Population Viability Analysis (PVA) is a class of quantitative tools for assessing status, projecting population growth, evaluating threats, and exploring conservation and management options for wildlife populations (Beissinger and McCullough, 2002; Morris and Doak, 2002; Lacy, 2019). PVA usually involves simulation models, in order to consider the many biological, environmental, and human factors and processes that can impact a wildlife population, in a framework that includes consideration of the uncertainties in our knowledge of the system (McGowan et al., 2011), uncertainties about future conditions, and inherent unpredictability of many stochastic biological processes (Lacy, 2000a).

The first applications of PVA were developed to estimate “minimum viable population” size (MVP), the size below which instability of local environments, stochastic demographic processes, disrupted population structure (e.g., Allee effects), genetic decay, and feedbacks among these threats would create a significant threat of population collapse and extinction even if the factors that originally drove the population to small size were ameliorated (Shaffer, 1981; Gilpin and Soulé, 1986). Subsequently, PVA has become a standard tool in conservation and wildlife management (Beissinger, 2002), with the focus on assessing threats and evaluating conservation options. However, applications of PVA are often based on inadequate data to provide confidence in the model projections (Ludwig, 1996; Ralls et al., 2002), and a common recommendation is for PVAs to be used for comparative analyses of threats and options, rather than predictions of population trajectories far into a mostly unknowable future (Reed et al., 2002; McCarthy et al., 2003). Thorough PVAs should include a number of components (Ralls et al., 2002; Lacy, 2019), including:

Population Viability Analysis is most commonly applied to threatened taxa, for which there are concerns about extinction risk, but it is also important to evaluate local populations of wildlife species not currently threatened with extinction. It is often difficult to obtain adequate data on highly threatened species to permit accurate projections with PVA, but data on related, more common species can be used, with appropriate caution, to provide plausible estimates in cases where data gaps hinder analysis of the threatened species (Cervin et al., 2020). Moreover, loss or decline of local populations of even widespread species can damage community and ecosystem structure and function (Ebenman et al., 2004; Gaston and Fuller, 2008). Attention to such processes will be increasingly important as the global environment changes at an accelerating rate (Scheffers et al., 2016). For any population, PVA can serve as a framework for compiling, synthesizing, and documenting what we know and what we do not know about the demography of the population.

The Sarasota Bay common bottlenose dolphin (Tursiops truncatus) community has been studied more intensely and continuously, for more years, than almost any other wildlife population, with research initiated in 1970 (Wells, 2020). These studies provide a more complete picture of the demography of this population than is available for most other species that have been examined with PVA. A few other notable examples of species for which there have been long-term demographic studies that enabled detailed PVA include chimpanzees (Pan troglodytes: Pusey et al., 2007; JGI et al., 2011), red wolves (Canis rufus: Simonis et al., 2018), whooping cranes (Grus americana, Traylor-Holzer, 2019), and the Shark Bay, Australia, population of bottlenose dolphins (T. aduncus: Manlik et al., 2016). Such long-term studies make it possible to build a robust PVA model of the population and validate the model projections against historical trends. With the PVA model, we can examine the current status, project population growth rates, assess tolerance of, or resistance to, threats, test resilience to catastrophic events, quantify sustainable removals or habitat degradation, evaluate management options, and identify needs for research into uncertain parameters that are primary drivers of the populations’ fates. Beyond the significance for the specific population, a PVA of the Sarasota Bay dolphins can provide a template (or baseline for comparison) for examining viability of other populations of small cetaceans.

Long-term monitoring of the Sarasota Bay dolphin community provides demographic information at a level of detail and precision rare in studies of cetaceans or any wild populations. Thus, the description of population demography for this population can serve as a standard for PVAs on small cetaceans. A PVA of the Sarasota Bay population of bottlenose dolphins was therefore undertaken with multiple purposes:

The Vortex PVA software has been used to assess population status, population growth and stability, relative threats, and management options for 100s of species⁴, including a number of marine mammals (e.g., dugong: Heinsohn et al., 2004; bottlenose dolphins: Thompson et al., 2000; Vermeulen and Bräger, 2015; Manlik et al., 2016; Blázquez et al., 2020; offshore dolphin species: Ashe et al., 2021; beluga whale: Williams et al., 2017; killer whale: Lacy et al., 2017; Murray et al., 2021; and Baltic harbor porpoise: Cervin et al., 2020). Vortex is an individual-based simulation that allows modeling of details of individual variation (e.g., variable ages of maturation and senescence), stochastic processes affecting small populations (e.g., demographic stochasticity, age structure perturbations, disrupted social structure, and local inbreeding), as well as the primary drivers of population trends of mean demographic rates. While not all of the options were used in this study, and factors should be omitted from population models when they are not of concern or thought to be consequential in the focal population, the options might be important for PVAs on other cetacean species. If all sources of stochasticity, dependencies on external variables, and non-linear interactions are removed from the model, the population projections generated by Vortex match those of standard matrix models of population growth (Caswell, 2001). The stochasticity in simulation models (and in real biological populations) will usually depress mean long-term population growth below that predicted from simpler matrix models (Lacy, 2000a).

