A profound lack of data hinders managers’ abilities to set scientifically defensible recovery goals and criteria for all but a few species that are listed as threatened or endangered under the Endangered Species Act. Given such data gaps, managers tend to jump between generic conservation rules and expensive, time-consuming, and often unattainable single-species population viability analyses (PVAs). This project sought to develop an analytical framework that would take advantage of shared traits and threats across many species to develop a pathway towards a more defensible system of developing recovery criteria. Multiple database resources containing compiled information on species recovery data, population trajectories, and life-history traits were developed.

This project had two objectives in undertaking an informatics approach to conservation. The first was to develop an analytical framework for inferring critical conservation information based on shared threats and traits. The second was to develop a resource that managers and agencies could use to make better conservation decisions, possibly following analytical frameworks or using the data in new ways. Database resources were used to make breakthroughs in understanding the linkages between species biology, conservation potential, and recovery criteria. However, there still were not enough species-specific data to allow robust cross-species modeling. In adjusting to this reality, the objectives were adapted to include: (1) developing database resources to enhance conservation management; (2) understanding key patterns in recovery criteria; (3) improving the utility of PVA; (4) determining the potential to infer recovery criteria for poorly studied species; and (5) developing new approaches for inferring traits for poorly studied species.

The technical approach was to build a series of databases from the literature and recovery plans and use those data to carry out novel analyses. The largest resource is a set of databases of information from 288 recovery plans for 642 plants and approximately 400 plans for 528 animal species. Information was extracted on every aspect of listing and conservation status, habitat requirements, and from more than 100 traits of biological importance. In addition, resources on well-studied species, especially plants, birds, and mammals, were compiled and made available. These databases were used to carry out analyses on patterns in recovery criteria, patterns in PVAs, the ability to model recovery criteria based on traits, and the ability to infer traits using phylogenetic approaches.

A full examination of different aspects of recovery was undertaken, including how recovery is defined and how recovery criteria are linked to patterns of decline and species’ biology. Overall, despite years of criticism, recovery criteria continue to be defined more by the current status of the species (e.g., the species’ listing status and population levels) than by the specifics of their biology or individual needs. This suggests minimal opportunities to link current recovery criteria with biological traits under an analytical framework.

In realizing that most plans lacked quantitative data to support recovery criteria, the research team closely examined one of the primary methods used to support recovery criteria – PVA. PVA is still considered by some scientists to be the “gold standard” in establishing defensible recovery goals. However, PVAs have also been criticized because uncertainty inherent in the modeling process may make it an inappropriate tool for assessing absolute outcomes or prescribing absolute population sizes. This study revealed that PVAs have seen very limited use in recovery plans. PVAs have only been used to help determine delisting criteria for five listed plant species and are included in the description for only nine listed species. Furthermore, despite a long history of criticism and suggestions to improve rigor, most PVAs, as carried out, fail to meet minimum standards for use in recovery planning.

As an alternative to data-hungry mechanistic PVAs, this project presented a statistical approach for extracting parameters from time-series data that are relevant to the establishment of recovery criteria. The approach is based on the idea that certain average properties of stochastic processes may be predictable even when the details of the underlying process are unpredictable and/or unknown. The goal was to extend this type of reasoning to the estimation of a specific property of stochastic population trajectories: the probability of decline below a pre-defined threshold (i.e., quasi-extinction). This model was successfully used to output quasi-extinction probabilities for a broad class of population change processes.

Because a rigorous link between traits and published recovery criteria could not be found, the project used the procedures developed in the quasi-extinction model to explicitly link species traits, threats, and population trends (instead of recovery criteria). However, this attempt to build sets of similar species was unsuccessful. The team was not able to produce enough stable and significant comparison sets to proceed with the inverse modeling efforts that were planned. Considering the focus was on relatively well-studied taxa (plants and birds), this does not bode well for applying the method broadly at this point. Yet, even exploring simpler surrogacy approaches proved elusive. Based on these results, it was determined that a new approach was needed for leveraging the information about species in a way that would usefully inform their recovery. The project switched to evolutionary statistical models and focused attention on predicting species maximum population growth rates, a fundamental metric in population biology known as “little r” or r, the intrinsic rate of increase.

Very generally, r describes trends in population density and abundance and is an indication of the potential for a population to replace itself. As a fundamental life history trait, r integrates how long a species lives, patterns in death over the course of a typical lifetime (referred to as survivorship curves), and lifetime reproductive capacity into a single metric. This project made advancements in estimating r and also showed how r is more strongly related to taxonomic ancestry than it is to body mass, as is typically believed. The realization that r had a strong phylogenetic signal led the team to develop a model that could predict r based on shared traits, phylogenetic structure, and knowing the value of r for a subset of species in each clade. This method was successful for birds and mammals.

In the course of this project, four substantial databases were developed and released to the public. A series of analyses were carried out that showed the limitations of the current “state-of-the-art” approaches to conservation science. These limitations have two causes, still too-sparse data and also that generalizations between species are still elusive. Two new analytical techniques were developed that could have widespread value to the conservation community, both in and outside of federal lands. The first is a new method of producing key parameters typically generated from PVAs without detailed process data and the second is a novel phylogenetic approach to estimating key life-history parameters for species where there is a continued paucity of data.