of habitat area and change in available habitat area through time into the stock assessment model was done by linking the capacity parameter in the stock-recruitment relationship directly to habitat changes over time (Hy ; Table 1, equation M.14). In this formulation maximum age1 1 at high egg deposition rates is a function of habitat-scaling factor (γ) and the time-varying habitat metric (Hy in this assessment is mean annual discharge times anadromous length). Weak priors were placed on the habitat scaling factors (Table 2, equation P.3). Age-1 recruits are also dependent on relative spawning stock biomass (Table 1, equation M.12), and an estimate of the unfished relative spawning stock biomass per recruit (Table 1, equation M.13). The habitat-scaling factor γ is estimated within the sSRA framework for each river in the analyses. Although we have assumed that all density-dependent mortality is occurring during the first year, there is the possibility that it occurs at other life stages.
We fit the sSRA to the Suwannee and Apalachicola rivers because these rivers have the longest history of removals via harvest and population monitoring efforts since the 1980s. Using equation (M.8) in Table 1 we calculated the number of Gulf Sturgeon in each age-class for every year that accounts for both natural survival and fishery removals with age-1 individuals entering the population modeled using the Beverton-Holt recruitment function described above. The leading parameters in this calculation are the compensation ratio (κ;) and adult natural mortality rate (M), which were shared between rivers. This assumption of a common M is not unrealistic given adult sturgeon are known to use specific locations in marine habitats in winter and are known to move between different river systems (Parauka et al. 2011; Rudd et al. 2014). In addition to the leading parameters, lognormal age-1 recruitment deviations and their associated variance were estimated for each river (Table 2, equation L.2; Table 1, equation M.9). Recruitment deviations can only be estimated for years with observations, but deviations were included for every year to incorporate recruitment variability in the posterior distribution of population numbers over time.
Estimates of leading parameters were obtained by fitting predicted population trajectories to observed mark-recapture estimates of vulnerable numbers assuming lognormal observation error with associated variances for each river (Table 1, equation L.1) in AD Model Builder (ADMB; Fournier et al. 2012). Of the total variability in the observed vulnerable numbers 40% was assumed to be the result of process variation (recruitment variation). The ADMB's Metropolis-Hastings Markov chain Monte Carlo (MCMC) algorithm was used to map marginal posterior probability distributions of the leading parameters. Posterior distributions were assumed to have converged when the Gelman-Rubin scale reduction factor (Gelman et al. 1995) for each leading parameter was <1.1 (two chains each running for 10 million iterations of which the first 1 million iterations were discarded and the remainder thinned at a rate of 1,000).
Application to Other Rivers Included in Recovery Criteria
Currently seven critical habitat units are included in the Gulf Sturgeon recovery plan (USFWS 1995). Five of these units have limited historical information on landings, and current population trends in these systems are uncertain. This situation is typical for rare or listed species; the absence of this information precludes the use of traditional stock-assessment approaches to assess historic population carrying capacity and current stock status. However, we can utilize sSRA output, including estimates of M, κ;, the γ distributions in conjunction with anadromous length, and mean annual discharge from other systems to inform estimates of population size and carrying capacity. If we assume that the values for these rivers with sparse data are within the distribution of estimated parameters for the Apalachicola and Suwannee rivers (Figure 3), we can then estimate juvenile carrying capacity based on known riverspecific habitat metrics (mean annual discharge × anadromous length) for all river systems. Vulnerable adult population size given the estimated juvenile carrying capacity in each river can be estimated using equilibrium analysis (equilibrium solutions to equations M.8 and M.9 in Table 1). This estimation is done within a Monte Carlo framework where we randomly draw from the posterior distributions of k and M and a distribution for y. Capacity estimates for the Suwannee and Apalachicola rivers, then, are based on river-specific y distributions, while estimates for other rivers are based on a normal distribution with a mean and SD equal to the mean and SD of the combined posterior distributions calculation for the Suwannee and Apalachicola rivers. Posterior draws (n = 100,000) are combined with randomly selected mean annual discharge for each river over the last 20 years as well as the river's anadromous length to estimate juvenile carrying capacity. Juvenile carrying capacity was estimated with and without migration barriers in the Apalachicola and Pearl rivers. Age-1 capacity estimates were then used to estimate equilibrium vulnerable numbers (approximately ages 4+). These estimates were compared with the best available estimates of vulnerable populations from ongoing mark-recapture programs (USFWS 2009).
