Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 91
6 Integrating Demographic Information with Abundance Estimates Sea­turtle management has been focused on reducing mortality from as many sources as possible on all possible life stages. That is a laud­ able goal for any endangered species, and it is reasonable to assume that minimization of anthropogenic mortality would result in population recovery. Yet, in spite of decades of monitoring and litigation, some U.S. populations (e.g., northwest Atlantic loggerheads [Caretta caretta]) do not appear to be recovering, and the status of most is unknown or inferred exclusively from nesting­beach trends (see Table 1.1). Wildlife and conservation researchers understand that using abun­ dance measures of a single life­history stage can be misleading in diag­ nosing the status and trends of a population (Van Horne, 1983; Thomson et al., 1997; Brooks et al., 2004), including the diagnosis of sea­turtle trends (Bjorndal et al., 1999; Hays, 2000; Chaloupka, 2001b; Solow, 2001; Chaloupka and Limpus, 2002; Heppell et al., 2003). Integrating abundance measures with demographic processes in a framework of modeling and data fitting provides a more robust basis for diagnosing trends, evaluat­ ing the effects of anthropogenic hazards, and defining recovery criteria (Brooks et al., 2008). In this chapter, the committee reviews some of the quantitative tools used in assessment of populations, reviews which tools have been applied to sea­turtle assessments and discusses the procedures that are routinely used in fishery assessments to ensure scientific rigor and could be adopted for future assessments of sea turtles. 

OCR for page 91
 ASSESSMENT OF SEA-TURTLE STATUS AND TRENDS MODELS FOR POPuLATION ASSESSMENT Mathematical models are powerful tools for species assessment and evaluation. The reliability and utility of models depend on the quality and availability of data and on the assumptions conferred by model struc­ ture. Population models for sea turtles have been reviewed by Chaloupka and Musick (1997), Heppell et al. (2002), and others. Published models have ranged from regression fits to nesting­numbers data, deterministic lifecycle analyses, and complex simulation models—all with varied data requirements and assumptions. There are tradeoffs in model construction among precision, realism, and generality. Levins (1966) argued that a par­ ticular model can achieve at most two of those three qualities. Appropri ­ ate model complexity depends heavily on the question asked. The results of a simple model might be robust in uncertainty in lifecycle parameters but qualitative or incapable of supporting the precise estimates of popu ­ lation size or the effects of removal of individuals from a population. In contrast, detailed simulation models may require a large amount of bio­ logical information to produce precise or reliable estimates of population size or to predict response to perturbations. Regardless, models that are to be used for assessment, prediction, and management decisions require solid demographic data, preferably as time series of information that can be analyzed for changes in response to stressors, population density, or environmental variability (Hilborn and Mangel, 1997). TOOLS FOR ASSESSMENT Sea­turtle management issues vary by region, but quantitative assess­ ment generally focuses on the following four primary issues: • Evaluation of trends in nesting and foraging population abundance as an indicator of population status • Diagnosis of the potential causes of those trends • Evaluation of the effects of natural and anthropogenic hazards on population viability • Definition of recovery criteria Here the committee reviews a variety of available modeling approaches to questions about sea­turtle status and trends and notes the data require­ ments for each (Table 6.1). Unlike fishery assessment, the focus for sea­ turtle management in the United States is not on sustainable harvest. Nevertheless, many of the quantitative tools used in fishery assessments are applicable to sea turtles and other threatened species. The results of the approaches identified here vary from qualitative to highly quantita­ tive (Table 6.1).