In the past 27 years of consistent monitoring, the Sarasota Bay community of bottlenose dolphins has been growing at a mean rate of about 2% annually, but with considerable fluctuations across years (mean r = 0.019, SD(r) = 0.061, SE = 0.012, range r = −0.086 to 0.188). Consequently, even with 27 years of data, the precision with which we can predict the long-term mean population growth rate from the observed population trajectory is still limited, with a 95% CI (−0.006 to 0.044) that includes no population growth. This suggests that even for a long-lived mammal, with census data covering more years than is typically available for wildlife populations, fluctuations in population size make it difficult to estimate expected population growth by a simple extrapolation from recent census counts.

The census data used in our analyses should not be assumed to be a precise description of the resident population. The demographic rates were calculated from those identifiable animals observed to be present in the study area each year. These tallies omitted the small number of dolphins that were unidentifiable (estimated at <5% of total) and several that were long-term residents but had moved to adjacent waters for one or a few years⁵. However, unless the uncounted dolphins were a biased subset with respect to demographic rates, the estimates made from the animals observed in each year provide estimates of demographic rates with more accuracy and reliability than is available for almost any other free-ranging, wild population of vertebrates.

Among the advantages of using demographic models to project population growth from demographic rates, rather than simply extrapolating from recent observed census numbers, are: short-term fluctuations that result from perturbations of the age distribution will not carry forward and amplify projected trends in population growth; components and causes of variation in growth can be examined; effects of changes to specific demographic rates can be explored; and our understanding of the population dynamics can be tested by comparing results from mechanistic models to observed population trends and patterns.

The range of projections in the simulations encompassed the observed trends from 1993 to 2020, and the Validation scenario projected the same population growth rate as observed (r = 0.019), thereby validating that the model is a plausible representation of the population demography. The census trend with larger than projected size up through 2005, but with an apparent plateau subsequently (Figure 2), might indicate that the population has reached the ecological carrying capacity of the habitat. However, apparent trends could result from random fluctuations in population growth. In the decade since 2010, the observed population growth has been approximately parallel to the mean prediction of the simulations. With respect to validating the estimated demographic rates, it is notable that the variance in annual growth across years of the simulation was due much more to environmental variation and intrinsic demographic stochasticity than to uncertainty in estimates of mean demographic rates (Table 6). The mean demographic rates in this population have been sufficiently accurately estimated from the 27 years of surveys to provide reliable estimates of mean birth and death rates for modeling the population, while short-term fluctuations are expected due to the natural variability in the environment and demographic processes.

The Baseline population model that includes all additions (births plus immigrants) and losses (mortalities plus emigrants), projects a long-term mean population growth rate of r = 0.021, with variation among independent iterations of SD (100 years mean r) = 0.007, 95% CI = 0.007 to 0.034. Thus, although precise estimates of the mean of simulated trajectories can be made arbitrarily small by running a large number of repeats of the simulation [SE(mean r) = 0.00002 with 100,000 iterations], the intrinsic uncertainties of population processes, augmented by uncertainties in model parameter estimates, leaves us unable to predict accurately the fate of any specific population (whether the real one or one of the simulated trajectories).

The contributions of immigration and emigration to the observed population growth are hard to quantify, but a plausible assumption is that immigration and emigration are approximately off-setting on average, although much of the observed annual fluctuations in census numbers could be due to short-term net inflow or outflow of dolphins in response to local conditions. Without immigration, and with mortality estimated from the reported ratio of documented deaths relative to total deaths of 0.33 (the Estimated mortality scenario), we project intrinsic growth rate of r = 0.023, close to the projected growth when all immigration and emigration is included in the model. Even when all disappearances are assumed to be deaths and no immigration is included in the model (the Maximum mortality scenario), there was still a projected growth of r = 0.016. Thus, the Sarasota Bay population is not dependent on the few immigrants per year to be demographically stable, immigration and emigration contribute only moderately to the population dynamics, and it is appropriate to consider this community to be a biological population.

