A Bayesian statespace markrecapture model to estimate exploitation rates in mixed-stock fisheries (original) (raw)
Related papers
ICES Journal of Marine Science, 2014
ABSTRACT We developed a hierarchical Bayesian integrated life cycle model for Atlantic salmon that improves on the stock assessment approach currently used by ICES and provides some interesting insights about the population dynamics of a stock assemblage. The model is applied to the salmon stocks in eastern Scotland. It assimilates a 40-year (1971–2010) time-series of data compiled by ICES, including the catches in the distant water fisheries at Faroes and West Greenland and estimates of returning fish abundance. Our model offers major improvements in terms of statistical methodology for A. salmon stock assessment. Uncertainty about inferences is readily quantified in the form of Bayesian posterior distributions for parameters and abundance at all life stages, and the model could be adapted to provide projections based on the uncertainty derived from the estimation phase. The approach offers flexibility to improve the ecological realism of the model. It allows the introduction of density dependence in the egg-to-smolt transition, which is not considered in the current ICES assessment method. The results show that this modifies the inferences on the temporal dynamics of the post-smolt marine survival. In particular, the overall decrease in the marine survival between 1971 and 2010 and the sharp decline around 1988–1990 are dampened when density dependence is considered. The return rates of smolts as two-sea-winter (2SW) fish has declined in a higher proportion than return rates as one-sea-winter (1SW) fish. Our results indicate that this can be explained either by an increase in the proportion maturing as 1SW fish or by an increase in the mortality rate at sea of 2SW fish, but the data used in our analyses do not allow the likelihood of these two hypotheses to be gauged.
Canadian Journal of Fisheries and Aquatic Sciences, 2007
Models for fluctuations in size of fish stocks must include parameters that describe expected dynamics, as well as stochastic influences. In addition, reliable population projections also require assessments about the uncertainties in estimates of vital parameters. Here we develop an age-structured model of population dynamics based on catchat-age data and indices of abundance in which the natural and fishing mortality are separated in a Bayesian state-space model. Markov chain Monte Carlo methods are used to fit the model to the data. The model is fitted to a data set of 19 years for Northeast Arctic cod (Gadus morhua). By simulations of the fitted model we show that the model captures the dynamical pattern of natural mortality adequately, whereas the absolute size of natural mortality is difficult to estimate. Access to long time series of high-quality data are necessary for obtaining precise estimates of all the parameters in the model, but some parameters cannot be estimated without including some prior information. Nevertheless, our model demonstrates that temporal variability in natural mortality strongly affects perceived variability in stock sizes. Thus, using estimation procedures that neglect temporal fluctuations in natural mortality may therefore give biased estimates of fluctuations in fish stock sizes.
The goal of our probabilistic model is to provide a formal framework which uses all relevant information about the juvenile population and ex- presses this knowledge in terms of probability distributions for relevant pop- ulation parameters. One important focus is the estimation and prediction of the size of the smolt run. For this task, biological understanding of the life- cycle of salmon, as well as measurements of parr and smolt abundances are all relevant information. The probability model which combines these sources of information can be divided to three sub models, which can be developed independently: 1) a parr abundance model that describes the rela- tionship between the true abundance of parr and corresponding measure- ments, 2) a smolt abundance model that connects the true abundance of smolts to the corresponding mark-recapture measurements and 3) a popu- lation model, based on the knowledge about the life-cycle of salmon, de- scribing the linkages between parr and smol...
Canadian Journal of Fisheries and Aquatic Sciences, 2007
Models for fluctuations in size of fish stocks must include parameters that describe expected dynamics, as well as stochastic influences. In addition, reliable population projections also require assessments about the uncertainties in estimates of vital parameters. Here we develop an age-structured model of population dynamics based on catchat-age data and indices of abundance in which the natural and fishing mortality are separated in a Bayesian state-space model. Markov chain Monte Carlo methods are used to fit the model to the data. The model is fitted to a data set of 19 years for Northeast Arctic cod (Gadus morhua). By simulations of the fitted model we show that the model captures the dynamical pattern of natural mortality adequately, whereas the absolute size of natural mortality is difficult to estimate. Access to long time series of high-quality data are necessary for obtaining precise estimates of all the parameters in the model, but some parameters cannot be estimated without including some prior information. Nevertheless, our model demonstrates that temporal variability in natural mortality strongly affects perceived variability in stock sizes. Thus, using estimation procedures that neglect temporal fluctuations in natural mortality may therefore give biased estimates of fluctuations in fish stock sizes.
Environmental and Ecological Statistics, 2006
A general model is developed to examine the patterns of the regional movement of tagged and released fish from mark-recapture experiments. It is a stochastic model that incorporates fishing mortality, natural mortality, fish movement, tag-shedding, and different rates of reporting. A likelihood function is constructed for estimating its parameters. We used this model to analyze data on the Pacific halibut from mark-recapture experiments conducted by the International Pacific Halibut Commission (IPHC), with a total of 36,058 releases from 1982 to 1986 and 5,826 recoveries from 1982 to 2000. We estimated their rates of movement among IPHC management areas, along with their instantaneous rates of natural and fishing mortalities. Our analysis revealed that fish movement was not significant among areas, with a resident probability of > 0.92. This lends support to the IPHC catch-at-age stock assessment model (which has no built-in movement components). The estimated instantaneous rate of natural mortality (0.198 year −1 ) lies between that assumed in all IPHC stock assessments before 1998 (0.20 year −1 ) and that from 1999 onwards (0.15 year −1 ). The estimates of the instantaneous rates of fishing mortality were consistent with those from the IPHC stock assessment model.
Canadian Journal of Fisheries and Aquatic Sciences, 1999
The large variability in Baltic cod (Gadus morhua) recruitment has been attributed both to environmental factors dependent on the inflow of saline water (oxygen and salinity in spawning deeps) and to the size of the spawning stock. Due to the complex interactions between hydrographic and biological processes, future recruitment levels of cod will remain highly uncertain and increase uncertainties in stock predictions and management advice. We assessed the effects of the exploitation level and mesh size used by a trawl fishery on some variables of management interest under different environmental conditions. The modeling consisted of three separate steps: (i) modeling of selectivity, (ii) estimation of uncertainties by Monte Carlo simulations, and (iii) decision analysis by Bayesian influence diagrams, focusing on the structural uncertainties and model selection. Realistic assumptions of environmental conditions and present fishing mortality rates suggest that the current Baltic cod fishery is unsustainable. We use our approach to identify robust management measures that reduce the risk of overfishing and the sensitivity to management information. The value of information analysis demonstrates the advantages of a larger mesh size as a management measure.
Experiences in Bayesian Inference in Baltic Salmon Management
Statistical Science, 2014
We review a success story regarding Bayesian inference in fisheries management in the Baltic Sea. The management of salmon fisheries is currently based on the results of a complex Bayesian population dynamic model, and managers and stakeholders use the probabilities in their discussions. We also discuss the technical and human challenges in using Bayesian modeling to give practical advice to the public and to government officials and suggest future areas in which it can be applied. In particular, large databases in fisheries science offer flexible ways to use hierarchical models to learn the population dynamics parameters for those by-catch species that do not have similar large stock-specific data sets like those that exist for many target species. This information is required if we are to understand the future ecosystem risks of fisheries.
Canadian Journal of Fisheries and Aquatic Sciences, 2009
Multistate mark-recapture models can be used to model migration through stratification of the study area into states (location). However, the incorporation of both tag loss and reporting rates is new to the multistate paradigm. We develop a migration model for fish that incorporates tag loss and reporting rates but has as its primary purpose the modelling of exploitation and natural mortality rates. This model is applied to a 2000-2004 yellowtail flounder (Limanda ferruginea) tagging study on the Grand Bank of Newfoundland, Canada. We found that exploitation rates varied over both location and years, ranging from 0.000 to 0.047. Migration into the centre of the Grand Bank (state 2) was three times higher than migration out. The estimate of the instantaneous annual natural mortality rate was 0.256, which is equivalent to an annual survival rate of 0.880. We describe how these mortality estimates will be quite valuable in specifying an assessment model for this stock.
Unbiased Methods for Calculating Mortality in Mark-Selective Fisheries Models for Ocean Salmon
North American Journal of Fisheries Management, 2013
Oncorhynchus kisutch and Chinook Salmon O. tshawytscha on the west coast of North America. Mark-selective fisheries allow anglers to keep legal-size Coho Salmon or Chinook Salmon with a missing adipose fin (typically hatchery fish) and require the release of those with an adipose fin (unmarked fish, which are usually wild fish). The objective of MSF is to provide meaningful fisheries on abundant stocks of hatchery salmon while reducing the impact on wild (unmarked) salmon stocks. As has been previously shown, the model currently used in the Pacific Fishery Management Council's preseason planning process to project mortalities for proposed Coho Salmon and Chinook Salmon fisheries underestimates the number of unmarked salmon mortalities occurring in MSF and concurrent nonselective fisheries. We propose equations that provide unbiased estimates of salmon mortalities that occur in these fisheries due to the release of fish. The performance of the proposed methods is evaluated and compared to the current methods using a simulation model. The methods are shown to provide unbiased calculations of total mortalities for unmarked salmon in both mark-selective and concurrent nonselective fisheries. The unbiased methods are able to incorporate different release-mortality and mark-recognition rates for the fisheries modeled.
ICES Journal of Marine Science, 2016
Knowledge of current fishing mortality rates is an important prerequisite for formulating management plans for the recovery of threatened stocks. We present a method for estimating migration and fishing mortality rates for anadromous fishes that combines tag return data from commercial and recreational fisheries with expert opinion in a Bayesian framework. By integrating diverse sources of information and allowing for missing data, this approach may be particularly applicable in data-limited situations.Wild populations of anadromous sea trout (Salmo trutta) in the northern Baltic Sea have undergone severe declines, with the loss of many populations. The contribution of fisheries to this decline has not been quantified, but is thought to be significant. We apply the Bayesian mark-recapture model to two reared sea trout stocks from the Finnish Isojoki and Lestijoki Rivers. Over the study period (1987–2012), the total harvest rate was estimated to average 0.82 y–1 for the Isojoki River...