Patterns of uncertainty in life-history and extinction risk for Arctic vertebrates

Conserving Arctic wildlife will be challenging given the ongoing environmental changes in the region. In addition, there is a lack of fundamental biological information for many Arctic species, including a dearth of knowledge surrounding conservation threats and the risk of extinction. In this study, we gather all available data on research effort and life-history traits to assess the current state of scientific knowledge surrounding 389 Arctic vertebrate species. We also quantify patterns of species evaluation by the IUCN Red List, a global database of conservation risk used to measure success and prioritize resources in many conservation programs. We find that 10% of Arctic vertebrates have been the subject of no peer-reviewed studies during the last 30 years, and that we have little life history knowledge for many species. Arctic marine fishes are especially poorly known with an average of 3.5 (out of six) key life-history traits unknown. In a multivariate analysis, whether an Arctic species had been evaluated by the IUCN Red List was most strongly predicted by research effort and varied among taxonomic groups. In addition, we found that species that have been evaluated by the IUCN Red List continue to receive more research attention than species which have not been evaluated. Protecting all Arctic species may, therefore, require research programs and methods to halt research inertia and shift more attention onto species that are poorly known.


Introduction
The Arctic is composed of unique ecological communities considered to be among the earth's most biologically valuable areas (Olson and Dinerstein 1998). Yet, the conditions in northern polar areas present unique challenges for conservation. Large areas of Arctic wilderness are still intact (i.e., undisturbed by industrialized human development and transportation; Sanderson et al. 2002;Venter et al. 2016), but the Arctic is experiencing rapid climatic change, including greater climate warming than any other area of earth (Hansen et al. 2006). In addition, human activity in the Arctic is projected to dramatically increase as the loss of sea ice opens shipping corridors and allows increased industrial activity (Stephenson et al. 2011). Conserving Arctic wildlife in the face of these environmental changes will be difficult.
Although Arctic wildlife has been the focus of considerable conservation efforts, a major challenge in conservation decision-making in northern polar areas is a lack of population, demographic, geographical, and basic life-history data for many Arctic animals. Indeed, this scarcity of information was highlighted in the 2013 Assessment of Arctic Biodiversity (CAFF 2013) which stated, "Basic knowledge on the vast majority of Arctic biodiversity is limited" (p. 57), and "There are no abundance or trend estimates for many key populations and species" (p. 122). In many cases, local knowledge can address gaps in scientific understanding (Huntington 2000;Gilchrist et al. 2005). However, conservation success is thought to be maximized when local and scientific knowledge are both integrated in conservation plans (Becker and Ghimire 2003;Kainer et al. 2009). As such, a lack of scientific knowledge of Arctic species poses a problem for biodiversity conservation.
In addition to an overall lack of data, it is possible that research effort for Arctic animals is biased towards certain fauna. For example, certain charismatic species may be highly studied because of taxonomic biases in research funding (Tisdell and Nantha 2007;Fisher et al. 2011). Similarly, species that are frequently harvested may be relatively well understood because harvest programs provide a source of data (e.g., demography and population data, samples, etc.) that would not otherwise be available. Such biases pattern would not be surprising, given that global research effort shows geographical and taxonomical biases, biases towards larger species, and biases towards less endangered species (Brodie 2009;De Lima et al. 2011;Fisher et al. 2011;Brooke et al. 2014;Jaric et al. 2015). However, such patterns could be exacerbated in the Arctic because of the remoteness of many Arctic areas, the high cost of Arctic research (Mallory et al. 2018), and the importance of subsistence harvest of predominantly large-bodied species.
In this study, we compiled data on research effort and key life-history traits for 389 Arctic vertebrates. In addition, we collected data on IUCN Red List (IUCN 2017) evaluations to understand what factors determine whether a species has been evaluated by this program. The IUCN Red List is an objective, scientifically rigorous global database of conservation risk for species (IUCN 2017). Although many vertebrates have been evaluated (67% of global vertebrate species; IUCN Red List v. 2017-3 Table 1), there is considerable variation related to taxonomy and geography (IUCN Red List v. 2017-3 Table 1). These gaps in the IUCN Red List are important to understand because Red List data are used for a variety of international, national, and regional conservation initiatives (Gärdenfors et al. 2011), measure progress in the Convention on Biological Diversity Aichi Targets (CBD-AHTEG 2015), inform the Convention on International Trade in Endangered Species and the Convention on Migratory Species (UNEP-WCMC 2013;UNEP 2015), and influence resource allocation for research and conservation (Possingham et al. 2002). Species that have not been evaluated by the IUCN Red List will therefore not count towards the measurement of success in these programs, potentially imposing different conservation value on different species.
In this study, we address the following two primary research questions: (1) what is the state of scientific knowledge surrounding Arctic vertebrates, and how is scientific knowledge related to species traits and taxonomy? and (2) what factors drive IUCN Red List evaluation for Arctic vertebrates? By answering these questions, we hope to improve our understanding of conservation uncertainty for Arctic vertebrates and ultimately enhance biodiversity conservation in Arctic regions.