OCR for page 91
 INTEGRATING DEMOGRAPHIC INFORMATION WITH ABUNDANCE ESTIMATES TREND EvALuATION AND EXTINCTION RISK Trends in Abundance and Abundance Indexes The most common evaluations of sea­turtle population status are those of nesting­beach trends, which may be based on counts of nests or nesting females (see Chapter 4). Linear regression is often used to identify an exponential growth rate for each nesting beach and used with data that is pooled by region (e.g., National Marine Fisheries Service Southeast Fisheries Science Center, 2001). Regression methods have also been used to evaluate trends in abundance indexes derived from juvenile and adult sampling at sea. Slopes and confidence intervals from simple regression analysis are easy to interpret but may fail to include important biological complexities that relate what is counted (such as nests) to a trend at the population level. The numbers of nests or nesting females may be highly variable because of environmental effects on the probability of breeding and other factors (Solow et al., 2002) so data are sometimes smoothed by using a running sum or averaging (e.g., Turtle Expert Working Group, 2007; Snover and Heppell, 2009). Uncertainty in population trends has been evaluated with Bayesian state­space methods that are not restricted to parametric statistical evalua­ tion and permit a more transparent evaluation of the probability of popu­ lation decline (Turtle Expert Working Group, 2007, 2009). In the Bayesian approach, trends are expressed as probabilities of increase or decline rather than as slopes and confidence intervals but still require biological information for extrapolation of nest counts to population abundance. More complex trend­evaluation models that incorporate environmen­ tal drivers, such as nonparametric regression or Bayesian generalized additive models (Bjorndal et al., 1999; Chaloupka, 2001b; Balazs and Chaloupka, 2004b; Troëng and Rankin, 2005), have also been applied. The advantage of the Bayesian approach is that the confidence intervals do not require normal approximation assumptions but are based on the data themselves, and this provides a natural means of evaluating both sam­ pling uncertainty and process error caused by environmental variance. Without estimates of breeding probability (remigration interval) and of recruitment of new turtles to the breeding population, assessment of population trends on the basis of nesting­beach data is highly tenuous. A change in the number of nests may be due to a change in the frequency of nesting, a change in adult­female survival, or a change in the number of first­time breeders, none of which is monitored by the agencies. Esti­ mates of trends in juvenile­turtle abundance through in­water surveys, aerial surveys, and frequency of strandings have generally been evalu ­ ated with regression analysis after an evaluation of data uncertainty (e.g., Turtle Expert Working Group, 2009).

OCR for page 91
 ASSESSMENT OF SEA-TURTLE STATUS AND TRENDS TABLE 6.1 Common Evaluation Methods and Modeling Tools That Have Been Applied to Sea­Turtle Assessment and Their Basic Data Requirements.a Focus Method Accuracy Quantitative Abundance Trend Linear regression of Yes X evaluation abundance index (nests) Bayesian trend evaluation Yes X In­water trends Yes X Diffusion approximation Yes X Trend Surplus production lower Yes X diagnosis Transition matrix Yes Aggregate simulation Yes X Individual­based simulation Yes Integrated models Yes X Ecosystem models higher Yes X Evaluating Bayesian belief network lower No anthropogenic Diffusion approximation Yes X impacts Potential biological removal Yes X Surplus production Yes X Aggregate simulation Yes X Individual­based simulation Yes X Integrated models Yes X Ecosystem models higher Yes X Defining Diffusion approximation lower Yes X recovery Aggregate simulation Yes X criteria Individual­based simulation Yes X Integrated models Yes X Ecosystem models higher Yes X a Methods are grouped according to three primary needs for management and ordered along a general gradient from lower to higher accuracy of model output. Increased accuracy is tied to model complexity and the need for detailed biological information. Stochastic Projections and Diffusion Approximation of Extinction Risk The simplest form of population­viability analysis projects a time series of abundance or an index of abundance and evaluates the prob­ ability of extinction (or recovery) on the basis of the proportion of projec­ tions that cross a predetermined threshold (Dennis et al., 1991; Holmes, 2001, 2004; Snover and Heppell, 2009). The model relies on estimates of the exponential trend and variance estimated from census data and can

OCR for page 91
 INTEGRATING DEMOGRAPHIC INFORMATION WITH ABUNDANCE ESTIMATES Vital Rate Breeding Clutch Adult Juvenile Age at Trophic Frequency Frequency Survival Survival Maturation Dispersal Dynamics X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X be evaluated analytically with a model that describes a diffusion process with drift, commonly referred to as a diffusion approximation of extinc­ tion risk (Dennis et al., 1991). Because time series of sea­turtle abundance are based on counts of nests or nesting females, the trend and variance through time must be adjusted to account for the relationship between nest number and adult­female number (clutch frequency, that is the num­ ber of clutches deposited by an individual turtle in a nesting season) and for autocorrelation (the similarity between observations as a func ­

