Close-kin Mark-recapture (CKMR): A Framework for Estimating Abundance and Demography from Genetically Related Individuals

Close-kin mark-recapture (CKMR) is a recently developed method for estimating abundance and demographic parameters (e.g. population trend, survival) from kinship relationships determined from genetic samples.  It is conceptually distinct from genetic mark-recapture as traditionally employed; instead of following distinct individuals through time, frequencies of certain relationships (notably, parent-offspring pairs and half-siblings) can be used to make inference about demography given an underlying population model and assumptions about sampling.  In essence, an offspring “marks” its parents and inference can thus be made from animals that are only sampled once. This method is potentially a “holy grail” for fish and wildlife agencies, given potential for making inferences about demography and abundance using catch or harvest data alone.  However, such models are quite novel and so far have been successfully employed on fish populations.  In this symposium, we summarize past and current research with the aim of providing attendees basic information on requirements for such analyses, including data processing, genetic analysis, statistical analysis, model assumptions, and sample size considerations.

1:10PM Close-Kin Mark-Recapture: Theory and Applications
  Mark Bravington
Close-Kin Mark-Recapture— CKMR— is a powerful technique that can provide cost-effective estimates of absolute abundance, survival rates, connectivity, and other demographic parameters critical for management. It takes advantage of the power of modern genotyping to find pairs of close relatives (parent-offspring and/or half-sibling) among large collections of biopsy samples. Then, it uses those pairs in an extended mark-recapture framework, based on the notion that each offspring “marks” its two parents. The samples can come from dead animals (e.g. commercial fish catches, hunting, bycatch) and there is no need to rely on other data sources that can be dubious (Catch-Per-Unit-Effort) or expensive (direct surveys, mark/recapture), although CKMR also fits happily alongside other data in a fisheries-style stock-assessment framework. This talk draws on 12 years of CKMR development at CSIRO, with examples ranging from large commercial fisheries through to otherwise-intractable endangered species. I will introduce the principles of CKMR, summarize what can and can’t be learnt from it, and comment on some issues that make all the difference between success and failure, including model construction, study design (sample sizes etc), genetics, and kin-finding.
1:50PM Identifying Kin Pairs for Close-Kin Mark-Recapture: Statistical-Genetic Models, Genetic Markers, and Software
  Eric C. Anderson
It has been nearly two decades since close-kin mark-recapture (CKMR) was first proposed in the literature. At the time, genetic data were in short supply, and allowances had to be made for the substantial uncertainty regarding inferred kin relationships. Even a decade after that, genetic technologies were still such that large-scale CKMR projects were generally feasible only for large agencies working on extremely valuable fishery species. Today, however, advances in sequencing capacity make CKMR projects within reach in a number of different contexts, and several kin-pair types can be identified with great accuracy. I will review the statistical procedures and underlying models used for identifying kin pairs, and then talk specifically about the advantages and disadvantages of different types of genetic marker data that are applied to CKMR. I will then present some recently available software packages that facilitate planning for CKMR experiments, especially by helping predict the number of markers necessary to resolve different levels of kinship. I will conclude with a discussion of physical linkage between markers. For resolving a few kin categories, this linkage is of little concern; however it greatly curtails our ability to distinguish between some other types of kin pairs.
2:10PM Close-Kin Estimates of Census Size and Effective Population Size
  Robin Waples
Close-kin mark-recapture (CKMR) methods provide new opportunities for estimating adult abundance (N) and other demographic parameters of natural populations. Initial applications used established procedures of parentage analysis to identify parent-offspring pairs (POPs). Advances in DNA-sequencing technology now provide enough power to extend the methodology to include identification of siblings, which alleviates some of the limitations of POPs-only analyses. However, there already is a widely-used method (Wang 2009 Molecular Ecology) to use the incidence of siblings to estimate effective population size (Ne). Therefore, it is essential to identify conditions under which CKMR methods based on siblings estimate N, when they estimate Ne, and when they estimate something else altogether. I show that a key factor is whether the siblings are from the same or different cohorts. Furthermore, estimates based on siblings from different cohorts are sensitive to correlations in realized reproductive success across time. Factors common to many species that can affect these correlations include skip breeding and persistent differences in individual reproductive success. I discuss these and other factors that can affect precision and bias of CKMR estimates of N and/or Ne, including population structure, sex ratio, vital rates that change with age, and non-random sampling.
