fishblicc Assessment Template

Seatrout Stock

Author

Paul A Medley

Published

March 28, 2025

Abstract

The article reports a stock assessment of seatrout (Cynoscion virescens) using a length-based catch curve method (fishblicc). This assessment was based on old (adapted) data presented to the CRFM 2007 Scientific Meeting, and as a result the assessment was for illustrative purposes only.

Introduction

Cynoscion virescens (Cuvier, 1830), also known as Guyana seatrout or green weakfish, is caught widely from Brazil to Panama through the Caribbean. There is a significant small scale fishery on this species in Guyana. A gillnet fishery catches the fish at a relatively large size compared to the trawl fisheries, such as the shrimp trawl, that has a bycatch of seatrout.

Data were collected 2000-2007 from length sampling of landings (Figure 1).

In 2007 at the Caribbean Regional Fisheries Mechanism Scientific Meeting (CRFM 2007), an attempt was made to assess Guyana seatrout using length frequency data. This was broadly unsuccessful because of the difficulties in fitting models to the length frequency data, the main problems being the lack of a consistent time series of data and the inability to account for the different gear selectivities.

Seatrout is predominantly caught in gillnet (GN), with some bycatch catch in trawl (TRAWL).

Figure 1: Length frequency data 2000-03 for Guyana seatrout (*Cynoscion virescens*) broken down by main gear type.

The analysis of multigear length frequency data can be conducted using a new Bayesian length-based catch curve method (Medley 2023). This method is a data-limited approach that estimates mortality from length frequency data together with relative catch in numbers taken by each gear. The method uses proposed selectivity functions that explain the observations and estimate the spawning potential ratio (SPR) to determine stock status. Being data limited, the model still requires input of important life history parameters in the form of Bayesian priors. The Bayesian fit helps explore the model uncertainties.

Life History Parameters

Information on Cynoscion virescens was very limited (Santos and Vianna 2017). The reported maximum length of C. virescens was 115cm (Froese and Pauly 2023), but there was no information on the \(L_\infty\). No growth parameters have been estimated for the Guyana stock. (CRFM 2007) used a \(L_\infty\) of 96cm. Given the observed lengths, any \(L_\infty\) in the range of 85-110 was possible.

Length-weight model parameters that have been estimated for this species include:

a = 0.005 b= 3.054 Brazil demersal trawl (Felizola Freire, Rodrigues Alves Rocha, and Lemos Souza 2009)
a = 0.00446 b= 3.100 Brazil demersal trawl (Passos et al. 2012)
a = 0.0108 b = 2.860 Brazil using different fishing gears (Viana et al. 2016)

All estimates from Brazil were based on smaller individuals not much larger than 30cm. Given the current uncertainty, the values a = 0.005 b = 3.054 were used as these are close to default values and in the middle of the range given.

The growth parameters used in the (CRFM 2007) stock assessment attempt were K = 0.3938, \(L_\infty\) = 95.8, M= 0.59 suggesting a default M/K of around 1.5, which is the default ratio suggested by (Kenchington 2013) and others. No justification was provided for the values used.

The length at maturity can be derived from the life-history invariant equation (Prince et al. 2014):

\[ L_m = {{b L_\infty} \over {b + M/K}} \tag{1}\]

Base Model

It might be expected that gillnet selectivity would follow a normal distribution, while the trawl selectivity would be logistic.

Note that catch numbers between gears are assumed to be proportional to the length frequency sample size. This would need to be set to a better estimate. The relative catch numbers is required for each gear.

Figure 2: Expected length frequency based on the prior parameter means (line) plotted over the observed frequency (histogram). If necessary these can be adjusted using the `blip_sel` function.

Inspecting the expected length frequency based on the priors (Figure 2), the trawl frequency was very different to the observed. This can be adjusted manually if the prior is too different as it provides the start point for the fit.

This model provided a reasonable fit to the data (Figure 3). The parameter estimates are reasonable and estimated SPR is relatively high (Table 1).

Figure 3: Observed and expected frequency for the maximum posterior density fit with mixed selectivity, constant natural mortality and logistic trawl selectivity.

Table 1: Model parameter estimates based on maximum posterior density fit.

Parameter	Value
Linf	94.242
Galpha	121.153
Mk	2.407
Fk[1]	0.553
Fk[2]	0.015
Sm[1]	77.272
Sm[2]	0.005
Sm[3]	0.027
Sm[4]	25.865
Sm[5]	0.731
NB_phi	112.456
Gbeta	1.286
lp__	-426.353
SPR[1]	0.799
B_B0[1]	0.839
YPR[1]	33.985
YPR[2]	3.057

The standardised residual plot can be used to identify outliers (Figure 4). Outliers are particularly important when using normal selectivity functions which predict very low probabilities of capture in their tails.

