[Paper Review] A Bayesian approach to recover the theoretical temperature-dependent hatch date distribution from biased samples: the case of the common dolphinfish (Coryphaena hippurus)

Vicenç Moltó, Andrés Ospina‐Álvarez+3 more|ArXiv.org|Mar 30, 2020

Marine and fisheries research29 references4 citations

TL;DR

This paper presents a Bayesian hierarchical model to correct for temporal sampling bias in fishery-dependent data, enabling accurate reconstruction of the true temperature-dependent hatch date distribution for common dolphinfish (Coryphaena hippurus). By integrating otolith age data, gonadosomatic index (GSI), and satellite sea surface temperature, the model corrects for mortality and fishing season biases, revealing a more accurate spawning phenology than raw samples suggest.

ABSTRACT

Reproductive phenology, growth and mortality rates are key ecological parameters that determine population dynamics and are therefore of vital importance to stock assessment models for fisheries management. In many fish species, the spawning phenology is sensitive to environmental factors that modulate or trigger the spawning event, which differ between regions and seasons. In addition, climate change may also alter patterns of reproductive phenology at the community level. Usually, hatch-date distributions are determined back-calculating the age estimated on calcified structures from the capture date. However, these estimated distributions could be biased due to mortality processes or time spaced samplings derived from fishery. Here, we present a Bayesian approach that functions as a predictive model for the hatching date of individuals from a fishery-dependent sampling with temporal biases. We show that the shape and shift of the observed distribution is corrected. This model can be applied in fisheries with multiple cohorts, for species with a wide geographical distribution and living under contrasting environmental regimes and individuals with different life histories such as thermo-dependent growth, length-dependent mortality rates, etc.

Motivation & Objective

• To address the bias in observed hatch-date distributions caused by fishery-dependent sampling and differential mortality.
• To recover the true theoretical temperature-dependent spawning phenology of common dolphinfish (Coryphaena hippurus) from biased field data.
• To develop a transferable Bayesian framework applicable to species with multiple cohorts, variable life histories, and contrasting environmental regimes.
• To improve stock assessment models by providing more accurate reproductive phenology estimates for data-poor fisheries.
• To integrate environmental drivers (especially sea surface temperature) with biological indicators (GSI) and age-at-capture data for improved population dynamics inference.

Proposed method

• The model uses a Bayesian hierarchical approach to estimate the probability of hatching on a given date for individual fish based on otolith readings and capture dates.
• It incorporates gonadosomatic index (GSI) data as a proxy for population spawning state, linked to seasonal spawning patterns.
• Sea surface temperature (SST) data from satellite imagery are used to model thermal triggers for spawning, linking environmental conditions to reproductive timing.
• The model estimates two key parameters per cohort: the mean hatch date (μj) and the spread of the distribution (σj), which are informed by biological and environmental data.
• A likelihood function models the probability of observing a fish's age at capture, conditional on its true hatching date and growth/mortality processes.
• Markov Chain Monte Carlo (MCMC) sampling is used to infer posterior distributions of hatching date parameters, accounting for uncertainty in age estimation and sampling bias.

Experimental results

Research questions

• RQ1 How can temporal sampling bias in fishery-dependent data be corrected to recover the true hatch date distribution of a fish population?
• RQ2 To what extent does incorporating sea surface temperature (SST) improve the accuracy of reconstructed spawning phenology in dolphinfish?
• RQ3 How do mortality processes and fishing season timing distort the observed distribution of hatching dates in age-0 dolphinfish?
• RQ4 Can a Bayesian model effectively integrate otolith age data, GSI, and SST to predict the underlying spawning distribution?
• RQ5 What is the impact of environmental variability on the inferred timing and spread of spawning events in a widely distributed fish species?

Key findings

• The Bayesian model successfully corrected the observed hatch-date distribution, revealing a more accurate and less skewed spawning peak than raw data suggested.
• The model identified a primary spawning peak in July for the NW Mediterranean population, with a secondary peak in late August, consistent with thermal and GSI data.
• Incorporating SST improved model fit, showing that thermal conditions strongly influence the timing and duration of spawning.
• The model demonstrated robustness to sampling bias, particularly in correcting for the underrepresentation of early-hatched individuals due to higher mortality.
• The posterior distributions of μj and σj provided reliable uncertainty estimates, enhancing the credibility of the reconstructed spawning phenology.
• The approach is transferable to other species with multiple cohorts, variable growth, and length-dependent mortality, especially in data-poor fisheries.