OCR for page 91
 ASSESSMENT OF SEA-TURTLE STATUS AND TRENDS tion of the time separation between them) caused by remigration inter­ vals (breeding frequency). The diffusion­approximation model has been applied recently to sea­turtle status assessment as a method of estimating trends and evaluating the risk of decline while accounting for uncertainty (susceptibility to quasi­extinction; Snover and Heppell, 2009). It has also been applied to evaluation of removals (Merrick and Haas, 2008; Snover, 2008), although the ability of the analysis to detect changes in extinction risk has yet to be evaluated fully. Trends in an index of abundance and simple stochastic extinction risk can provide benchmarks for status determination, but it is the diagnosis of a trend that is more critical for decision making. Predicting how and why changes in abundance have occurred requires tools that provide additional biological details, particularly the mechanisms of population dynamics that are linked to the sea­turtle lifecycle (Chapter 3). Surplus-Production Models The surplus­production models is the most commonly used population­ assessment approach when one is limited to datasets that consist only of harvest and relative abundance time series (Hilborn and Walters, 1992). Surplus­production models implicitly account for density­dependent demography—the change in population growth rate that is anticipated with changes in population size. The models do not include age struc­ ture but can be modified to include time lags. To determine parameter estimates through data fitting, the models require a time series of abun­ dance data that can accurately demonstrate density­dependent popula­ tion processes. Chaloupka and Balazs (2007) used a Bayesian state­space modeling approach to fit a stochastic surplus­production model to the Hawaiian green turtle (Chelonia mydas) nesting­abundance data series given the known commercial harvest history. This Bayesian­inference approach enabled prior knowledge of green turtle demography to be incorporated to supplement the limited information available on this population. The model accounted for both process and observation error. The approach also enabled uncertainty in model­parameter estimates and the temporal vari­ ability in nesting abundance to be accounted for explicitly. The main objec­ tive was to determine whether it was possible to derive useful estimates of population and management parameters for the Hawaiian green turtle population with the data available. Age- and Stage-Structured Matrix Models These structured models aggregate individuals into life­history stages or ageclasses, allowing incorporation of time lags. They can be deter­

OCR for page 91
 INTEGRATING DEMOGRAPHIC INFORMATION WITH ABUNDANCE ESTIMATES ministic or stochastic (random) and can (but often do not) include non­ linearities, such as density dependence (Caswell, 2001). Analytical sensi ­ tivity analyses of deterministic matrices have been used extensively for sea turtles to identify vital rates that have a large effect on asymptotic­model outputs, such as population growth rate and stage­specific reproductive value (reviewed in Heppell et al., 2003). Most deterministic matrix­model evaluations are useful for learning and discovery purposes to compare relative changes in abundance that may occur with changes in stage­ specific vital rates (e.g., Crowder et al., 1994) or to compare qualitatively the potential effect of removals of turtles of different ages (Wallace et al., 2008). They can be used to predict population size only if vital­rate means and variances have remained relatively constant and if initial conditions of abundance and age structure can be determined. Matrix models that describe lifecycles can be simple or include complex population struc ­ ture, such as the life stages shown in the conceptual model in Chapter 3; values are assigned to parameters on the basis of empirical estimates of survival, growth, or fecundity and estimates of dispersal if life stages are spatially explicit. Matrix models for simulation purposes can include assigning values to parameters through model fitting when time series of abundance, recruitment, or age structure are available (e.g., the model for Kemp’s ridley [Lepidochelys kempii]; Turtle Expert Working Group, 2000; Heppell et al., 2005); however, this has not been done in most of the exist­ ing assessments because of uncertainty in age­ or stage­specific vital rates and unknown population age structure. Diagnosis of observed popula ­ tion change can potentially be performed by using a life­table response experiment if the magnitude of the effects of different vital­rate changes in two or more periods can be evaluated (Caswell, 2001). Age­structured models used in fishery assessment, although not matrix models them ­ selves, operate with the same principles of age­specific tracking through time and recruitment tied to adult abundance. Stochastic Simulation Models A number of stochastic, ageclass­specific, and individual­based simu­ lation models have been developed to account for sea­turtle demography. Chaloupka (2003a) developed a stochastic simulation model for the south­ ern Great Barrier Reef green sea­turtle population to foster better insight into regional metapopulation dynamics. The model (based on a system of ordinary differential equations) was sex­structure and ageclass­structure linked by various density­dependent, correlated, and time­varying demo­ graphic processes that are subject to environmental and demographic stochasticity. The density­dependent processes included depensatory or Allee effects that occur at low abundances when the per capita growth

