Answers are grounded in this paper and cite their sources.
A fundamental goal in conservation biology is the recovery of at-risk species. It is easy to assume that when a species is listed as threatened or endangered under the IUCN Red List or US Endangered Species Act processes, that conservation planning and species protection will follow, but this does not necessarily happen (Farrier, Whelan & Mooney 2007). In the United States, if a species is added to the Endangered Species list, the obligation of the responsible agency [U.S. Fish and Wildlife Service (USFWS) or National Marine Fisheries Service (NMFS)] is to direct implementation of a recovery plan which includes quantitative and measurable criteria by which to evaluate whether recovery ensues to the point at which the species can change status or be removed from the list. In contrast, the IUCN Red List serves to inform national agencies about the status of species within their boundaries. Farrier, Whelan & Mooney (2007) observe that the assumption that there is an automatic nexus between IUCN listing and conservation action is ‘illusory’. Broad reviews of global conservation efforts indicate that conservation research is not meeting the need to produce practical and applicable recommendations to implement in conservation actions around the world (e.g. Braunisch et al. 2012). In the United States, recovery criteria embedded in recovery plans are the focus of criticism from academics (e.g. Zeigler, Che-Castaldo & Neel 2013). The undertone of many of these articles is that government agencies are not adequately doing their job. The general conclusion is that recovery plans fall short, the criteria fall short, and species are not recovering. In part, these critiques are correct. They find that recovery criteria poorly reflect the biology of species of interest (Zeigler, Che-Castaldo & Neel 2013) and that recovery criteria are often qualitative and subjective (e.g. requiring that populations be ‘viable’ without specifying how an agency might quantify viability; Tear et al. 2005). At the same time, academic critiques often fail to acknowledge the constraints and practical realities of the recovery process. For example, it is not immediately obvious that statistical rigour is a primary limitation for recovery in most cases. In this article, we discuss our experience developing recovery criteria for an endangered butterfly, the Fender's blue. As leader (Schultz) and consultant (Crone) of the Recovery Team for Fender's blue, we worked closely with USFWS and other interested parties to develop recovery criteria that are grounded in ecological theory yet practical to implement (U.S. Fish and Wildlife Service 2010). To date, recovery efforts appear to be effective, in the sense that there are larger populations and more habitat for Fender's blue throughout its range (Fig. 1, Fitzpatrick 2013). We, first, describe the biology of the system, secondly, outline the recovery criteria and their theoretical basis and, thirdly, discuss these criteria in terms of practical and academic critiques. Fender's blue Plebejus (=Icaricia) icarioides fenderi is endemic to Willamette Valley prairies in western Oregon, USA. It persists in prairie remnants that support its larval host plants, Kincaid's and spur lupine Lupinus oreganus = L. sulphureus spp. kincaidii and L. arbustus. Only ~1% of historic prairie remain. At the time of listing in 2000, there were 32 small sites and ~2000–3000 individuals (Fig. 1). A decade of surveys, research and discussion on these species and several other rare native species followed, leading to the Recovery Plan for Prairie Species of Western Oregon and Southwest Washington which includes Fender's blue butterfly (U.S. Fish and Wildlife Service 2010). Fender's blue, like many butterflies and annual plants, has high variability in population size from year to year. In part, these fluctuations reflect weather conditions; adult butterflies fly in May, near the end of the winter rainy season. In cloudy and rainy years, butterflies are often unable to fly during much of their adult lives, and populations tend to crash (Schultz, Franco & Crone 2012). On very long time-scales, populations were probably unlikely to persist indefinitely at local sites. Historically, prairies in western Oregon were maintained by fires during the dry season, leading to mortality of larvae if fires came through a site, and succession to unsuitable habitat if they did not. Over management time horizons (years to decades), populations persist at the same sites, but over centuries, the species likely persisted as a classical metapopulation (Schultz & Crone 2005). Recovery calls for populations in each of three regions, spanning the historic distribution of Fender's blue. This decision was largely subjective. There is theoretical justification for having multiple independent populations (‘redundancy’ sensu Tear et al. 2005), and ‘recovery’ in its literal sense implies distribution throughout the historic range. But the decision for three regions (over two, four or five) was largely based on the locations of