Stochastic Stock-Reduction Analysis: Model Fit to Time Series of Historical Landings, Habitat Changes, and Current Population Size
The inclusion of the habitat scaling metric 7 to account for changes in habitat availability, particularly within the Apalachicola River following completion of the Jim Woodruff Dam, allowed us to capture the different recovery trajectories of these populations (Figure 2) and helped us to develop an understanding of the population’s carrying capacity in each river under present habitat availability. For both the Suwannee and Apalachicola Gulf Sturgeon populations, historical removals caused a rapid decline in model populations beginning in the late 1800s. Within the Suwannee River, the Gulf Sturgeon population model biomass and vulnerable numbers declined during the early decades of the commercial fishery to below one-half of the unfished abundance of about 7,000 individuals, and peak exploitation rates of less than 30% occurred in 1900 (Figure 2). This model population slowly recovered in the first half of the 20th century (Figure 2) and then declined again as landings increased in the 1950s–1970s before the fishery closed in the early 1980s. The Apalachicola River Gulf Sturgeon population is estimated to have experienced a much larger decline in population vulnerable numbers of more than 90% from around 18,000 prior to 1900 to fewer than 100 individuals by the 1940s–1950s. Unlike the Suwannee River where landings continued throughout the 1960s and 1970s (Huff 1975), high removals during this time period did not occur in the Apalachicola River, allowing for limited population recovery. However, around the time of the completion of the Jim Woodruff Dam on the Apalachicola River in 1957 there was an increase in reported Gulf Sturgeon catch, which may have resulted in exploitation rates of nearly 100% of vulnerable individuals (Figure 2). Current estimates of vulnerable population size from our sSRA model explicitly included habitat changes, (e.g., completion of the Jim Woodruff Dam), which explain why our current abundance estimates for the Apalachicola River are extremely low compared with historical levels (Figure 2). These results also suggest that Gulf Sturgeon population recovery in the Apalachicola River much beyond these low levels is likely strongly curtailed by the reduction in habitat area following the completion of the Jim Woodruff Dam.
Parameter estimates.—Posterior probability distributions of κ;, M, and river-specific γ values are shown in Figure 3. There is a notable contraction in the posterior distribution of the κ; from the prior distribution and a decrease in the mode from 5 to 2.6, suggesting that recruitment compensation is weaker than that estimated using the approach in Martell et al. (2008). The posterior distribution for adult natural mortality rate contracts slightly with a slight increase in the mode. Posterior distributions for the habitat scalar 7 are well defined for the Suwannee and Apalachicola rivers. The habitat scalar for the Suwannee River was estimated to be greater than that for the Apalachicola River, suggesting that, per unit of habitat metric used, the Suwannee River has a greater capacity for juvenile production than the Apalachicola River. As anticipated, there was a strong positive correlation between natural mortality and 7, particularly for the Apalachicola River, with a correlation of 0.99 (Figure 3).
Habitat-based carrying capacity estimates.—We estimated the 90% credible interval for habitat-based carrying capacity for the Suwannee River to be about 2,359–9,555, which is similar to current estimates of population size suggesting that this population is approaching the preexploitation levels (Figure 2). Our Apalachicola River results suggest that the Apalachicola River before the dam would likely have supported an estimated 7,447–23,271 Gulf Sturgeon of age 4 and older, while postdam estimates of carrying capacity are approximately 1,923–6,066. Current abundance estimates in the Apalachicola River range from only a few hundred to a few thousand vulnerable individuals (Figure 2; Table 4). Carrying capacity estimates were sensitive to the imputed landings. A 20% change in the imputed values resulted in a 10% change in the Suwannee River capacity estimate and a 15% change in the Apalachicola River capacity estimate. Our results provide key information to resource managers in updating recovery targets for these populations that are currently not available, even with the uncertainty in the landing data and the associated sensitivity to the carrying capacity estimates.