OCR for page 91
 ASSESSMENT OF SEA-TURTLE STATUS AND TRENDS rate decreases as abundance declines. The simulation model was based on extensive demographic information derived for the population from a long­term sea­turtle research program established and maintained by the Queensland Parks and Wildlife Service. Model validation was based on comparison with empirical­reference behaviors, and sensitivity was eval­ uated by using multifactor perturbation experiments and Monte Carlo simulation within a fractional factorial sampling design. The model was designed to support evaluation of the effects of habitat­specific compet­ ing mortality risks on population abundance and on the sex and age­ class structure. Similar but simpler stochastic simulation models have been developed for the southern Great Barrier Reef green (Chaloupka, 2002a) and loggerhead (Chaloupka, 2003a) sea­turtle populations. The southern Great Barrier Reef green turtle model presented in Chaloupka (2002a) was extended (Chaloupka, 2004) to account for a simple meta­ population structure based on distance­dependent dispersal. Mazaris et al. (2009) developed an individual­based stochastic simulation model that accounted for various density­dependent biological and behavioral attributes (e.g., nest­site selection) of nesting loggerhead sea turtles in the eastern Mediterranean. The model was designed to evaluate the potential effect of nesting habitat loss due to coastal development and sea­level rise on hatchling production and population dynamics. Similar individual­ based stochastic simulation models have been used by Mazaris and col ­ leagues to evaluate various risk factors, such as ageclass­specific mortality on nesting Mediterranean loggerhead population dynamics (Mazaris et al., 2005, 2006). Integrated Population Dynamics Models Stochastic simulation models outlined above are the most compre­ hensive models developed so far to explore the population dynamics of sea turtles and to evaluate the potential effects of exposure to anthro­ pogenic hazards on those populations. It is possible to fit the process­ based models developed, for instance, by Chaloupka (2003b) to a variety of ageclass­specific abundance and demographic data. This modeling approach, comprising integration of various data and model compo­ nents and simultaneous estimation of values for all parameters, presents a number of challenges, including the availability of long time series (Fonnesbeck and Conroy, 2004). Maunder (2003, 2004) presents an inte ­ grated population­modeling framework applied in recent fishery stock assessments that warrants further investigation for sea­turtle population assessments when suitable data series exist. A similar approach was used by Fonnesbeck and Conroy (2004) to model the effects of harvesting on black­duck populations.

OCR for page 91
 INTEGRATING DEMOGRAPHIC INFORMATION WITH ABUNDANCE ESTIMATES Multispecies and Ecosystem Models Sea turtles interact directly and indirectly with other species, and changes in environmental factors have effects on vital rates (see Figure 3.1). There has been an increasing effort to incorporate multispecies and eco­ system interactions in fishery­assessment models (Plagányi, 2007), and any mechanistic model of sea­turtle dynamics has to account for changes in prey, predators, competitors, and habitat. However, comprehensive eco­ system models include a large number of parameters and uncertain inter­ actions; thus, they may prove to be more heuristic than predictive (Fulton et al., 2003). Qualitative approaches, such as loop analysis of community models, can evaluate stability and trophic responses in data­poor sys­ tems (Dambacher et al., 2003). Biomass­balance models, such as EcoPath with Ecosim, require more information on food­web structure and energy transfer but have been applied to a number of ecosystems that include sea turtles (Walters et al., 1997). Comprehensive tools for ecosystem­based fishery assessment, such as Atlantis (Fulton et al., 2005), may have future application to sea­turtle management in well­studied ecosystems. Bayesian Belief or Probability Network Models There are few robust tools available to assist risk assessment and policy development in data­poor and knowledge­vague situations. One approach to support better decision making in data­poor situations is to apply a method known as Bayesian belief networks, also known as probability networks or Bayes nets (Varis and Kuikka, 1999; Castelletti and Soncini­ Sessa, 2007). It provides a structured framework to integrate information from several sources, including simulation models, published material, and stakeholder and expert opinion. Chaloupka (2007) introduced this probability­based approach at a recent workshop of the Food and Agricul­ ture Organization of the United Nations as a robust way to evaluate the relative risk of effects of ageclass­specific anthropogenic hazards—such as fishing gear, coastal development, and climate change—on the long­term viability of Southeast Asian sea­turtle populations. The Bayesian belief net­ work model constructed for the workshop showed (given limited data and uncertainty about turtle­fisheries interactions) that trawl fisheries, gillnet fisheries, and coastal development were hazards most likely to have major effects on the viability of the Southeast Asian sea­turtle populations. Potential Biological Removal Some models are designed specifically to address particular man­ agement questions, such as identification of a threshold bycatch level (see Table 6.1). Potential biological removal (PBR) was developed for