known populations, expert opinion that having multiple regions spreads the extinction risk across the range, and local planning units. Thus, this criterion is quantitative, measurable and based on ecological theory in a general way, but the choice of three zones was largely practical. Criteria for what ensures viability in each region are described below. We initially calculated population viability and extinction risk using standard population viability analysis (PVA) approaches. Specifically, we performed a count-based PVA (Dennis, Munholland & Scott 1991) using annual monitoring data, that is transect counts of butterflies at all known sites (Schultz & Hammond 2003). Extinction probability over a given time horizon is a function of the current population size, population growth rate and variance in growth rate and is a density-independent calculation (Dennis, Munholland & Scott 1991). For Fender's blue, USFWS chose a criterion of 95% persistence for 100 years for recovery and 90% persistence for 25 years for downlisting from an endangered to a threatened species. In our initial analyses, we assumed that environmental variation was not under management control. Of the two remaining parameters, variation in extinction risk, as calculated for each site, depended more on growth rates than population size. Therefore, our published recommendation for recovery was based on managing habitat to achieve suitable growth rates for persistence. In the presence of observed levels of variation, a single population would require a very high growth rate to persist for 100 years (λ > 1·8, where λ is the annual rate of increase). The need for this extreme high growth rate can be moderated by spreading the risk of extinction among multiple patches (Schultz & Hammond 2003). If the probability that each of several populations goes extinct is independent, network extinction risk is PeM where M is the number of subpopulations in the network. We recommended creating networks of three independent populations with λ > 1·55 in each population (Schultz & Hammond 2003). When we reached the point of recovery planning, this criterion was not implemented for two reasons. First, although it is objective and measurable, achieving a specified population growth rate was perceived as too abstract. Secondly, λ > 1·55 for population persistence implies that the populations would grow unrealistically large. As populations recover, they could be regulated by density dependence. We did not have data on density dependence because few, if any, populations were near carrying capacity (McIntire, Schultz & Crone 2007). Therefore, we used the fitted PVA models to calculate the minimum population size needed to ensure 95% persistence for 100 years, given the observed population growth rate and variance at the site we believed was closest to recovered (‘Willow Creek’ in Fig. 1). For three independent populations per recovery region, this calculation implied that a minimum population size of over 70 000 butterflies in each of three sites would be needed to ensure persistence. Like λ = 1·55, this was an unrealistic number to expect. Therefore, we extended the notion of spreading the risk to multiple independent sites, until we achieved reasonable numbers of sites, and butterflies per site. With nine sites, a minimum of ~300 butterflies per site are required to achieve 95% probability of persistence in each region. Based on the distribution of existing and potentially restorable habitat patches, these numbers seemed reasonable, as did the assumption that habitat patches would fluctuate independently (McIntire, Schultz & Crone 2007). As a nominal design, we recommended three networks of three habitat patches, with a minimum of 1000 butterflies per network (~300 × 3 ≈ 1000), in each of the three recovery regions. We note that this criterion does not consider emigration and immigration because these processes are less important in parts of this highly fragmented system and because populations function independently, even those in close proximity (Schultz & Hammond 2003). Overall, this criterion is strongly rooted in ecological theory. It meets repeated calls for incorporating population viability-based metrics into recovery planning and contributes to a growing literature that uses expected minimum abundance as a viability metric (EMA, McCarthy & Thompson 2001). At the same time, many of the decisions we made were strongly rooted in pragmatism. Using a population minimum, rather than a mean, is relatively robust to unknown factors such as density dependence; by the time a population reaches carrying capacity, it should be well above the minimum number of butterflies for persistence. Naïve application of population viability models could easily have hindered recovery efforts. For example, it would have been counterproductive to use λ > 1·55 or highly precise statistical estimates of average population size, as the viability criterion. Given limited resources, attainable management goals become a higher priority than unattainable ones. Unrealistic targets or impractical time horizons would