Search

Chapter 5 in the book Science Communication in South Africa., Published by Africa Minds.

The Statistical Catch-at-Age (SCAA) assessment methodology applied to the Greenland halibut resource in 2017 is updated to take further data now available for the years 2017 to 2019 into account. Results are reported for a Base Case corresponding to that considered in 2017, and some sensitivities related to uncertainties about some recent catch-at-age data. Apart from a slight downward shift in the overall biomass scale, results are essentially unchanged from those in 2017, and the sensitivity runs show little difference from the Base Case. There has been a recent slight downward trend in exploitable biomass, but this is likely to reverse quite soon given that the estimates of incoming recruitment are of above average strength. New data and recent resource trends are consistent with predictions made in 2017 when a revised management procedure for Greenland halibut was adopted.

This document presents the proposals from the poaching task group for time series of annual poaching tonnages from the resource as a whole for use in population models.

This document provides a summary of past recovery targets for west coast rock lobster, along with recent estimates of resource trends

This document reports updated 2019 horse mackerel assessments, along with some initial projections. The updated assessments provide indications of improved resource status, with consequent implications for management advice. However, before final calculations to facilitate development of such advice are specified and conducted, the DWG needs first to give attention to whether this updated information as yet provides a sufficiently strong scientific basis to allow firm conclusions to be drawn concerning such improvement in status

This document explores why the horse-mackerel survey data require changes in selectivity over time.

This document reports further updated 2019 horse mackerel assessments, along with constant catch projections for the base case model, in accordance with requests made the Demersal Working Group.

This document provides the detailed and comprehensive specifications for the 2019 Reference Case assessment model for the South African hake resource, including the algebraic specifications for the assessment model, tables listing the input data and key assessment results for the Reference Case. Note that the RC results presented here do not correspond to the most recent assessment results (as the specification document has not been updated since the last assessment update). Results for the most recent update for the 2019 RC can be viewed in FISHERIES/2019/OCT/SWG-DEM/22rev.

Updated results are given for an approach which simultaneously models the “old” and the “new” databases as recommended by the Panel for the 2018 International Fisheries Stock Assessment Workshop, in which poaching data for 2018 is now included. These results are compared to those obtained previously. For cases where the “old” database is upweighted the 2018 poaching indices estimated for both the northern Super-areas (3-7) and for the southern Super-area 8+ show statistically significant decreases compared to 2017. A new smoothing method is put forward for converting the output poaching series obtained to the input series used in the assessments.

An updated GLMM standardisation of trotline toothfish CPUE data is presented which includes new data for the 2018 “fishing”-year. After a marked decrease in CPUE from 2015 to 2017, there has been a notable increase in the 2018 index, which is close to the larger estimated values for the first four years (2010-2013) of the series.

This document details the current 2018 Reference Case Operating Model which is to be proposed to be used for simulation testing of a new OMP for South Coast rock lobster.

The analyses undertaken to provide a scientific recommendation for the 2019 sardine TAC and TAB, under Exceptional Circumstances were subject to a number of more serious uncertainties than would normally apply. Given the critically low level of the sardine resource, care should be taken to either recommend catch limits that are robust to these uncertainties, or to undertake further research to minimise such uncertainty. One key uncertainty is how survey (and commercial) length frequencies should be weighted given that concerns frequently arise that one or other (or both) may be unrepresentative of the population length structure. Future recruitment to this short‐lived resource is another key uncertainty. While the primary uncertainty surrounding the weight‐at‐length has been alleviated, and the growth curve adjusted, some uncertainty surrounding the choice of growth curve remains. Finally, under Exceptional Circumstances, the decision‐making process involves a greater degree of subjectivity. While the Small Pelagic Scientific Working Group aimed at making this as objective as possible to recommend 2019 sardine catch limits, that approach could possibly be further improved. This document is aimed at providing current research and analyses to assist discussion on the key questions to the panel at the MARAM International Stock Assessment Workshop.

This document provides the most recent updates of all five super-area assessments for west coast rock lobster, as well as projections taking into account recommendations made at the December 2018 International Workshop.

A list is put forward of suggestions and criteria to provide a basis for developing a sardine directed TAC and by-catch TAB for 2019

OMP-based TAC recommendations for rock lobster TACs for the Tristan group of islands for the 2019/20 season are summarized.

This document examines the shifts in catch-at-length proportions for the Tristan group rock lobster fisheries.