Figure 4: Standardised residuals for the logistic selectivity mixture model.

MCMC Fit

The base model was fitted using Stan MCMC algorithm (mc-stan.org). The MCMC converged (Table 2) and the model fitted the data well (Figure 5).

It can be seen that the observations were informative on natural mortality (Figure 6), favouring higher estimates. This was mostly the result of the length data from the trawl gear where the range in lengths allows estimation of the mortality over a wide range of ages and the fishing mortality from this gear was estimated to be very low.

Figure 5: Expected length frequency based on the MCMC posterior parameters (line) for the base model plotted over the observed frequency (histogram).

Table 2: MCMC parameter estimates with the effective sample size and MCMC convergence statistic (Rhat). Effective sample size needs to be 500 or more for reliable estimates. A Rhat value close to one indicates convergence.

Parameter	Mean	SD	N (eff)	Rhat
Linf	94.206	0.942	1,736	0.999
Galpha	123.739	27.293	2,026	0.999
Mk	2.413	0.123	880	1.003
Fk[1]	0.533	0.293	655	1.005
Fk[2]	0.015	0.008	678	1.004
Sm[1]	77.236	0.567	572	1.010
Sm[2]	0.005	0.000	807	1.009
Sm[3]	0.027	0.003	706	1.012
Sm[4]	25.880	0.335	1,219	1.001
Sm[5]	0.731	0.079	1,249	1.000
NB_phi	117.237	47.617	1,289	1.002
Gbeta	1.314	0.290	2,013	0.999
SPR[1]	0.812	0.089	619	1.005
lp__	-427.339	2.413	757	1.000

Figure 6: MCMC parameter plots indicating relative support for each parameter from the data. The parameters are drawn from standard normals (mean=0, sd=1). If no data were present, 95% of the lines would be spread in a normal density pattern within the vertical dotted lines, so departures from this indicate estimation support from the data.

Figure 7: Spawner potential ratio probability density posterior estimated from the MCMC.

There is a large difference between the gillnet and trawl selectivities (Figure 8).

Figure 8: Estimated selectivity with 80%CI for the four gear types.

Discussion

Discussion or conclusion of results.

References

CRFM. 2007. “Report of Third Scientific Meeting – Kingstown, St. Vincent and the Grenadines, 17-26 July 2007 - Fishery Management Advisory Summaries.” Vol. 2.

Felizola Freire, K. Meirelles, G. Rodrigues Alves Rocha, and I. Lemos Souza. 2009. “Length-Weight Relationships for Fishes Caught by Shrimp Trawl in Southern Bahia, Brazil.” Journal of Applied Ichthyology 25 (3): 356–57. https://doi.org/10.1111/j.1439-0426.2009.01220.x.

Froese, R., and D. Pauly, eds. 2023. “FishBase.” World Wide Web electronic publication. www.fishbase.org.

Kenchington, Trevor J. 2013. “Natural Mortality Estimators for Information-Limited Fisheries.” Fish and Fisheries 15 (4): 533–62. https://doi.org/10.1111/faf.12027.

Medley, Paul A. H. 2023. “Fishblicc: Bayesian Length Interval Catch Curve.” https://github.com/PaulAHMedley/fishblicc.

Passos, A. C., R. Schwarz, B. F. C. Cartagena, A. S. Garcia, and H. L. Spach. 2012. “Weight-Length Relationship of 63 Demersal Fishes on the Shallow Coast of Paraná, Brazil.” Journal of Applied Ichthyology 28 (5): 845–47. https://doi.org/10.1111/j.1439-0426.2012.01973.x.

Prince, Jeremy, Adrian Hordyk, Sarah R. Valencia, Neil Loneragan, and Keith Sainsbury. 2014. “Revisiting the Concept of BevertonHolt Life-History Invariants with the Aim of Informing Data-Poor Fisheries Assessment.” ICES Journal of Marine Science 72 (1): 194–203. https://doi.org/10.1093/icesjms/fsu011.

Santos, Sérgio Ricardo, and Marcelo Vianna. 2017. “Scientometric Analysis of the Fisheries Science for the Species ofCynoscion(Sciaenidae: Perciformes) from the Western Atlantic, with Emphasis in the Comparison of the North American and Brazilian Fisheries Catch Data.” Reviews in Fisheries Science & Aquaculture 26 (1): 55–69. https://doi.org/10.1080/23308249.2017.1337078.

Viana, A. P., F. Lucena-Frédou, F. Ménard, T. Frédou, V. Ferreira, A. S. Lira, and F. Le Loc’h. 2016. “Lengthweight Relations of 70 Fish Species from Tropical Coastal Region of Pernambuco, Northeast Brazil.” Acta Ichthyologica Et Piscatoria 46 (3): 271–77. https://doi.org/10.3750/aip2016.46.3.12.