Data collection
For this study, we collected data from the peer-reviewed literature and readily available secondary sources (i.e., field guides and handbooks, online databases) for a suite of life history and conservation-related traits, for each of 389 Arctic vertebrates. Full details related to the data collection, including details related to species inclusion, are provided in the Supplementary Materials. 1 In brief, we collected data from hardcopy and electronic databases for as many species as possible related to (i) primary habitatclassified as either marine, terrestrial or freshwater, (ii) current IUCN Red List statuswith possible values of Least Concern, Near Threatened, Vulnerable, Endangered, Critically Endangered, Extinct in the Wild, Extinct, in increasing order of extinction risk, (iii) year of the most recent IUCN Red List assessment, (iv) number of threatsmeasured as the number of threats to the species listed in the IUCN Red List threat classification scheme, (v-viii) binary scores of whether each species is threatened by climate change, habitat alteration, invasive species, or overexploitation, respectively, (ix) adult body mass, (x) maximum longevity, (xi) age of female maturity, (xii) fecundity, (xiii) reproductive rate, and (xiv) trophic level. These life-history traits (ix-xiv) were chosen because they are important variables driving population dynamics and because they are commonly reported in secondary sources. If separate values were reported for males and females (e.g., for variables ix, x), we recorded the mean of the values. If a range of values was reported (e.g., for variables ix-xiv), we recorded the center of the range.
In addition, we provide data on (xv) research effort, measured as the number of papers published on each species between 1987 and 2016 that were archived on Web of Science (Clarivate Analytics 2017). This metric includes research conducted outside of the Arctic region, which we permitted because conservation decision-making often draws on information from outside the focal region. As such, there is a stronger scientific basis for managing Arctic species that are well researched even if that research occurs outside of the Arctic, compared with species that are not well studied. Finally, we collected data on the (xvi) history of IUCN assessments for each species in our data set, including the year and IUCN Red List conservation threat level of all previous evaluations.
Data sources were ranked (see Table 1) and data collected in an iterative fashion; we first searched the highest-ranking sources (i.e., those judged to be highly reliable and to contain many data points required for our analyses). If data were found in a high-ranking source, we no longer searched for this value in the remaining sources. This approach facilitated a progressive reduction in "missing data" in our database and allowed the searching of increasingly specialized sources.
The data from this study are publicly available at doi.org/10.5281/zenodo.1149680 (Dey et al. 2018). Note that we do not use all of the data collected as part of the analysis described below, but we provide all of the data collected for the use of other researchers.

Statistical analysistaxonomic patterns
We first explored general patterns of research effort and trait knowledge across all species using descriptive statistics. Then, we tested for differences in total research effort among taxonomic groups. We classified animals into six taxonomic groups [marine mammals, terrestrial mammals, marine birds, terrestrial birds, marine fishes, freshwater fishes (amphibians were excluded from this and all subsequent analyses as there were only three species in our dataset)] based on their primary habitat use (i.e., trait i, above) and taxonomic classification.
Because the processes driving the number of published papers were unknown, we used model selection to select among models which all included the total number of publications as the response variable and taxonomic group as the sole fixed effect. During exploratory data analysis, we noticed that a large number of species had zero published papers, and that the number of papers published per species was overdispersed relative to a Poisson distribution. As such, we formulated all of our models with negative binomial error structures ( Table 2).
The models we selected among were (a) a negative binomial generalized linear model (nbGLM), (b) a zero-inflated nbGLM with a single intercept for zero-inflation, (c) a zero-inflated nbGLM with a fixed effect of taxonomic group for zero-inflation, (d) a hurdle nbGLM with a single intercept for zero-inflation, and (e) a hurdle nbGLM with a fixed Note: The data types column describes which data types were taken from each source, and refer to the 15 data types described in the "Data collection" subsection of the "Methods". effect of taxonomic group for zero-inflation. In each model, we included taxonomic group as the sole fixed effect.
Next, we tested for taxonomic differences in trait knowledge and IUCN evaluation using two binomial family GLMs. These used the number of known life-history traits, or a binary score of whether a species was Red List evaluated, as the respective response variables, and included taxonomic group as the sole fixed effect. In the latter case, species that were listed under the Red List as Data Deficient were categorized as not evaluated.