Brandão and Butterworth (2019a) proposed a simple Candidate Management Procedure (CMP) control rule based upon recent trends in CPUE by to provide future TAC recommendations for the Prince Edward Islands toothfish resource. Several suggestions made by a Task Team that evaluated the performance of this CMP are addressed in this paper. In particular, the CMP considered is tuned to target median final depletion levels of 30%, 40% and 50% under OM10 (which is the Operating Model (OM) with a better fit to the last two low trotline CPUE indices). These three CMPs are referred to as CMP30, CMP40 and CMP50 respectively. These CMPs are able to secure an increase in the TAC without adversely affecting the status of the resource under most OMs, especially for the more conservative CMP50. However, the performance under OM17, which addresses the concern with the poor fit to the trotline CPUE data in the last two years by assuming an increased tag loss rate, is not satisfactory. In order to tune these CMPs to react satisfactorily under the scenario assumed by OM17 would heavily penalise TACs under the scenarios reflected by the other OMs. A CMP which incorporates other information, to hopefully be able in some sense to better distinguish what the status of the resource is, might perform better under all OMs. Deterministic projections of mean length and the cumulative number of tags recaptured are examined for this purpose, and are suggestive that incorporation of these data could improve CMP performance.

Genetic data, while compatible with the hypothesis of a single M. paradoxus stock off Namibia and South Africa, does not exclude the possibility of two M. paradoxus stocks with a soft boundary somewhere between Saldahna Bay and the Orange River. The same conclusion follows from the survey catch-at-length distribution information available. Thus, while a single M. paradoxus stock hypothesis likely remains the most plausible, the possibility of two stocks of this species (with no need for their joint management by South Africa and Namibia) remains. Some suggestions are made for further work in the short term which may throw more light on this issue.

This document provides the full specification and rationale for the south coast rock lobster OMP 2019

Results are reported for three further CMPs for the south coast rock lobster. Simulations now take the 2017 catches and CPUE values into account.

The length of the additional closed season which would need to be applied to the squid jig commercial sector to keep effort within the TAE has been re-calculated taking into account that the TAE recommendation for the 2019/20 fishing season by the Squid Scientific Working Group is 295 000 person-days, and further information about effort levels applied in the fishery over recent years. These calculations suggest that the additional closed season should remain of three months duration.

The longline CPUE series for Inaccessible and Gough islands are GLMM standardised through to 20181. For Nightingale, the fishery was closed for the 2011 season and catches were set at precautionary levels for the 2011 and 2012 seasons. The Nightingale GLMM model thus excludes 2011 and 2012, although it now includes 2013-2018 for this island. Year, month, area, trap-type, soak time, depth and year-area interactions are treated as fixed effects, and year-month interactions treated as a random effect. The standardised CPUEs for Inaccessible and Nightingale continue to be optimistic, although both show slight decreases since the previous season. For Gough, the standardised CPUE value for the current season is higher than that of the previous season.

Stock-recruitment relationships are developed and included when forward-projecting the west coast rock lobster population. This results in appreciably lower extents of resource recovery under the current TAC than when this effect is ignored.

The Reference Case Ricker model is updated with commercial and survey data from 2018 and 2019. The updated data points lie within the bounds predicted by OMP2018 projections, except for the 2018 proportion of M. capensis in the South Coast offshore trawl fishery catch, which is markedly higher than predicted. The updated assessment results are compatible with the previous Reference Case results and suggest an increase in spawning biomass since 2017 for both species.

Specifications and projection results for the 2018 Operational Management Procedure used for setting South African hake Total Allowable Catches are provided along with various background information, including details of the metarule processes.

This document presents stochastic projections for the west coast rock lobster resource for two variants of future recruitment.

Results for a total of 14 robustness tests re-run for the updated 2019 Reference Case Operating Model (OM) are contrasted with results for the same robustness tests conducted for the 2018 Reference Case OM. Three of the robustness tests were also run for the 2019 Beverton-Holt central-year OMs (which estimate a heavily depleted M. capensis population) to check whether performance of OMP2018 is satisfactory even under those circumstances. Results for this selection of robustness tests indicate that the performance of OMP2018 remains adequate.