2:30PM Inference for Abundance, Recruitment, Survival, and Parentage Using Molecular Data
  Jamie Sanderlin, Richard Barker, Mathew Schofield, Oliver Berry, Roger Kirkwood
Molecular parentage data is often employed to derive parameters that support wildlife management, including reproductive success, mating strategies, relatedness among individuals and effective population size. Parentage data has also recently been incorporated into capture-recapture framework to provide key additional ecological parameters, including abundance, with application primarily to fisheries. These models are based on discrete units of time while we focus on continuous time with latent times of birth and death. Also, previous inference is generally based on a maximum likelihood approach using pseudo-likelihoods, while we adopt a Bayesian approach implemented by Gibbs-sampling for all unknowns, except for abundance which was updated with a reversible-jump Metropolis-Hastings step. Here, we describe a new Bayesian open population mark-recapture model using the Crosbie-Manly-Arnason-Schwartz parameterization of the Jolly-Seber model for molecular parentage data that tracks survival and recruitment of individuals through time, in addition to estimating population size and relatedness. We demonstrate the method with data consisting of multi-locus genotype samples of offspring and potential parents sampled over fifteen years from a small introduced population of red foxes on Phillip Island, Victoria, Australia. Our method relies on data augmentation, since parents of sampled offspring may not be sampled and genotypes may contain errors. It provides the first estimates of recruitment and survival for this population, as well as estimates of abundance consistent with those derived by other means. We discuss the challenges of fitting models with parentage data within this framework.
2:50PM Refreshment Break
3:20PM Statistical Design of Close Kin Mark Recapture Experiments
  Hans J. Skaug, Mark V. Bravington
Close-Kin Mark-Recapture (CKMR) is a powerful new technique for estimating abundance, survival rates, and other demographic parameters. By collecting samples from many animals (alive or dead) and making pairwise comparisons between their genotypes, we can find parent-offspring and sibling pairs; the number of “recaptures” (i.e. kin-pairs found), and the sample covariates (age, sex, time, place, etc.) can be embedded in a mark-recapture framework, allowing estimation of the demographic parameters (Bravington et al., 2016, Statistical Science). CKMR is being applied to previously-intractable problems in fisheries (Bravington et al., 2016, Nature communications), game management / hunting, and marine endangered species. To get precise estimates, it is necessary to have enough “recaptures”. All else being equal, to reach a desired total of say 50 or 100 kin-pairs will require a sample size proportional to the square root of adult abundance, but the “constant” of proportionality varies substantially depending on biology and sampling stratification. While CKMR can be perfectly feasible and cost-effective for large populations (certainly up to tens of millions of adults), it does require large sample sizes (e.g. 14,000 Southern Bluefin Tuna genotypes in Bravington 2016, Nature communications) and consequent expense. Thus, to avoid disappointment it is essential to carefully design the sampling strategy using whatever prior knowledge is available, before embarking on a full-scale CKMR project. This talk shows how to tackle sampling design for CKMR using a statistical framework based the notion of pairwise Fisher information. Specifically, we present an algorithm that computes expected precision for any quantity-of-interest in a CKMR model, under any proposed sampling scheme, and without needing any simulation or estimation. We also show how “optimal” designs can be found automatically— though of course they should not be interpreted too literally. The statistical and computational ideas involved are powerful, and may have application to other mark-recapture settings.
