Changes in French salmon population and climate change

This was my PhD Project. It aimed at highlighting ongoing changes in French salmon population, assessing the possible implication of climate, and starting building models for forecasting. The main hypothesis was that rising temperature would modify juveniles’ growth in river and disrupt population dynamics. This revealed rather wrong. Indeed after developing two Bayesian models allowing to first reconstruct and forecast stream temperature and weight the impact of temperature ion growth compared to other covariates, we found that changes in marine conditions leading to lower fecundity by decreasing growth and survival are more susceptible to deeply impact the future dynamics of French stocks. Changes in French salmon spawners mass, length and migration timing have then been quantified. It highlighted a broad impact of changing marine conditions on all French stocks and revealed that two sea winters fish exhibited lesser changes than their one sea winter counterparts. Result also suggest a canalization of length who appeared buffered against environmental variations.

Improving models used to manage Pacific salmon stocks

This project aim at improving the management of Pacific salmon populations. The two main lines of research. The first presents the usefulness of Bayesian model averaging in assessing evidence for both positive and negative density dependence in populations and prioritize management options. It highlights the utility of Bayesian model averaging in a conservation context wherein errors in choosing the best model may have more severe consequences than errors in estimating model parameters themselves. The second line of research aims, based on mark recapture data, at estimating straying rates between salmon Evolutionary Significant Units and the extent to which populations’ characteristics as well as human management practices affect them.

Real time Incentives to manage fisheries

This project aims at maximizing sustainable yield from commercial marine fisheries using a real-time incentive system (RTI) depending on spatiotemporal dynamic fisheries models. So far this approach has been tested on simulated data and compared to more traditional management methods such as quotas. This revealed the potentially high efficiency of RTI in reducing discards, tuning effort to resources in time and space, eliminating loss of fishing opportunities due to TACs and incorporating ecosystem objectives. Further studies and consultations are in progress to allow for a potential real test of this method in Irish waters. Please visit the Facebook page!

Improving the management of Irish salmon using a Bayesian hierarchical approach

Currently salmon returns to Irish rivers are estimated through Monte Carlo analyses based on angling catch and exploitation rates to determine pre-fisheries abundances.  Estimated returns are compared against conservation limits, set at the necessary number of spawners to achieve maximum sustainable yield, defined from stock-recruitment analyses of index-rivers, to determine if a surplus is available in a river. The aim of this project is to make the assessment fully Bayesian. While doing so, were are interested in assessing the spatial scale at which fish traits and dynamics fluctuate. In particular, we are assessing the extent to which these variations cope with the description of the genetic structure of the whole Irish stock whom the hierarchy has levels including country, clusters, regions and rivers.  

Bayesian hierarchical state space models for zero inflated survey data

Estimating how the size of populations fluctuates through time and space is the cornerstone of further conservation biology. Because of time and other resources constraints, a full count of individual is most of the time impossible and scientists or stakeholders rely on partial observations to derive abundance indices. Often, data coming from surveys are zero inflated. To deal with this particular feature, zero inflated and delta GLMs approaches have been developed. We believe that these methods are inherently giving biased results when applied on species having a herding behavior or some patchy distributions. We are currently working on proving this and on the development of a Bayesian Hierarchical state space model as a way to circumvent this issue.

Trends in sharks populations from recreational fisheries mark recapture data

In this project, we aim at assessing historical population sizes of three shark species (tope shark, blue sharks and angel shark) using mark recapture data from an Irish coastal recreational fishery. This work is being performed within a Bayesian framework. It presents two main challenges. First, recapture rates are low and the variability in fishing efforts across years is suspected to be high. We thus have to include fish caught year round as the number of data would otherwise be too low. We are also working on getting some indices of fishing effort to help building robust populations dynamics models.