We used the distribution of 7 values from the Apalachicola and Suwannee rivers to estimate population carrying capacity for the five other rivers considered as part of the Gulf Sturgeon recovery plan: Pearl, Pascagoula, Escambia, Yellow, and Choctawhatchee rivers (Table 4). Our results for these rivers demonstrate that carrying capacity is highly dependent on estimated available juvenile habitat and river discharge. Our 90% credible intervals of carrying capacity estimates for vulnerable numbers range from about 185–811 in the Yellow River to 4,022–18,077 in the Pascagoula River (Figure 4; Table 4). Much like the Apalachicola River, results for the Pearl River, where access to historic habitat is blocked by a sill on the Bogue Chitto River, suggest capacity estimates of 6,336– 28,856 without any migration barriers but only about 10% of this level (686–3,076) under current habitat availability.
Current status relative to capacity.—To assess their recovery potential we compared the ratio of current mark-recapture-based estimates of vulnerable population numbers to model-based capacity estimates of vulnerable Gulf Sturgeon abundance for each river (Table 4). We found that for the Yellow, Choctawhatchee, and Suwannee rivers, current population size is similar to the estimated posterior distribution for carrying capacity, while the Pascagoula, Apalachicola, Escambia, and Pearl rivers are likely <30% of their estimated carrying capacity (Figure 5). For the Apalachicola and Pearl rivers we calculated these ratios both in terms of their current available juvenile habitat with the existing impediments to migration (noted with “S” for short in Figure 5) as well as in the absence of migratory barriers (noted with “L” for long in Figure 5).
Median as well as the lower (5%) and upper (95%) limits of the 90% credible interval values for estimated vulnerable numbers of Gulf Sturgeon produced at juvenile production capacity for sturgeon rivers flowing into the eastern Gulf of Mexico. Carrying capacity estimates for the Pearl and Apalachicola rivers are shown with (S) and without (L) migration barriers. Recent estimates of vulnerable population size were taken from the 2009 Gulf Sturgeon 5-year review document.
Using the Gulf Sturgeon as a case study, we developed an approach to informing restoration programs for listed species by estimating the carrying capacity for species where habitats have been highly altered. This approach is particularly useful for situations where portions of the species' range are data poor, such as the absence of long-term population trends in the Pearl River in this example. Yet conservation decisions such as removing dams to increase habitat availability or directing research efforts to improve abundance estimates and trends must still be made. We recognize that our estimates of riverspecific carrying capacities are highly reliant on the prior distributions of leading parameters used, as well as the habitat metrics chosen. We also recognize the key assumption that these metrics are related across rivers. These carrying capacity estimates could be improved by refining our understanding of Gulf Sturgeon population ecology in any one river, particularly through direct assessment of juvenile production annually. However, given limited funding is available for such highly detailed, long-term research programs, this information is unlikely to be developed for Gulf Sturgeon in the near term. We take an approach where key parameters, such as natural mortality, are estimated for rivers where data are relatively rich and share these estimates with data-sparse rivers. There are three key uncertainties in our approach. (1) Our estimates of landings are likely underreported such that the actual removals were greater than what we have included. In this case our estimates of carrying capacity are likely underestimated. (2) Our estimates of natural mortality may not reflect the natural mortality in any one river system. For example, rivers in the western Gulf of Mexico have experienced hurricanes and severe flooding in recent years, which may have led to higher local natural mortality. This uncertainty was previously recognized (Pine and Martell 2009) and research efforts are currently underway to assess natural mortality across the Gulf Sturgeon's range. (3) Direct estimates of κ;, the Goodyear recruitment compensation ratio, are difficult due to the required time and expense and because of the strong correlation with M. If κ; or M were substantially higher than the values used, our estimated carrying capacities would be much lower. However, if the κ; value was much higher, population recovery would have been much more rapid than has been observed. The framework presented uses the best available information to set reasonable bounds on recovery goals. As new information becomes available, these benchmarks can be easily updated.