OCR for page 91
00 ASSESSMENT OF SEA-TURTLE STATUS AND TRENDS marine­mammal populations to determine a maximum removal rate that a population can absorb without a large increase in the probability of decline (Barlow et al., 1995; Wade, 1998). PBR is based on the precaution ­ ary approach in a very explicit way. A simple algebraic formula is based on the concept of optimum sustainable yield (Taylor et al., 2000), which is a function of population productivity. PBR determines a maximum human­caused removal of individuals from a population on the basis of half its potential net productivity rate, adjusted by a recovery factor (F) that varies from 0.1 to 1 depending on the status of protection. The equa ­ tion requires a minimum population estimate (Nmin), the maximum rate of increase predicted (or measured) for a population (Rmax), and predeter­ mined risk criteria (low risk to minimal risk) for the recovery factor. PBR is generally applied to an entire population or stock but could be set for specific life stages; the PBR value represents cumulative removals due to all anthropogenic sources. PBR and various modifications to accommo­ date sea­turtle life history have been explored (Bolten et al., 1996; Turtle Expert Working Group, 2000) but not yet used to set bycatch limits or evaluate human­caused mortality. Each of those modeling approaches has merit in potential application to sea­turtle demographic analysis and assessment. However, no model can be useful without data for both setting values of parameters and evaluating model behavior, particularly for applications that require pre­ cision. Increasing model complexity provides biological realism and the ability to estimate population status precisely, but data need to increase also (see Table 6.1). The most biologically realistic and complex models for sea turtles have been developed for populations with long time series of in­water abundance, breeding frequency, survival­rate estimates, and nesting abundance (e.g., Chaloupka, 2003a, b). All of the published sea­ turtle assessment reports (e.g., the Turtle Expert Working Group reports and National Marine Fisheries Service [NMFS] technical memoranda summarized in Table 1.2) have highlighted the paucity of basic data for population modeling, as have reviews of sea­turtle modeling efforts in the United States (e.g., Heppell et al., 2003). The most recent sea­turtle status assessments (National Marine Fisheries Service and U.S. Fish and Wildlife Service, 2007a, b, c, d, e, f) also comment on the need for basic information on population structure and vital rates to identify changes in populations and their listing designations properly. ASSESSMENT PROCEDuRES FOR SCIENTIFIC REvIEW OF DATA AND MODELS In addition to identification of appropriate assessment tools, it is important to have standard procedures for evaluation that ensure rig­

OCR for page 91
0 INTEGRATING DEMOGRAPHIC INFORMATION WITH ABUNDANCE ESTIMATES orous scientific review in all phases of assessment. A thorough review process that covers all elements of a stock assessment is invaluable when it is undertaken by knowledgeable teams of scientists that also include independent experts. It ensures that the “best available science” (National Research Council, 2004; Sullivan et al., 2006) is used to manage our nation’s resources, especially when the process is transparent and open to the public. The need for “best available science” is encoded in legislation directly applicable to sea turtles under the Endangered Species Act of 1973 and in Standard 2 of the Fishery Conservation and Manage ­ ment Act of 1976 (reauthorized in 1996 as the Magnuson­Stevens Fishery Conservation and Management Act). To achieve the use of best scientific data and practice, assessments may include several components that each include peer review. Fishery Assessment The review procedures for stock assessments vary regionally in the United States depending on the fishery management council that is responsible for managing the stock, but they follow a general pattern wherein panels of experts review input data series, models, and refer­ ence points. The review workshops are the Stock Assessment Workshops and the Stock Assessment Review Committees in the northeastern United States; the Southeast Data, Assessment, and Review in the southeast and Gulf of Mexico regions; and the Stock Assessment Review and the West­ ern Pacific Stock Assessment Review in the Pacific region. Typically, the expert panels include one or more members of the management council’s Scientific and Statistics Committee (SSC) and state, federal, and academic scientists but may also include international reviewers from the Center for Independent Experts (CIE). The reviews entail workshops that last up to a week; the workshops result in a series of written reports that are available through the NMFS Web site. CIE reports are prepared separately and are also available to the public online. The CIE reports provide an independent and critical review that are outside agency procedures or oversight and can provide valuable insights. The process of fishery assessment and formula­ tion of management recommendations involves a series of workshops. Data Workshops Participants in data workshops are experts who are responsible for data programs and collections. Some data­review workshops also include CIE representatives who evaluate data quality and the statistical analy ­ ses used in data summaries. During the data workshops, input data are submitted by state agencies and NMFS that include (1) fishery­dependent