3:40PM Validation of the Close Kin Mark Recapture Method for Estimating Abundance
  Daniel Ruzzante, Gregory R. McCracken, Brage Forland, John MacMillan, Daniela Notte, Colin Buhariwalla, Joanna E. Mills Flemming, Hans Skaug
Knowing how many individuals there are in a population is a fundamental problem in the management of exploited marine fish where abundance is traditionally estimated with catch per unit effort (CPUE) data. CPUE statistics can, however, be subject to bias and uncertainty, and are therefore often considered relatively unreliable and contentious. We compare abundance estimates (census size, Nc) in seven brook trout (Salvelinus fontinalis) populations using standard mark-recapture (MR) and the close-kin mark-recapture (CKMR) method. Our purpose is to validate close-kin mark-recapture as a method to estimate population size. CKMR is based on the principle that an individual’s genotype can be considered a “recapture” of the genotypes of each of its parents. Assuming offspring and parents are sampled independently, the number of parent-offspring pairs (POPs) genetically identified in these samples can be used to estimate abundance. We genotyped (33 microsatellites) and aged (~2400 brook trout individuals) collected over 5 consecutive years (2014-2018). Despite various sources of uncertainty, we find close agreement between standard MR abundance estimates obtained through double-pass electrofishing and CKMR estimates, which require information on age-specific fecundity, and population- and age-specific survival rates. Population sizes are estimated to range between 300
4:00PM Exploring Close-Kin Mark-Recapture As a Method for Assessing Bearded Seal Population Abundance and Status
  Brian Taras, Lori Quakenbush, Paul Conn, Jay Ver Hoef
Bearded seals (Erignathus barbatus) have a wide Arctic distribution that includes the Bering, Chukchi, and Beaufort seas. Bearded seals are associated with, and use sea ice for pupping, nursing, and molting, and they are a vital subsistence resource to coastal Alaska Native communities. In 2012, bearded seals were listed as threatened under the Endangered Species Act (ESA), not because of the population had declined, but because they would decline in response to a predicted reduction in sea ice over the next century. Their ESA listing results in higher scrutiny during stock assessments, elevating the need for better estimates of abundance and vital rates. The National Marine Fisheries Service (NMFS) and the Alaska Department of Fish and Game (ADF&G) are exploring close-kin mark-recapture (CKMR) as a better method to assess and monitor the population status of bearded seals than aerial surveys, which are difficult and costly to implement over large and remote expanses of ice covered water and because they need to be corrected for availability bias. In addition, CKMR is effective for hunted populations for which genetic samples are available and accessible through time. Since 2000, ADF&G has archived samples from >2300 harvested animals, including >1700 with accurate ages from teeth. Preliminary sample size calculations indicate available and future samples can provide reasonably precise estimates of abundance (CV=0.20-0.40), trend, and vital rates depending on model complexity and monitoring duration. Life history information such as mating strategy and regional patterns in relatedness is also obtained. The initial kin-finding effort provides insights into bearded seal life history that will inform a preliminary population dynamics model. Initial model results and additional kinship information can provide further model refinement. The results of this iterative process can greatly improve our ability to monitor the bearded seal population, particularly as more samples are collected and analyzed.
4:20PM Close-Kin Mark-Recapture in Dispersal Limited Populations
  Paul Conn
Close-kin mark-recapture often assumes that genetic data represent a random sample from the population of interest. In practice, this requires either complete mixing of individuals or even spatial sampling coverage. However, in patchy populations or species where dispersal is limited, these assumptions are unlikely to be met, particularly if sampling is spatially biased (e.g., in many harvested populations). In this talk, I use individual-based simulation to investigate the degree of negative bias in abundance estimators when naive CKMR estimators are employed on dispersal limited populations. I also describe spatially explicit CKMR estimators that can be used in their place, noting that their use will often require auxiliary data sources (e.g., resource selection functions or movement rates from telemetry studies and/or spatial indices of relative abundance).
4:40PM Using Ckmr Methods to Empirically Estimate Viability Measures in Freshwater Systems
  Scott Blankenship
Status and trends monitoring often suffers from a lack of information about basic biological parameters. Characteristics of rivers make visual surveys difficult and datasets from different life-history stages are often incompatible, which reduces the credibility of population estimates. Associating adults and juveniles together using genetics-based parentage and relatedness methods provides a powerful enhancement to existing population monitoring programs. Three applications will be presented that employed genetic-marking approaches and evaluation of closely-related individuals to estimate fundamental population metrics. The first example illustrates using genetic tagging in a mark-recapture framework for a Columbia River Basin Chinook Salmon population. Genetic mark-recapture proved capable of estimating all annual Viable Salmon Population metrics in a single wholistic sampling design. The second example describes the use of parentage and relatedness information to evaluate habitat restoration performance. While reproductive success information suggested spawning habitat space limitations, this approach provided direct evidence of habitat restoration benefits. The last example illustrates the use of kinship to estimate White Sturgeon breeder numbers in a challenging monitoring environment that is both data and monitoring tool limited.
5:00PM Panel Discussion

Organizers: Paul Conn, Jamie Sanderlin, Jay Ver Hoef
Supported by: TWS Biometrics WG, TWS Molecular Ecology WG, AFS Marine Section, AFS Genetics Section

Location: Reno-Sparks CC Date: September 30, 2019 Time: 1:10 pm - 5:20 pm