have limited the ability of managers to focus on this species. Next, we set a recovery criterion for minimum habitat area. We based this criterion on Fender's blue movement behaviour (Crone & Schultz 2003). Specifically, Fender's blue is a mobile insect, with only partial affinity for its habitat, which is defined by the presence of its host plant, lupine (Schultz 1998). Fecundity is proportional to the amount of time females spend in lupine patches. For organisms with these traits, a long-standing body of theory (Skellam 1951) relates movement behaviour and patch size to population viability. In brief, the realized reproduction of a female butterfly is proportional to the time she spends in her natal patch, multiplied by her maximum fecundity if she spent her entire life in a habitat patch. The time she spends in the patch depends on movement behaviour and patch size. Animals that move quickly and have low habitat affinity leave their natal patches quickly; animals that move slowly or have high habitat affinity leave their natal patches more slowly. In smaller patches, animals encounter edges more frequently and are more likely to leave the habitat. Through a series of field and modelling studies, we estimated this minimum patch size, based on movement. First, we used biased correlated random walk (BCRW) models (i.e. models with random movement, except for attraction to host plant patches at edges) to estimate the relationship between patch size and residence time (Schultz & Crone 2001). Next, we scaled residence time to population growth rate by assuming that lifetime reproduction was proportional to time spent in the habitat (Crone & Schultz 2003). We calculated that ~6 ha of high-quality lupine habitat would support a viable Fender's blue population in the absence of immigration from other patches. This calculation corroborates our decision to use movement, rather than a food-based estimate of carrying capacity, to estimate minimum area requirements. For example, ‘high’ butterfly densities can be on the order of 500 ha−1 (McIntire, Schultz & Crone 2007). Thus, a patch size based on estimates of food and area requirements, for example to support our minimum number of 300 butterflies if they were to stay in the patch, is much smaller than the 6 ha needed to keep them in the patch long enough to maintain viable populations. As the final major recovery criterion, we wanted to ensure connectivity among habitat patches. Landscape connectivity is an important measure for species, with a metapopulation structure and/or populations inhabiting a highly fragmented landscape. Before Fender's blue was listed, we used movement data and a correlated random walk model (Kareiva & Shigesada 1983) to estimate the inter-patch distance that would support connectivity in a restored system (Schultz 1998). This work suggested that typical lifetime movement of Fender's blues was ~2 km. Simulations including attraction to habitat patches indicated that at 1 km distance, we would expect to see routine movement among patches (Schultz & Crone 2005). Thus, for a recovery criterion, we combined these two results as follows: source populations (≥6 ha in size) should be <2 km apart or include stepping stones (<6 Ha lupine patches) with an inter-patch distance <1 km. This criterion is supported by more detailed simulation work (McIntire, Schultz & Crone 2007) that indicated that dispersal over longer distances occurs sporadically when there are large source populations in ‘good’ years. These rules-of-thumb were used to set clear guidelines for landscape structure. Based on this network design, it led to identification of potentially restorable networks. In addition, it allowed USFWS to prioritize monitoring and management of these networks over scattered, isolated small patches that are unlikely to substantially contribute to recovery. Our experience highlights several tensions between academic and practical aspects of conservation. Notably, there is a tension between the need for data-based decision-making and limited data available for many species, especially for non-flagship species. Numerous recovery plan reviews advise PVA frameworks for recovery criteria (e.g. Zeigler, Che-Castaldo & Neel 2013). Himes Boor (2014) recommends that criteria specify significance level to reduce type II error, the likelihood of falsely concluding that a species has recovered when it has not. These analyses require substantial data to conduct informative assessments. In the absence of sufficient data, Zeigler, Che-Castaldo & Neel (2013) review alternatives to PVA approaches such as genetic monitoring, occupancy models and the IUCN framework. In fact, the fundamental issue for interpreting population viability of Fender's blue was its extreme variability. In this light, rigorous statistical precision will never be possible. Nonetheless, we feel confident that Fender's blue recovery criteria do a good job of balancing pragmatism and rigour. Attention to population viability has also led to increased rigour in monitoring (discussed below), although the fundamental issue of variability