Several caveats about the projections should be recognized. First, a portion of the uncertainty in the estimated long-term mean growth was due to uncertainty in the demographic rates entered into the model. Thus, there remain improvements in precision of population growth projections that can be achieved by more data on demographic rates. Moreover, the census population size fluctuated more from year to year (SD(annual r) = 0.061) than occurred in the simulation (SD(annual r) = 0.035 in the Validation scenario). This suggests that there were causes of variation in population growth that were not captured in the model. Although almost all residents present in the population each year were tallied in the census data, some of the variation in observed population growth could have been caused by small sampling errors in the annual counts.

Further monitoring of the population size and demographic rates will be necessary to confirm the actual carrying capacity of the Sarasota Bay habitat. Although the population size has remained relatively flat over the last few years, suggesting that it might be near its carrying capacity, there is not yet evidence of density-dependent reductions in reproductive success. Moreover, the concept of carrying capacity simplistically assumes that the environment is stable over time, rather than increasing, decreasing, or being in a constant state of flux. Several recent changes likely affected carrying capacity and birth and death rates. A ban on commercial fishing with large gill nets in inshore Florida waters was imposed in 1995, and the increased population growth of dolphins from 1995 to 2000 could have resulted from increased prey availability and shifts in foraging patterns that might have improved reproduction (the 2 years with the highest birth rates were 1996 and 1999) or led to increased immigration into the local population. (There was a mean of 3.67 immigrants/year from 1995 through 2000, compared to a mean of 1.43 per year in the other 20 years.) Declines in the population in the few years after 2005 and in the most recent few years (apparently continuing into 2020) could have been consequences of the severe red tide events that occurred during 2005-2006 and 2018-2019 (Berens McCabe et al., 2021). The future environment might be very different from that of the recent past. The 100-year simulations represent the expected trajectories if conditions remain as they are, but PVA projections should not be assumed to be a prediction of the future.

An important contribution of PVA theory was the recognition that the stability and even long-term persistence of populations depends not just on mean demographic rates but also on the variation over time (“environmental variation,” EV) caused by fluctuations in the environment (Gilpin and Soulé, 1986). Moreover, the temporal variation in demographic performance determines the number of years of data that must be sampled in order to obtain reliable estimates of long-term means in demographic parameters. However, data from a number of years are needed in order to obtain estimates of EV, especially if there are occasional extreme years (i.e., catastrophes) that lie outside of the more typical annual variations. EV must be partitioned from the contribution of individual sampling error or demographic stochasticity to total annual variation in observed rates (Akçakaya, 2002; Lacy et al., 2020), and the sampling error can be a larger component than is EV when the numbers of observed demographic events are small, making accurate estimation of EV difficult. Consequently, many PVAs use crude estimates of EV based on few years of data, use total annual variation as an upper bounds estimate, use expert opinion based on the life history of the species, or use unvalidated rules of thumb such as assuming that EV will be approximately 20% of the mean for each demographic rate.

The long-term data on the Sarasota Bay population provides an opportunity to estimate EV for demographic modeling of a long-lived species. However, even with 27 years of data, the small number of births and deaths each year results in large binomial sampling variation around population probabilities of each type of demographic event, especially since separate estimates must be made for each age class that is found to have statistically different mean rates (e.g., calves, juveniles, young adults, peak age adults, and older adults). Thus, our calculated values for EV are still only approximate, albeit more accurate than is available for most other comparably long-lived animals. For adult mortality and young female fecundity, the expected binomial sampling error was as large or larger than the observed annual variation, indicating that we have no evidence for EV in those rates. For calf mortality, juvenile mortality, and breeding rates for females beyond 12 years of age, there was substantial EV (Tables 2, 3), and those variations in rates contributed substantially to the variation in projected growth rate from year to year (Table 6). This suggests that reproduction and survival of still-dependent offspring are life stages that are more susceptible to environmental conditions or specific threats than is survival of adult females.

Uncertainties in the future of wildlife populations – even in PVAs that use the most complete demographic data – arise from residual sampling error around demographic rate estimates made from finite (and often small) sample sizes, intrinsic demographic stochasticity, environmental variation in key habitat characteristics, changing local and global environments due to increasing human activity, and future management or conservation actions that impact the species. Consequently, exploration of the extent of our uncertainty and the component causes via sensitivity analysis is a necessary part of any PVA (McCarthy et al., 1995; Mills and Lindberg, 2002).

