Uncertainty in empirical estimates of marine larval connectivity

Abstract

Despite major advances in our capacity to measure marine larval connectivity (i.e. the pattern of transport of marine larvae from spawning to settlement sites) and the importance of these measurements for ecological and management questions, uncertainty in experimental estimates of marine larval connectivity has been given little attention. We review potential uncertainty sources in empirical larval connectivity studies and develop Bayesian statistical methods for estimating these uncertainties based on standard techniques in the mark-recapture and genetics literature. These methods are implemented in an existing R package for working with connectivity data, ConnMatTools, and applied to a number of published connectivity estimates. We find that the small sample size of collected settlers at destination sites is a dominant source of uncertainty in connectivity estimates in many published results. For example, widths of 95% CIs for relative connectivity, the value of which is necessarily between 0 and 1, exceeded 0.5 for many published connectivity results, complicating using individual results to conclude that marine populations are relatively closed or open. This “small sample size” uncertainty is significant even for studies with near-exhaustive sampling of spawners and settlers. Though largely ignored in the literature, the magnitude of this uncertainty is straightforward to assess. Better accountability of this and other uncertainties is needed in the future so that marine larval connectivity studies can fulfill their promises of providing important ecological insights and informing management questions (e.g. related to marine protected area network design, and stock structure of exploited organisms). In addition to using the statistical methods developed here, future studies should consistently evaluate and report a small number of critical factors, such as the exhaustivity of spawner and settler sampling, and the mating structure of target species in genetic studies.

Dispersal of a nearshore marine fish connects marine reserves and adjacent fished areas along an open coast

“…For example, the ~1,000 additional full‐sibling matches identified in our data set using colony are probably an artefact caused by a mismatch between the characteristics of our data and the capabilities of this methodology, specifically the behaviour of unpenalized maximum likelihood with very large samples (Almudevar & Anderson, ). Genetic results indicating high levels of self‐retention should always be coupled with a complete explanation of the methods used to generate those results (Harrison, Saenz‐Agudelo, Planes, Jones, & Berumen, 2013a, 2013b; Saenz‐Agudelo, Jones, Thorrold, & Planes, ), because the ability to accurately identify true relationships hinges on the statistical power of a given panel of genetic markers, and uncertainty involved in genetic and parentage analyses can lead to uncertainty in estimates of dispersal (Kaplan et al, ). Note that it is exceedingly difficult to discriminate full and half siblings, as well as other categories of related individuals, with genetic data (Baetscher et al, ).…”

“…Genetic results indicating high levels of self-retention should always be coupled with a complete explanation of the methods used to generate those results (Harrison, Saenz-Agudelo, Planes, Jones, & Berumen, 2013a, 2013bSaenz-Agudelo, Jones, Thorrold, & Planes, 2009), because the ability to accurately identify true relationships hinges on the statistical power of a given panel of genetic markers, and uncertainty involved in genetic and parentage analyses can lead to uncertainty in estimates of dispersal (Kaplan et al, 2016). Note that it is exceedingly difficult to discriminate full and half siblings, as well as other categories of related individuals, with genetic data .…”

The transboundary nature of the world’s exploited marine species

“…For example, shrimp (e.g., Litopenaeus stylirostris or Farfantepenaeus californiensis) along the Pacific coast of Mexico 16 and hammerhead shark (Sphyrna zygaena) in Peru 17 are managed at the species level, yet include multiple populations exploited by each country's fisheries. Moreover, recent research shows connectivity across fish stocks through larval dispersal 18 and adult migration [19][20][21] , although considerable level of uncertainty exists at different life stages 22,23 . For this analysis we only considered shared species between neighboring EEZs, rather than the species' extended distribution (e.g., we did not include the high seas).…”

MPA network design based on graph theory and emergent properties of larval dispersal

“…All analyses were performed using R v.3.6.0 (R Core Team 2019). Matrix, network graph and community detection analyses were performed using R packages 'igraph' v.1.2.4.1 and 'ConnMatTools' v.0.3.3 (Csardi & Nepusz 2006, Kaplan et al 2017. Network visualisations were made with R packages 'mapdata' v.2.3.0, 'ggplot2' v.3.2.1 and 'ggraph' v.2.0.0 (Chang 2012, Wickham 2016).…”