Brandão and Butterworth (2019a) proposed a simple target-based Candidate Management Procedure (CMP) control rule based upon recent average values of CPUE to provide future TAC recommendations for the Prince Edward Islands toothfish resource. Several suggestions made by a Task Team that evaluated the performance of this CMP were addressed by Brandão and Butterworth (2019b). Although the proposed CMP performed satisfactorily under most OMs, it was unable to react appropriately especially under the scenario assumed by OM17, which addresses the concern with the poor fit to the trotline CPUE data for the last two years by assuming an increased tag loss rate. Further CMPs are investigated in this paper to try to achieve better performance. These CMPs include changes to the form of the CMP proposed by Brandão and Butterworth (2019b), as well as the incorporation of the cumulative number of recaptured tags. These CMPs are able to secure an increase in the TAC without adversely affecting the status of the resource under OM01 and OM10. However, their performances under OM17 are still not satisfactory, although CMP(dep t) (which incorporates a time-dependent target) performs better than the other CMPs in so far as the median final spawning biomass remains near the current (2017) abundance level and at about BMSY. However, all CMPs still fall well below the target value of median final depletion under OM17. CMP(slope) (based on the recent trend in CPUE) is the only CMP to perform satisfactorily under OM18. Incorporation of the cumulative numbers of tag returns in the CMP has not been successful in improving its performance for OMs which show problematic resource trends such as OM17 and OM18. However, this might mean that further adjustment of this CMP is needed rather than that the tag information is not useful in discriminating amongst the various scenarios covered by the OMs. Further investigation/adjustment of these CMPs is needed.

The GLMM (General Linear Mixed Model) reported by Brandão and Butterworth (2019) to standardise toothfish CPUE for trotlines is rerun under assumptions of a different catchability coefficient for the Koryo Maru for the period 2016 to 2017 and for the period 2016 to 2018. Very little change is estimated, and the pattern of the standardised CPUE series is scarcely changed.

The 2018 Reference Set of Operating Models used for testing OMP2018 was not fully endorsed by the Panel for the 2018 International Stock Assessment workshop. This document provides results for an updated RS that take account of the recommendations made by the Panel to address the reasons why they did not fully endorse these OMs, and projects them forward under the rules of OMP2018. There are some substantial changes to certain of the Beverton-Holt models (primarily as a result of finding improved minima for model fits rather than of correcting other of the issues identified by the Panel). These reflect more instances of an M. capensis population currently below its MSY level, but this status is not well determined in terms of the likelihood; furthermore, arguably these results are less plausible than those for the Ricker model, the fits for which have an appreciably better likelihood and reflect this population to be above its MSY level at present. This uncertainty about the status of the M. capensis population notwithstanding, projections for the updated Reference Set under OMP2018 indicate that the performance of this OMP remains fully adequate.

The sole resource is modelled by a dynamic Schaefer production model which allows for a drop in the value of the intrinsic growth rate parameter from 2000 onwards. The model is fit to the available CPUE and survey abundance indices. These data are not sufficiently informative to be able to distinguish amongst fairly wide ranges of pre- and post-2000 intrinsic growth rate parameters. Nevertheless, all suggest that the sole resource has never been substantially depleted (being well above its MSY level), and furthermore that the current replacement yield and MSY are reasonably robustly estimated in the ranges of 300-400 and 500-650 mt respectively.

Results are given for an approach which simultaneously models the “old” and the “new” databases as recommended by the Panel for the 2018 International Fisheries Stock Assessment Workshop. These results are compared to two previous GLM methods reported by Brandão and Butterworth (2018a); a) a GLM applied to the “new” compliance database, and b) a GLM method applied to the “old” database in which estimates of relative effort efficiencies obtained from the “new” database are used to link effort to the number of confiscations (“old”-linked). By simultaneously modelling the “old” and the “new” databases, information on the trend in poaching, as well as on the relative efficiencies of different effort types from the “new” database, influence the estimates of poaching trends obtained from the combined model. However, due to the fact that the “old” database is a summation of operations which produced a number of instances of confiscations (including cases of zero confiscations), while the “new” database reports individual incidents with non-zero confiscations, undue weight is being given to the information available in the “new” database if data from each are equally weighted. We therefore advocate that an analysis which upweights the data in the “old” database should be used in the model which combines use of both the ”old” and “new” database values, and that this should serve as the primary basis for inferring poaching trends.

This document reports yet further updated 2019 horse mackerel assessments, along with constant midwater catch projections for the base case model, in accordance with requests made by the Demersal Working Group. These include differentiating survey selectivity on the basis of old and new gears, and exploring alternative values for natural mortality at age.