When demographic rates were varied across + 2 × SE, we found that uncertainty in the breeding rate had the largest effect on the estimated long-term population growth, with the mean annual population growth resulting from the lower and upper bounds tested (approximately the 95% confidence interval) ranging from about 1% to about 3% (Figure 7). This might indicate that research to improve the accuracy of estimates of breeding rate could help reduce considerably the uncertainty in projected population growth. However, given that virtually all surviving births in the population are being documented and most females are of known age, the only option to improve estimation of breeding rate might be to extend the data over more years or over a wider geographic area. If any calves are lost before being counted in the surveys, those neonatal deaths would not contribute to our estimates of breeding rate or calf mortality, nor to the population growth.

When demographic rates were varied across the observed environmental variation (±2 × SD), the demographic rate that had the greatest influence on population growth was again the breeding rate (Figure 8). Variation in calf mortality and juvenile mortality also had large effects. The importance of environmental variation in adult female mortality was not tested because the data do not show that there was any annual variation in adult female mortality beyond that expected due to binomial sampling. This contrasts with the generalization in population ecology that adult female mortality is expected to have the largest effect on population growth for species with low fecundity and long reproductive lifespans (Brault and Caswell, 1993; Caswell, 2001). However, that generalization derives from sensitivity analyses that vary each demographic parameter by equal proportions. Yet, if environmental variation affects fecundity and survival of early age classes to a much greater extent than it affects adult survival, then proportional sensitivity analyses do little to inform us about the demographic rates that drive differences in population growth across time or space (Mills et al., 1999; Mills and Lindberg, 2002; Mills, 2013). Similarly, the extent to which human activities can affect each life history stage needs to be considered when evaluating which stage should be the focus of management (Manlik et al., 2018). For the Southern Resident Killer Whales, Lacy et al. (2017) found that recovery actions needed to address the recently low reproductive success, because there was little scope for improving the already high adult survival.

It is notable that the largest contributor to the year-to-year variation in population growth in the model was the intrinsic demographic uncertainty resulting from the stochasticity of biological processes. This intrinsic stochasticity accounted for 71% of the variation in annual change in population size, in contrast to 22% of such variation being due to EVs in the probabilities of reproduction and mortality, less than 6% due to our imprecision of estimating mean demographic rates, and less than 2% due to the variation in numbers of immigrants. These results indicate that the natural variation in environmentally determined annual demographic rates and the probabilistic nature of individual demographic events are much greater factors limiting our ability to predict the population change in any given year than are the uncertainties in our estimates of mean demographic rates for this population. However, it should be reiterated that the model projected less annual fluctuation in population growth than has been observed, indicating that some sources of variation are underestimated or not accounted for in the model.

The use of population data from recent years to project future population trajectories implicitly assumes that the environment and the population’s responses to it (including any evolutionary adaptation) will remain unchanged. Thus, PVA baseline models should be seen as projections from current or other hypothesized conditions, rather than as predictions of the future. PVA modeling can be used to explore alternative possible futures through testing alternative demographic rates, predicted trends in rates, or functional relationships to environmental variables that can themselves be projected from other models (Keith et al., 2008; Aiello-Lammens et al., 2011; Lacy et al., 2013). The assumption of constancy of demography quantified from past observations will be increasingly challenged by the impacts of global warming and other aspects of climate change. PVA has been used to predict direct and indirect impacts of climate change on species viability (e.g., Brook et al., 2009; Lacy et al., 2016; Williams et al., 2017).

Increasing water temperatures occurring with global climate change might cause increased mortality to the dolphins in Sarasota Bay and the Gulf of Mexico due to increased metabolic demands (Costa et al., 2013), mobilization of organic pollutants that had been sequestered in blubber (Yordy et al., 2010), or increased exposure to pathogenic microorganisms (Buck et al., 2006). Although the extent of any such increases in direct and indirect threats due to climate change has not yet been quantified, the higher observed mortality in summer months suggests that future increases in water temperatures could have negative impacts (Wells, 2010). Other anthropogenic changes to the environment such as amount of accumulated toxins (see text footnote 3) and frequency of direct human interactions (Wells et al., 2008)⁶ can also potentially cause increased mortality in the future. Although we do not yet know if and to what extent mortality will change, we can test what increase in mortality would be tolerated before causing the dolphin population to go into decline. Not surprisingly, a 0.5% decrease in survival was projected to reduce population growth by about 0.5% (from r = 0.021 to r = 0.016), and a 2.5% reduction in survival results in a population with no growth. Thus, with an estimated K = 200, the loss of an additional one dolphin per year (0.5%) would cause significant harm. Conversely, a reduction in the number of deaths due to human interactions will yield comparable improvements in population growth (McHugh et al., 2021).