OCR for page 91
0 ASSESSMENT OF SEA-TURTLE STATUS AND TRENDS measures, such as catch per unit effort (CPUE), total catch, and age­length matrices to convert total catch to catch at age among others; (2) fishery­ independent measures, such as survey catch abundance and CPUE, and biological metrics; and (3) other ancillary data that might affect abundance or distributional characteristics of the species. Those data are evaluated for consistency and quality. Data that are chosen for analysis are then rec­ ommended for use in the modeling process. Although some data that are typically used in fishery assessments are not available or directly appli­ cable to sea turtles, the approach of comprehensive data review holds value as a potential component of sea­turtle management. It might have value in evaluating surveys, such as nesting­beach counts, strandings and in­water mark–recapture efforts, and length distributions. Model Workshops Participants in model workshops include assessment scientists and demographers, CIE reviewers, SSC members, and other knowledgeable experts. During the model workshops, the adequacy of input data for modeling, model performance, and stability are evaluated. In large part, the evaluations are based on the fits of model outputs to time series of population data, including abundance and age distribution. Results from several different models (e.g., biomass versus age­structured) are often evaluated after the recommendation given by the National Research Council (1998). The model results are reviewed as to whether there is evidence of sustainability of population abundance and excess mortality. Models are also reviewed for retrospective patterns in residua that indi ­ cate poor model fit as parameter values are updated over time. Reference-Point Workshops Participants in the reference­point workshops include experts from state and federal agencies, CIE, SSC, academics, and other knowledgeable experts. Participants evaluate the adequacy of point values that demark the level of overfishing or excess fishing mortality and the level of stock abundance or biomass that results in sustainable populations, which are sufficiently productive of new recruits. These workshops require informa­ tion on population growth and productivity for evaluation of the appro­ priate reference points. Management-Strategy Evaluation Workshops The management­strategy evaluation (MSE) concept was developed in Europe and Australia to provide a simulation approach to evaluate

OCR for page 91
0 INTEGRATING DEMOGRAPHIC INFORMATION WITH ABUNDANCE ESTIMATES management strategies by simulating the effects of different input data, reference points, and modeling frameworks on virtual “populations” (see, for example, Smith et al., 1999). MSE deals directly with uncertainty by simulating the entire process of population dynamics and manage ­ ment from data input to reference points and management response; this is the simulation version of adaptive management. The approach has many advantages because sensitivity analyses can reveal how data quality, assessment­model structure, reference points, and the manage­ ment process itself affect the performance of a given management model. MSE concepts have been introduced into the stock­assessment process by CIE reviewers. Marine-Mammal Assessment Like sea turtles, marine mammals are protected species in the United States that face threats of mortality often caused by direct and indirect interactions with fisheries. Section 117 of the Marine Mammal Protection Act specifies requirements for stock assessments of marine mammals. The act requires formation and support of regional scientific review groups consisting of experts in marine­mammal ecology, population dynamics and modeling, and commercial fishing practices. The groups are respon­ sible for reviewing stock assessments and updates and data and models used to estimate abundance and trends and for advising the agency on uncertainty and research needs. In addition, take reduction teams (TRTs), consisting of scientists and industry representatives, are formed when fishery interactions exceed the allowable take determined through PBR analysis. TRT plans are reviewed according to independent guidelines that have been established for all assessment procedures, including take evaluation, PBR calculation, and review and revision of stock­assessment reports (Wade and Angliss, 1997). Sea-Turtle Assessment A variety of assessments of sea turtles have been conducted by NMFS, all with considerable peer review but not as part of a standardized pro ­ cedure (see Table 1.2). Sea­turtle assessments are conducted as part of a status review required by the Endangered Species Act or in response to a specific management concern. Turtle Expert Working Groups consisting of agency scientists, academics, and scientists associated with stakeholder groups have been formed at irregular intervals since 1995 to review data and conduct analyses related to conservation concerns (Turtle Expert Working Group, 1998, 2000, 2007, 2009). Status reports, required for each species every five years, are conducted by biological review teams that