The GLMM model to estimate the impact of fishing on penguins at the St Croix and Bird Island colonies is updated given further data. For trip length and duration data, there is no evidence to suggest a statistically significant impact of fishery on the penguin populations on these two islands, unless the influential 2016 data points are excluded from the analyses, in which case a significant impact would be indicated for the St Croix forage trip length variable. However, no a priori independent reason to treat 2016 data differently has been put forward to justify such an exclusion. If the maximum distance from the colony is considered, a significant (at the 5% level) negative impact of the fishery is indicated for St Croix.

A summary is provided of issues that need to be considered in planning further analyses needed of penguin data that might provide a basis for penguin-related management recommendations. These issues include the data (and particularly co-variate values) available, the conduct of two- and one-stage estimation approaches, and the relation of possible response variable changes in response to island closures to penguin population growth rates.

The Powerboat CPUE series for Tristan was last updated in 2018 (MARAM/Tristan/2018/MAY/05) and took data for the 1994-20171 seasons into account. This GLM incorporated a method whereby the full GLM for Tristan can be rescaled to take into account the overall fishing efficiency changes as informed by the data on fishermen’s names (available for the 2005-2007 and 2013-2018 seasons only). This document updates this GLM analysis by incorporating data from the most recent 2018 season. A very slight decline in the 2018 CPUE relative to 2017 is evident.

For stochastic future recruitment and for annual catch levels of 0, 200 and 400 tonnes, the median of the spawning biomass estimates for the New Base case model drops initially, but returns to its current level within a decade. For higher catch levels of 543 and 643 tonnes, the median of the spawning biomass estimates return to current levels within only 15 and 20 years respectively. However, for variants of this model that attempt to improve the fit to the trotline CPUE or which adjust the projected recruitments to reflect a possible regime shift, it is only annual catches of 200 tonnes that maintain the spawning biomass above its current level at the end of a 20 year period. This suggests that consideration should perhaps be given to recommending some decrease in the present annual TAC of 543 tonnes.

OMPs have recently been accepted as the basis to recommend rock lobster TACs for all four islands of the Tristan da Cunha Group. The application of these OMPs using the most recent standardized CPUE data for input result in the following TACs. The OMP for Tristan recommends the TAC for 2019 at 120 MT (2018 TAC 120 MT). The OMP for Nightingale recommends the TAC for 2019 at 87 MT (2018 TAC 83 MT). The OMP for Inaccessible recommends the TAC for 2019 at 93 MT (2018 TAC 89 MT). The OMP for Gough recommends the TAC for 2019 at 105 MT (2018 TAC 111 MT).

The Operating Models (OMs) presented in Brandão and Butterworth (2019) for the toothfish resource in the Prince Edward Islands region are used for initial trials of a Candidate Management Procedure (CMP) which could provide future TAC recommendations for this resource. The performance of a simple CMP control rule based upon recent trends in CPUE is investigated. This CMP is able to secure an increase in the TAC without adversely affecting the status of the resource under most OMs. However, the performance under OM17, which addresses the concern with the poor fit to the trotline CPUE data in the last two years, is not satisfactory. In order to tune this CMP to react satisfactorily under the scenario assumed by OM17 would heavily penalise TACs under the scenarios reflected by the other OMs. A CMP which incorporates other information, to hopefully be able in some sense to better distinguish what the status of the resource is, might perform better under all OMs.

Projected TAC outputs for OMP-2014 at that time are compared to those that subsequently eventuated. A minor variant of OMP-2014 is applied to the adjusted base case SCRL operating model for a number of values for the central parameter CPUE_targ.

A Bayesian “Replacement Yield” model is applied to the total annual catches and the survey abundance estimates for the South African kingklip resource off the South coast and that off the West coast over the 1986 to 2019 period. A posterior median replacement yield (RY) of 1 152 tonnes is estimated for the South coast and of 3 218 tonnes for the West coast; these values are suggested as upper bounds for the catch limit recommendations. Setting the catch limit at the 25th percentile of the posterior distribution results in values of 919 t and 2 986 t for the South and West coasts respectively. The corresponding posterior median rates of increase over the last five years are estimated at about -0.7% and 2% for the South and West coasts respectively.

Conditioning results for the Reference Set of Operating Models identified at a Task Team meeting and described in Brandão et al. (2019) are given. The assumptions that show more marked differences in the conditioning results from those of the Base case OM01 are changes in natural mortality (OM02 and OM03), up-weighting all CPUE data by a factor of 10 (OM10), a change in recruitment residual standard deviation (OM11), alternative von Bertalanffy growth parameters (OM12) and large values of tag loss which are assumed to force good fits to the trotline CPUE (OM16 and OM17).