Statistical analysisresearch effort
Then, we examined how Red List evaluation and taxonomic grouping influenced changes in research effort over time using a negative binomial generalized linear mixed model. In this model, we used the number of papers published in each year for each species as the response variable. The fixed effects included year, taxonomic group, and whether a species had yet been evaluated by the Red List (in the given year). For example, if a species was first evaluated in 2004, all observations for this species up to 2004 would be scored as "Not Evaluated", whereas observations after 2004 would be classified as "Evaluated". To test how research effort changed in different taxonomic groups over time, we included the interaction of group and year in our model. To test how research effort changed over time as a factor of Red List status, we included the interaction of status and year in our model. To account for repeated sampling, we also included a random intercept for each species.

Statistical analysis -Red List evaluation
Finally, we performed a multivariate analysis of factors predicting Red List evaluation. This model was constructed as a binomial family generalized linear mixed model, with log10 (research effort + 0.1), number of known life-history traits, log10 (body mass) and trophic level as predictor variables. The response variable was a binary score of whether a species had been evaluated or not (the latter category included Data Deficient species). To control for differences in evaluation among taxonomic groups, taxonomic group was included as a random intercept.

Results
Across 389 Arctic species in our database, the median number of papers published from 1987 to 2016 was 42 (range: 0-18 679). One hundred and eleven species in our database (29%) have been the subject of fewer than 10 published papers, and 37 species (10%) have been the subject of zero published papers (Fig. 1A) over the 30-year period. The median number of life-history traits known for each species was 4 (range: 0-6). For 32 species (8%), four or more life-history traits were unknown. Two hundred and sixty-six species (68%) have been evaluated by the Red List.

Taxonomic patterns
We found considerable taxonomic bias in research effort, with marine fish species being the most likely to have no publications [28%: Fig. 1A; Table 3 (hurdle component)]. For species with at least one published paper, fish species, along with marine mammals, had the highest number of mean publications [freshwater fishes = 655 papers/species; marine fishes = 559 papers/species; marine mammals = 326 papers/species; Fig. 1A; Table 3 (count  component)]. In addition, there was relatively low life-history knowledge for fishes and relatively high life-history knowledge for mammals ( Fig. 1B; Supplementary Fig. S1 1 ; Table 3). Although most (68%) Arctic vertebrates have been evaluated through the Red List, only 15% of Arctic marine fishes have been evaluated, which is much lower than for other taxa (Fig. 1C; Table 3).

Research effort
Research effort on Arctic vertebrates has generally increased from 1987 to 2016 (effect of year, Table 4; Fig. 2B). However, research effort has increased significantly faster for certain taxonomic groups. After controlling for the confounding effect of Red List evaluation, our Fig. 1. The state of scientific knowledge of Arctic vertebrates. Panel A shows the density distribution (i.e., total area under the curve is equal for all groups) of research effort (number of papers published from 1987 to 2016) for six groups of Arctic animals. This panel demonstrates a large number of marine fishes have not been the subject of a published study, which is reflected in a low percentage of known life-history traits (body mass, longevity, age at maturity, fecundity, reproductive rate, and trophic level) for marine fishes in our database (B) and a low percentage of species being evaluated for extinction risk in the IUCN Red List (C). Note, light grey coloring in C indicates species evaluated as "Data Deficient". Sample size for each group is as follows: terrestrial mammals (27), marine mammals (35), terrestrial birds (105) analysis shows that publications on marine and terrestrial mammals, as well as marine and terrestrial birds, have increased faster than have publications on marine fishes [ Table 4 (interaction effects)]. In addition, there are significantly more papers on species which have Note: Raw model estimates are reported along with the 95% confidence intervals (CI). Details of the statistical analysis are presented in the methods. *, because of complete separation (Albert and Anderson 1984), these estimates, confidence intervals and P values are unreliable. Note: Raw model estimates are reported along with the 95% confidence intervals (CI). The intercept term represents a reference level with status = not evaluated and group = marine fishes. Details of the statistical analysis are presented in the methods.
been evaluated through the Red List compared with species that are not evaluated [Table 4 (main effects); Fig. 2A]. However, the number of studies on evaluated species has increased more slowly than species that are not evaluated [Table 4 (interaction effect)].

Red List evaluation
After controlling for taxonomic group, Red List evaluation was significantly predicted by research effort, with more research effort increasing the likelihood that a species has been evaluated (Fig. 3A). No other variable included in our model was a significant predictor of Red List evaluation (Fig. 3).