Similarly, PVA models can be valuable in informing estimates of the sustainable take under conditions of uncertainty. Incidental mortalities of marine mammals resulting from fisheries (bycatch) is regulated in the United States (NOAA, 2004) and elsewhere (Williams et al., 2008; Lonergan, 2011) based on the potential biological removal (PBR), calculated as PBR = N_min (R_max/2) F_R, in which N_min is a minumum population size estimate, R_max is the net reproductive rate (symbolized r in this paper, to distinguish it from the lifetime reproductive rate, which is often designated as R), and F_R is a recovery factor (Wade, 1998). Following Wade (1998), R_max is usually assumed to be 0.04 for cetaceans, unless there is population-specific demographic data to suggest otherwise. The use of a minimum estimate of N, division of R_max by 2, and application of the recovery factor are intended to make the calculation of sustainable losses conservative, thereby accommodating uncertainty in the values (Wade, 1998; Moore et al., 2009). However, Punt et al. (2018) noted that the efficacy of PBR approaches to protect vulnerable cetaceans from excess human-induced mortality was highly dependent on the accuracy of the default assumption of R_max = 0.04.

The Sarasota population is projected to grow at only half the rate often assumed in many PBR calculations for cetaceans. Prior PVAs on cetaceans have generally found r to be substantially less than 0.04 (e.g., Manlik et al., 2016; Lacy et al., 2017; Williams et al., 2017; Cervin et al., 2020; Ashe et al., 2021; Murray et al., 2021), but often those analyses were completed on populations already known to be in decline or at risk of decline. The Sarasota population is perhaps a better model for a population that is growing. Sarasota Bay is not a pristine habitat, as deaths occur due to human interactions (see text footnote 6), dolphins accumulate harmful levels of pollutants (Wells et al., 2005; Yordy et al., 2010; see text footnote 3), and red tide events might be exacerbated by both climate change and introduction of anthropogenic nutrients from the watershed (Van Dolah, 2000; Brand and Compton, 2007; Wells, 2010). However, no parts of the oceans are pristine and unaffected by anthropogenic activities, and the Sarasota Bay population of Tursiops might be considered to be typical of moderately affected coastal populations of marine mammals. Thus, the baseline maximum population growth for calculating PBR of cetaceans perhaps should be adjusted downward, as the standard formulation of PBR might be less conservative than has been assumed.

Among the threats that can affect dolphin survival and reproductive success are the harmful algal blooms such as red tide (Karenia brevis). The most severe red tide events in recent years (in 2005-2006, and 2018-2019) were associated with higher calf and juvenile mortality and population declines in 2006 and 2019. Model scenarios that either omitted or doubled the years with such increased mortality projected very small differences in population growth compared to the Baseline scenario. This indicates that the resistance of adult dolphins to the impacts of red tides (perhaps reflecting their behavioral flexibility and ability to adjust foraging patterns [Wells, 2010]) and the long reproductive lifespan of dolphins provides tolerance of the population to such events. The apparently increasing frequency of severe red tide events (Van Dolah, 2000; Brand and Compton, 2007) and the possibility of other or longer-term impacts on the population should be monitored to determine if the few simple red tide scenarios that we tested adequately represent the threats that the population might face in a changing environment.

Although the Sarasota Bay population of dolphins is not at risk of being so small and isolated that it would be demographically and genetically unstable if recent demographic trends continue, the detailed demographic model does provide an opportunity to evaluate the minimum viable size (MVP) of isolated coastal populations of dolphins. When we modeled a closed population with no immigration, populations of size N < 30 were found to be vulnerable to extinction, even in the absence of any inbreeding depression. The high survival and long reproductive lifespan of bottlenose dolphins affords demographic stability at much smaller sizes than are reported for MVPs of most mammalian species (Traill et al., 2007). However, N > 50 was necessary to keep genetic decay over 100 years (5 generations) to less than 90% of initial heterozygosity (a common criterion for genetic viability; Soulé et al., 1986), which is equivalent to accumulated inbreeding remaining below F = 0.10 (Lacy, 1995). This translates also to an effective population size of N_e = 23 for a population of size N = 50, giving an estimated N_e/N ratio of 0.46. This is likely an over-estimate, because the Vortex model assumes that all individuals of an age class are equally likely to reproduce and that mating is at random. Any greater than random variation in reproductive success would depress N_e/N and accelerate loss of genetic diversity.