Large-scale habitat modification coupled with intensive harvest pressures altered much of the Gulf of Mexico's natural resource landscape during the 19th and 20th centuries. As such, most conservation efforts are in part motivated by overutilization of resources in the past. Many conservation actions from both governmental and nongovernmental parties are, in some respects, designed to “turn back the clock” and preserve or restore portions of the landscape to what is perceived to be pre-European settlement levels. These types of retrospective planning efforts (using historical baselines: Sanderson 2006) are difficult given the uncertainties of what these landscapes actually looked like prior to settlement. Recent work with Atlantic Sturgeon (Secor 2002), Atlantic Cod Gadus morhua, green turtles Chelonia mydas, and dugongs Dugong dugong (Jackson et al. 2001) all used retrospective analysis (statistical modeling and/or zooarcheological excavations) to reconstruct historical abundances of key coastal species. Results provided conservation targets and insight into the magnitude of anthropogenic alteration of coastal ecosystems. Conservation targets with quantifiable goals for a species such as Gulf Sturgeon are critical to ESA and MFCMA programs since they provide metrics to assess population responses to conservation and restoration actions.
Lessons for the Gulf Sturgeon Recovery Plan
Our results for Gulf Sturgeon suggest that population biomass for this species was severely reduced during a short, but intense, period of commercial harvest in the eastern Gulf of Mexico. After large-scale fishing was abandoned, a small fishery persisted, primarily on the Apalachicola and Suwannee rivers, until regulation closed all sturgeon fishing in the early 1980s. Gulf Sturgeon populations in the rivers that likely supported the largest fisheries (Apalachicola and Suwannee rivers) have demonstrated stable or increasing populations during the last 20 years (Zehfuss et al. 1999; Pine et al. 2001; Sulak and Randall 2002), but recent estimates of abundance may be uncertain (Pine and Martell 2009). Threats to Gulf Sturgeon populations and riverine and estuarine habitats are increasing from a wide variety of anthropogenic sources that include alterations to surface and subsurface waters (e.g., dams and water withdrawals: Seavey et al. 2011), expanding risks from increased oil and gas extraction activities (e.g.. Deepwater Horizon oil spill: USFWS 2010, http://tinyurl.com/cvbcmpk), and in-river industrial accidents (e.g., recent Temple-Inland pulp mill spill in the Pearl River), all of which could impede or reverse population recovery.
A key basic need in many population recovery and fishery management plans is to determine measurable, objective, population target levels to answer the question of how many individuals are needed for long-term viability of a population or species (Tear et al. 2005; Sanderson 2006; Wilhere 2008). Our results provide guidelines in using historical data to estimate current potential carrying capacities for use as recovery reference points. As an example, currently the Suwannee River population of Gulf Sturgeon is likely at more than 80% of its estimated carrying capacity or very near carrying capacity. In contrast, the Gulf Sturgeon populations in the Apalachicola and Pearl rivers in which critical habitat for sturgeon is reduced primarily by the Jim Woodruff Dam are estimated to potentially support only 50% of the age-4-+ individuals that would be supported if all of the identified critical habitat were available. Thus, for both the Pearl and Apalachicola rivers, recovery targets near the historical abundances are likely dependent on whether access to historic spawning and rearing sites are available. The need to develop measurable, quantifiable goals for recovery criteria have long been recognized (NRC 1995); yet, for many populations this information is not available because of sparse data. For Gulf Sturgeon we have shown how key population parameters can be estimated for a portion of the range and then used to inform decisions for areas where data are sparse. In this way, critical decisions, such as whether to define recovery goals with or without habitat change, or the potential benefits to a species from management actions, such as dam removal, can be better informed and thus lead to more effective conservation actions and an increased likelihood of the species' recovery.
We thank J. Flowers, C. Walters, and S. Martell for assistance with data compilation and with early versions of the SRA model. We thank O. Jensen and S. Stapleton for reviewing early versions of this paper. We thank the NOAA Fisheries Office of Protected Resources, U.S. Fish and Wildlife Service, and the Florida Fish and Wildlife Conservation Commission for research and funding support of earlier projects that led to the development of this paper.