WKGMSE3

ToR b) Develop more efficient ways of conducting searches over a grid to the required level of precision be investigated.

This is needed because of the high-performance computing requirements for full MSEs.
This work could include investigating statistical properties that relate sample size to required precision, GAMs to interpolate over an incomplete grid, etc.

WKLife

Workshop on the Development of Quantitative Assessment Methodologies based on LIFE-history traits, exploitation characteristics, and other relevant parameters for data-limited stocks

MyDas

Develop Proxy \(MSY\) Reference Points across the spectrum of data-limited stocks.
Diagnostics for screening reference points, assessment methods, HCRs and MPs
Management Strategy Evaluation to develop plans to manage risk given uncertainty

Emprical Control Rules

Used for ICES Category 3 stocks for which survey-based assessments indicate trends.

Includes stocks for which survey indices (or other indicators of stock size such as reliable fishery-dependent indices, e.g. lpue, cpue, and mean length in the catch) are available.

Catches are increased when the trend in an index is positive, and decreased when negative

\[TAC^1_{y+1}=TAC_y\times \left\{\begin{array}{rcl} {1-k_1|\lambda|} & \mbox{for} & \lambda<0\\[0.35cm] {1+k_2\lambda} & \mbox{for} & \lambda\geq 0 \end{array}\right. \]

where \(\lambda\) is the slope from the regression of \(\ln I_y\) on year (\(y\)) for a recent period, \(k_1\) and \(k_2\) are terms and action asymmetry.

Incorrectly increasing catch is not the same as incorrectly decreasing it. Since in the former a stock may become overexploited requiring a long-term closure, while in the case of the later catches are banked and can be taken once the error is realised.

Pollack Operating Model

Time series for an example simulation.

Machine Learning

Use Random Search to explore the impact of \(k_1\) and \(k_2\) on management objectives, i.e. risk of falling below \(B_{lim}\), \(MSY\) and inter-annual variability in catches.

No need to explicitly consider risk of falling below \(B_{MSY}\) or fishing above \(F_{MSY}\), since if yield is maximised and stock is sustainable then by definition \(MSY\) objectives are met.

Genetic algorithms: there is no single utility function as optimise as different stakeholders may weight objectives differenttly, Therefore use Pareto Frontiers to explore trade-offs between multiple objectives.

Support vector regression: to model the relationship between control parameters and performance measures. Makes no assumption about distributions of variables, and no need for transformations.

Objectives

Strict condition \(P(B>B_{lim}\)) > 0.95

Simulations for random combinations of control parameters (k1 and k2), colours showing probability of Blim > 95%.

Pareto Optimal Solutions

A Pareto plot for objectives corresponding to maximising yield and minimising average annual variation in yield (AAV), non-optimal solutions and optimal solutions that lie along the Pareto front are indicated.

Scenarios

Variability of index 10%, 20%, 30%
Length of series used in estimation of trend 3, 5 years.

Pareto Frontiers

Calibration Regressions

Calibration regressions identifing the values of k1 and k2 to achieve a given Pareto optimal solution. Vertical lines correspond to \(80% MSY\), AAV value derive from the Pareto optimum.

MSE

Summary statistics, for the MSEs run for the a single set of control parameters, k1=0.6 and k2=0.6.

Summary

In this example there was only 1 OM scenario, however, multiple OMs could have been considered and SVR used to model the impact on the Pareto Frontier.
Alternatively if probabilities can be assigned to OMs then a weighted random search could have been used.
Both methods would allow a Value-of-Infomation analysis to be performed, i.e. showing how a reduction in uncertainty reduced risk of management objectives not being met.

Conclusions

a) Develop guidelines for when and how reference points should be extracted from an MSE when one is conducted.

There are no explicit reference points developed in the MP, \(k_1\) and \(k_2\) are tuned, i.e. varied until the parameters that best meet management objectives are found.
The same procedure could be used for a model based HCR, i.e. where a biomass dynamic assessment model provides estimates of \(B_{trigger}\) based on exploitable biomass and \(F_{MSY}\) based on harvest rate. These may be related to, but are not the same as the reference points, based on \(SSB\) and instantaneous mortality, used in the OM to derive summary statistics, or in a base case assessment to review the performance of a management strategy after implementation.

b) Develop guidelines for how to treat the results of alternative operating models. Currently, these have been used as robustness tests for “optimised” management strategies.

Two choices, i.e.

evalute performamnce measures seperately, i.e. objectives must be met for all OM scenarios
weighted probabilities

In both cases there is a need to + Choose scenarios + Weight based on goodness of fit + Weight based on plausibility, hower in Carvalhoa et al., and Sharma et al., plausibility is never definded.

d) Develop more efficient ways of conducting searches over a grid to the required level of precision be investigated. This is needed because of the high-performance computing requirements for full MSEs. This work could include investigating statistical properties that relate sample size to required precision, GAMs to interpolate over an incomplete grid, etc.

Use Randon search and then GA & SVR to model relationship between control parameters and management objectives

e) Compare the short-cut and full MSE approaches, providing guidelines for use of the former as an approximation for the latter, if appropriate. Consideration should be given to MSE with alternative operating models (i.e. operating models not solely based on the currently-used assessment).