While species with long reproductive lifespans and low fecundity can be demographically stable at small population sizes, and long generation times also reduce the rate of decay of genetic diversity in closed populations, those same life history characteristics can prevent the populations from recovering quickly from any catastrophic declines, such as those caused by disease outbreaks and spills of oil or other toxins. Our tests of resilience to such declines show, for example, that the Sarasota Bay dolphin population would not be expected to recover to its current size for 28 years after a 50% decline. Scott et al. (1988) reported a 53% reduction in abundance of a mid-Atlantic migratory coastal population of dolphins due to a morbillivirus outbreak, and they estimated a median recovery time of 32.5 years, or a median of 50.5 years if human-induced mortality due to fisheries continued at pre-outbreak levels.

A long reproductive lifespan can, however, buffer a population from short-term suppression of reproductive success, because females can skip a year of breeding and then resume or even compensate with increased breeding rate the following year. Thus, a 25% 1-year reduction in the breeding rate (affecting only the fewer than 30% of adult females that would have bred that year) can have much less consequence to the population than would a 25% reduction in survival in a catastrophe year. When catastrophic events reduce survival or breeding over a number of years, however, then the damage compounds and even long-lived species cannot cope by delaying breeding and responding with more reproduction a year or two later. Using a sex- and age-structured demographic model, Schwacke et al. (2017) projected that the dolphin population in Barataria Bay, Louisiana, would require 39 years to recover to its prior population size following the Deepwater Horizon oil spill in the Gulf of Mexico. The long recovery time resulted from reductions in survival that were estimated to last up to 10 years (with a 35% cumulative reduction observed over the first 3 years after the spill), compounded by a reduction of about 70% in reproductive success in each of the first 3 years and an estimated 17 years before reproduction would return fully to pre-spill levels.

The availability of data from 27 years of consistent monitoring of the individual dolphins resident in the Sarasota Bay population provided birth and death rates, allowed analysis of transition points in demographic rates as dolphins age, and provided the basis for a detailed demographic model that was validated against census data. The remaining uncertainties in population projections were partitioned into those due to sampling limitations, environmental fluctuations, and inherent stochasticity of demographic processes. Beyond the assessment of the Sarasota Bay population, the analyses can inform estimates of Minimum Viable Population (MVP) size for isolated, small populations and Potential Biological Removal (PBR) for directly or indirectly harvested populations. We used the PVA for some initial assessment of the resistance and resilience to threats affecting survival or breeding, and it can provide a basis for further investigations of the population-level impacts of causes of mortality that might be partitioned and quantified (e.g., McHugh et al., 2021) or of projected improvements in reproduction or survival resulting from management actions. Williams et al. (2020) noted that as the multiple effects of chronic ocean noise on cetaceans are measured, PVA can be used to quantify the population level impacts. Schwacke et al. (2017) described their demographic model as an important first step in determining the type and amount of restoration action needed to compensate for damage from the oil spill. Demographic impacts of singular or occasional events on the Sarasota population such as red tide, disease outbreaks, and the ban on large gill nets are difficult to quantify amid the multiple causes of population fluctuations, but long-term monitoring, data on other affected and non-affected populations, and mechanistic models of effects can provide estimates of impacts that could be explored through PVA. Such analyses are likely to be especially useful when integrated with other modeling approaches that have been applied to threat assessments for marine mammal populations, such as Pathways of Effects (PoE) (e.g., Murray et al., 2021), Population Consequences of Acoustic Disturbance (PCAD) (Williams et al., 2020), accumulation of Persistent Organic Pollutants (POPs) (Hall et al., 2006; Lacy et al., 2017), environmental changes driven by climate change (Williams et al., 2017), and metamodel linkages (Lacy et al., 2013) to examine predator-prey interactions with multi-species PVA (Lacy et al., 2016).

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: The Vortex input files with all scenarios examined in this study are available in the zenodo.org repository at http://doi.org/10.5281/zenodo.5541834.

The animal study was reviewed and approved by the National Marine Fisheries Service Scientific Research Permits, with annual approvals by the Mote Marine Laboratory’s IACUC.

RW provided and summarized the census and demographic data used for these analyses. RL constructed the population model and ran the scenarios. RL and RW interpreted the results and wrote the manuscript. RW, MS, KU, SH, JA, and AB were extensively involved in the collection and/or processing of survey and/or life history data, and/or oversaw data collection or processing operations. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.