OCR for page 91
0 ASSESSMENT OF SEA-TURTLE STATUS AND TRENDS are composed of agency scientists in NMFS and the U.S. Fish and Wild­ life Service (USFWS); these are primarily data­update summaries but recently included quantitative analysis (Conant et al., 2009). Recovery teams update the recovery plans for each species, which are split by ocean basin; recovery plans use existing models or published model results to set recovery criteria (e.g., National Marine Fisheries Service and U.S. Fish and Wildlife Service, 2008). Expert workshops to evaluate particular assessment­related issues, such as survey techniques and fishery­impact assessment, have been conducted with assistance from academic scientists and fishery management councils (e.g., Bolten et al., 1996). In addition, the agency conducts and contracts out quantitative evaluations that result in internal agency reports, including take evaluations described in biological opinions. All documents are submitted to extensive internal review and various degrees of external review. Recent quantitative analyses used by Turtle Expert Working Groups, biological review teams, and recovery teams have undergone external review by CIE. Take Evaluation In accordance with Section 7 of the Endangered Species Act, all feder­ ally permitted activities that have potential interactions with sea turtles are evaluated for effect. Estimates of the number and severity of inter­ actions with sea turtles are developed by using observer data or other sources. The resulting biological opinions include the population level impacts of takes, where a take may be direct or indirect killing, injur­ ing, or harassment of individuals or their habitat. Activities may need to be reduced or restricted if they are likely to impede recovery of a listed species or stock. For sea turtles, which are under the joint jurisdiction of NMFS and USFWS, biological opinions are most often written in response to sea­turtle interactions with commercial fisheries or coastal develop­ ment activities. Under Endangered Species Act guidelines, the evaluations must include a determination of whether a proposed activity is likely to cause “jeopardy” to the affected population or species as a whole. Bio ­ logical opinions and jeopardy rulings are critical documents in litigation and are challenged regularly by environmental and industry groups. Standardized, quantitative tools are desirable to determine when a take is sufficient to cause jeopardy and to warrant a curtailment of the fishing or development activity. PBR was developed for marine mammals for a similar application (Taylor et al., 2000). Quantitative evaluation of the effects of bycatch on sea­turtle recov­ ery has been discussed in workshops (Bolten et al. 1996) and modeled in various ways by expert working groups (Turtle Expert Working Group, 2000), agency scientists (National Marine Fisheries Service Southeast Fish­

OCR for page 91
0 INTEGRATING DEMOGRAPHIC INFORMATION WITH ABUNDANCE ESTIMATES eries Science Center, 2001; Snover, 2008), and contractors. In all cases, the authors blamed a lack of basic demographic information for their inability to discriminate among alternative models. The uncertainty in past and present survival, growth, and reproduction rates was too high to make a proper assessment of the likely effect of bycatch at the popula ­ tion level. In one case, 64 alternative population­projection scenarios for loggerheads were presented, ranging in prediction from dramatic decline to rapid recovery (National Marine Fisheries Service Southeast Fisheries Science Center, 2001). A more complex evaluation of expected changes in population growth that might result from reductions in anthropogenic mortality used age­structured models with Monte Carlo sampling of vital­rate distributions to try to cope with uncertainty; the result was a nearly incomprehensible amalgamation of possible population responses (National Marine Fisheries Service Southeast Fisheries Science Center, 2009). Without demography, there is no way to predict the effects of fish­ ery bycatch for such a long­lived animal (Heppell et al., 2003). Threats Evaluation Recent recovery plans have included a semiquantitative evaluation of threats to sea­turtle populations based on rough estimates of the number of turtles affected. To compare the potential population­level effects of threats that affect different life stages of sea turtles, the recovery teams have developed an “adult equivalent” calculation that “discounts” the esti­ mated number of juvenile mortalities according to their reproductive value relative to the reproductive value of adults (National Marine Fisheries Ser­ vice and U.S. Fish and Wildlife Service, 2008; Wallace et al., 2008; Bolten et al., in press). Reproductive value is determined by a deterministic­lifecycle matrix, which requires estimates of survival, growth, and fertility. Uncer­ tainty in remigration interval or other reproductive parameters can have a substantial effect on the adult reproductive value used for scaling, and reproductive values depend on the underlying asymptotic growth rate predicted by the matrix (Caswell, 2001). Thus, methods based on repro­ ductive value and adult equivalents rather than quantitative assessment of threats or setting take limits are best for relative comparisons within spe­ cies that may be used to set priorities for research or conservation effort. Abundance Estimation Estimating population size of sea turtles is highly problematic because they inhabit vast areas and have many ageclasses that occur in different habitats. Extrapolation of nest abundance and trends to adult sea turtles, which probably make up less than 5% of the non­hatchling population