remains the same. From a practical perspective, we have also seen some tension between theory and management. For Fender's blue, connectivity and patch size criteria are central to the development of many local efforts including habitat conservation plans (Benton County, 2010; Yamhill County HCP, in development) and other initiatives. The inter-patch distance is central to putting specific guidelines on maps for county planners. However, the population viability criterion has been less popular with managers and decision-makers because the criterion requires at least 10 years above the minimum to be met; if a new site is discovered, it cannot contribute to recovery until the new population has been monitored for 10 years. This time frame is inherent for any abundance-based metric that may be fluctuating or declining. We use population minima as a criterion in part to minimize the length of monitoring. However, one unforeseen consequence is that monitoring for the minimum requires estimating population size every year, or at least in an unbiased set of survey years. In practice, field biologists like to survey butterflies at times and in places where butterflies are most abundant. Although using minima (as opposed to averages) does not reduce the length of the monitoring period, it creates a need to survey in poor conditions. Several recommendations advocated by academics were incorporated into the Fender's blue recovery plan; other recommendations seem impractical in the context of this case study. For example, revision of monitoring approaches to assess population relative to minimum population size criteria followed completion of the recovery plan. Setting rigorous statistical guidelines with set sample sizes would have hindered discussion and development of the new monitoring scheme. We also noted that recommendations for recovery planning often emphasize models and monitoring, but rarely the importance of experiments. In our experience, it is the combination of models, monitoring and targeted experiments that enabled progress towards recovery. In addition to using experiments to develop the movement rules (e.g. Schultz & Crone 2001), we have used experiments to develop and evaluate methods for habitat restoration and management (e.g. LaBar & Schultz 2012). We close with a few additional reflections and lessons we have learned. First, we believe the recovery planning process with Fender's blue has been a tremendous success. It has led to specific advances in species-specific monitoring protocols to gauge progress towards monitoring goals (Collins et al. 2011) and to the identification of potentially restorable networks, which has allowed agencies to focus research, management and land acquisition in a systematic fashion. As a result, optimism is high as populations have grown across the region and the acreage of high-quality lupine habitat has mushroomed. Secondly, Fender's blue has benefited from managers and academics who care, who seek out resources and who work together. At this point, it may appear that work on Fender's blue has been especially well-funded by management agencies. In fact, USFWS spends much less on invertebrates than vertebrates in general. Fender's blue is no exception, and much of the work summarized here was done only because we, and members of our research groups, contributed our time. Thirdly, academics need to be willing to use ecological theory coupled with limited data to propose quantitative metrics that are useful to managers, even if bounded by large but reasonable ranges. Academics often have too little confidence in the data to provide such criteria, or do not approach questions in a manner that is tangibly useful to managers (see Braunisch et al. 2012). Conservation would benefit from a change in this perspective. Fourthly, the goal of setting objective, measurable criteria is not a fixed framework to be applied to all listed species. Rather it is a creative process of people working together to draw out knowledge of species biology from limited data and from targeted focal grounded in ecological theory to develop practical criteria that can recovery planning efforts. We at Fish and Wildlife US of and other agencies in the Willamette the Prairie Species Recovery and other people in Fender's blue conservation for leading to these in We for to the Fender's blue monitoring protocols and Fitzpatrick for in and monitoring for the We Himes Boor for on the We Washington and for support this our work on this was by an to Schultz and an in to the to have not been because this does not Schultz is a conservation and has been working with Fender's blue butterfly and other prairie butterflies for over two is of USFWS Recovery Team for Western Oregon and Southwest Washington Prairie which includes Fender's blue butterfly and is an at Washington Crone is an of at is a theoretical who often with applied and managers to data and theory to on conservation planning for plant and populations.
No takes yet. Share an insight, caveat, or question.
Schultz et al. (2015) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: