This repository provides data, methods and results used to estimate the status of the sea garfish fishery in New South Wales (Australia). This stock assessment was originally developed using a maximum likelihood approach (Broadhurst et al., 2018) and later expanded with a Bayesian approach (Kienzle et al., 2021). The Bayesian approach provides a convenient method to propagate uncertainties from parameter estimates to quantities of interest for fishery management.
The data dashboard shows the four types of data used to produce this stock assessment: (1) age from a sample of commercial catches, (2) weights from individual fish in specific intervals of age, (3) yield by year, and (4) effort by year.
In 2004/05, only three age-groups, age-groups 0, 1 and 2, were present in the catch (data dashboard). The fourth age-group, age-group 3, became more frequent over the years: age-groups 3+ made between 2 and 16% of commercial catches in the last 5 years while they represented only 0 to 6% of catches in previous years.
Yields in this fishery have been relatively stable, in average 42 +- 18 tonnes. Catches in 2009/10 were reported to be 100 tonnes, an un-usually large figure in this time series of data. Fishing effort declined sharply since 2004/05, and have slowly declined to 136 boat-days in 2019/20.
Catch per unit effort (CPUE), right-hand panel, were approx. 50 kg/boat-day at the beginning of the time series when effort was very large. After the sharp decline in effort, CPUE increased and fluctuated around 200 kg/boat-day without showing any trend.
Several models describing how mortality rates varied by age-group through time (depending on fishing effort, catchability and age-specific gear selectivity) were developed. All models assume a constant rate of mortality due to natural causes which was estimated. These mortality models have been expressed using hazard functions and converted into probabilities of dying at age from fishing using a method developed in the field of statistics called survival analysis. These age-specific probabilities of dying from fishing were combined with age data sampled from commercial catches into a likelihood function. A range of hypotheses regarding gear selectivity and natural mortality expressed into the mortality models were compared using Akaike Information Criteria (AIC) and the Bayesian Information Criteria (BIC).
The method used to analyse the data is fully described into 2 refereed articles (Broadhurst et al., 2018 and Kienzle et al., 2021).
So far, three mortality models have been estimated from the data. All models assume constant natural mortality and full selection by the fishing gear of sea garfish age-group 1 (i.e. fish between 1 and 2 years of age) and older. None of those model include catchability varying through time.
- Model 1 estimates selectivity of age-group 0-1 year old to have remained constant throughout the entire time series.
- Model 2 estimates selectivity of age-group 0-1 year old to have change in 2010/11 as a result of a management decision to increase the minimum legal mesh size.
- Model 3 estimates selectivity in 1 block like model 1 and has natural mortality fixed at 0.7 per year according to estimates found in the scientific literature.
An estimate of MSY was obtained by simulating the exploitation of this stock, using the best fishery model calibrated to the data, 200 years into the future using various constant level of fishing effort. This approach produced the typical dome shaped curve between catch and fishing effort from which we derived MSY and effort at MSY.
The parameters of all three models were estimated using a Bayesian approach. Four Monte Carlo Markov Chains (MCMCs) converged after about 20,000 iterations (MCMC convergence for model 2). Model 2 was the model most supported by the data according to BIC).
The posterior distributions of the parameters are uni-modal and symmetric: they are essentially Gaussian (comparison with Gaussian). There is a strong negative correlation between catchability (parameter 1) and natural mortality (parameter 2).
Fishing mortality rates have declined below natural mortality from 2009/10 onward. The decrease in age-group 0-1 retention, induced by the increased mesh size regulation initiated in 2006 and fully adopted in 2009/10, reduced fishing mortality on the youngest age-group to a negligible level.
Sea garfish biomass estimates have been trending upward since 2008/09: they remained above 100 tonnes since 2009/10. Biomass increased in the last 3 years and stayed above 200 tonnes in the last 2 years.
Recruitment was estimated to have varied approximately between 1 and 5 millions fish each year throughout the time series. 2008/9 had the largest recruitment estimate, presumably leading to the increase in biomass estimated in 2009/10. 2013/14 was estimated to be one of the largest recruitment but did not translate into an increase in biomass in 2014/15. Recent years has had several strong recruitments.
There start to be enough recruitment estimates (14) to begin looking at the relationship between stock and recruitment. We assumed that seagarfish age 1+ are sexually mature. The plot of recruitment estimates against spawning stock biomass (SSB) estimates, fitted with a Ricker stock-recruitment function, suggests that the Ricker model might provide useful information about the level of SSB that produces the largest amount of recruits.
The simulation study that projected this stock into the future using constant level of effort produced the result shown below. From this graphics, we estimated that this stock can produce, in average, a maximum of 78 tonnes per year when a constant effort of 700 boat-days is applied.
The average spawning stock biomass (SSB) at the level of effort, SSB at MSY, is 62 tonnes.
The average stock biomass when fishing mortality is zero is 235 tonnes (according to model 2):
As of 2021, models seem to be affected by the drifting M syndrome.