OCR for page 91
0 ASSESSMENT OF SEA-TURTLE STATUS AND TRENDS (Crowder et al., 1994), requires data on sex ratio, recruitment rates (propor­ tion of nesters that are breeding for the first time), and annual survival; uncertainty in these parameters has been incorporated through resampling of known or presumed distributions to provide a range of possible popula­ tion sizes (Turtle Expert Working Group, 2007). Extrapolation of nesting data to estimate population size is even more problematic because of uncertainty in survival and cohort variability. A lack of sufficient infor­ mation on survival rates resulted in a range of a factor of five to ten in estimates of population sizes among best­fit models for Kemp’s ridley sea turtles even though cohort strength (annual hatchling production) was well known on the basis of extensive monitoring of nests for the entire species (Turtle Expert Working Group, 2000; Heppell et al., 2005). Population Trends and Probability of Extinction or Recovery Older sea­turtle assessments relied heavily on simple regression anal­ ysis of nesting­beach data to evaluate population trends, but recent assess­ ments published by NMFS have included Bayesian state­space modeling and diffusion approximation methods to estimate trends and uncertainty in population trajectories (Turtle Expert Working Group, 2007, 2009; Conant et al., 2009). The most recent status assessment of Atlantic logger­ head turtles also includes a “matrix threat analysis” that is essentially a deterministic matrix sensitivity analysis to ascertain potential changes in population growth that result from additional mortality (Conant et al., 2009). The analysis is far more comprehensive than past sensitivity analy ­ ses (e.g., Crowder et al., 1994; Heppell et al., 2003) in that it accounts for uncertainty in estimates of parameter values. The potential cumulative effects of anthropogenic stressors affecting all life stages of each popula­ tion unit are then modeled as additive mortality, and ranges of potential asymptotic growth rates are compared. The exercise is informative inso ­ much as it shows that even under the most optimistic scenarios, there is a high probability that current mortality is too high to be sustained by most loggerhead populations. However, it is a largely heuristic exercise with little or no real power for prediction because of the high level of uncertainty and of assumptions required for deterministic age­structured models. There is no attempt to fit models to data, in part because the time lags involved in sea­turtle life history make it very difficult to establish a likely past or current age structure of the population. CONCLuSIONS Population assessment for management requires an integration of abundance data and demography to account for species’ life history and

OCR for page 91
0 INTEGRATING DEMOGRAPHIC INFORMATION WITH ABUNDANCE ESTIMATES to determine the likely causes of observed trends. There are a number of modeling approaches of varied complexity and precision that can address management questions, but they all need accurate data at the population level. Vital­rate estimation is essential for these slow­growing species, as trends in nesting­beach abundance provide information about only a tiny fraction of a sea­turtle population. Some data that can be used to deter­ mine changes in vital rates already exist, including time series of juvenile abundance (or indexes of abundance) and size distributions. Assessments of managed fish populations include gathering and reviewing biological information and catch data, a variety of modeling workshops to determine the most appropriate tools for assessment and reference points for status determination, and extensive external peer review. Marine­mammal assessments also follow a prescribed path for evaluation. Sea­turtle assessments have included many of the elements required for those species but are not done in a set procedural framework that ensures consistency, transparency, and thorough evaluation. There has been no thorough attempt to assess sea­turtle status with population models that are fitted by using available data on bycatch, size distributions, and productivity. That is because of the following three primary factors that can be addressed by the agency: • Critical vital rates have not been monitored so there is high uncer­ tainty in estimates of parameter values and in interpretation of trends. • Data are scattered and require a thorough evaluation to determine their quality and their applicability to population assessment. • Sea­turtle assessment efforts have not been isolated from broader evaluations of status and threats and have rarely included scientists in other quantitative modeling fields, such as fishery scientists. RECOMMENDATIONS • NMFS and USFWS should develop a general framework for a sea­ turtle assessment procedure, including data evaluation, model review, and MSE. • NMFS and USFWS should conduct data­evaluation workshops, starting with Atlantic loggerheads, focused specifically on the evalua ­ tion of time series information that can contribute to setting values of parameters for demographic models. Data for evaluation include, but are not limited to, nesting abundance, in­water abundance, hatchling­cohort production, length distributions, and reproductive frequency. All sources of data should be evaluated for quality, consistency and spatial or tempo­ ral heterogeneity, and gaps. • Researchers should work with modelers in different fields to

OCR for page 91
0 ASSESSMENT OF SEA-TURTLE STATUS AND TRENDS develop a toolbox for sea­turtle assessment that can provide standardized methods for evaluation and review of data­poor and data­rich species. They would include methods that use available data on trends and size distributions of turtles to reduce the possible ranges of unknown values of parameters and estimates of abundance through model fitting. • The agencies should sponsor a cost­benefit analysis workshop to set priorities among research needs according to which parameters will provide the most useful information for diagnosis of population change.