Discussion
Although some Arctic vertebrates are well studied, many have never been the subject of a published scientific study. In addition, nearly one-third of Arctic vertebrates have not been evaluated by the IUCN Red List and we lack fundamental life-history data for many of these species. While these patterns were generally known prior to this study, we provide the first quantification of the scientific knowledge scarcity for Arctic species and identify patterns of conservation uncertainty present in Arctic regions. In addition, we demonstrate a strong relationship between research effort and IUCN Red List evaluation for Arctic vertebrates. Our analysis shows that species are more likely to be evaluated through the Red List if they have been the subject of more published studies (Fig. 3A). However, species that have already been evaluated receive more research effort Fig. 2. Changes in research effort for Arctic vertebrates from 1987 to 2016. Panel A shows the changes in median papers per species (line and dots) for species evaluated and not evaluated through the IUCN Red List, respectively. Dot size is scaled to the number of species in each category, and the interquartile range is shown in grey. Note that dot size changes across years in panel A because species can move from "Not Evaluated" to "Evaluated" during the time series. Panel B shows the changes in median papers per species for each of six taxonomic groups, which are indicated by icons on the right. Animal silhouettes accessed through PhyloPic 2018 via public domain dedication licensing. than unevaluated species (Fig. 2A). These findings suggest that shifting research focus onto species that are not currently evaluated would help to address data gaps that hinder conservation assessments.
For example, we show that there is relatively little biological and conservation knowledge of Arctic marine fishes. Although approximately 40% of global marine fishes have been evaluated by the Red List (IUCN 2017), we found that only 15% of Arctic marine fishes have been evaluated. In addition, 54% of Arctic marine fishes have been the subject of fewer than five scientific papers published, and we have a poor understanding of reproductive life-history of many species ( Supplementary Fig. S1 1 ). This is worrying given that commercial fishing in Arctic regions is expected to increase in the near future (Mcbride et al. 2014), and that Arctic marine environments are being invaded by poleward range expansions of southern species (Fossheim et al. 2015). Furthermore, some Arctic marine fishes are known to be keystone species (e.g., Arctic cod; Welch et al. 1992), which play an integral role in mobilizing energy from low to high trophic levels in Arctic food webs, and it is possible that some of the poorly studied species have similarly important roles in Arctic ecosystems. As such, protection of poorly studied Arctic marine fishes may be essential to maintain populations of higher trophic level species. Interestingly, some Arctic marine fishes are among the most well-studied Arctic vertebrates (e.g., many salmonids and cyprinids), perhaps because of their popularity for recreational angling, aquaculture and commercial harvest both within and outside of the Arctic. Although research effort should not only serve to address conservation needs, increased attention on understudied Arctic marine fishes is clearly warranted. Evaluation of species through the IUCN Red List is coordinated by a number of Red List Authorities and Global Species Assessment Projects, which represent different taxa and geographic areas. For example, between 2005 and 2015, the IUCN Species Program Marine Biodiversity Unit coordinated efforts to have 20 000 marine vertebrates, invertebrates, plants, and corals evaluated through the IUCN Red List. Although this group outlined priority regions for species evaluations (e.g., the Gulf of Mexico, the Arabian Gulf), the Arctic was not among these regions. As our study shows Arctic fishes (both marine and freshwater) are poorly represented in the Red List and should be a priority for future evaluations. Fortunately, there is already an expert group related to Arctic fishes established through the Arctic Council (the Fishes Expert Network), which could be engaged by the IUCN as a partner in these evaluations. However, to accomplish these evaluations, we will also require increased biological information for many species.
Depending on the true extinction risk of species that are not currently evaluated, somewhere between 9% and 41% of Arctic vertebrates are threatened with extinction (this study). These extremes paint very different pictures of the status of Arctic biodiversity and suggest that dealing with uncertainty in the risk of extinction will be required for Arctic conservation projects in the near-term. In general, researchers suggest using a precautionary principle for unevaluated species (i.e., treating them as Threatened) when considering conservation offsets and mitigation measures (Mace et al. 2008;Butchart and Bird 2010). However, treating unevaluated species as Threatened is not appropriate during protected areas planning processes because it de-values protection of species known to be Threatened and potentially protects areas of low biodiversity value (Butchart and Bird 2010). Regardless of how current uncertainty is approached, it would clearly be beneficial to reduce uncertainty in species conservation risk by prioritizing research effort for unevaluated species. However, our study demonstrates that unevaluated Arctic vertebrates receive lower research effort than species that have been evaluated ( Fig. 2A; Table 4). Addressing this bias in research effort would decrease conservation uncertainty and increase the success of Arctic biodiversity conservation.
The Arctic is one of the most ecologically intact places left on earth (Sanderson et al. 2002), and conservation of Arctic species is therefore of global importance. Although there is a current thrust to protect Arctic ecosystems as part of international obligations to meet the Convention on Biological Diversity Aichi Targets (CBD-AHTEG 2015), these measures may be ineffective if they are not informed by scientific knowledge of local species. To maintain Arctic socio-ecosystems, we need further study of many Arctic animals; in particular, we call for increased research effort for Arctic marine fishes, some of which are virtually unknown to science beyond taxonomic identification. Local ecological knowledge could help address gaps in scientific knowledge and should always be considered during conservation planning. However, severe research biases should also be addressed within scientific funding schemes to gain an understanding of the basic biology of all Arctic vertebrates, which will help to effectively conserve biodiversity across this unique environment.