An example of a workflow to build an Operating Model based on life history, project it forward for a \(F/F_{\mathrm{MSY}}\) trajectory, recover process-error residuals, then fit JABBA with and without depletion-dependent PE, then evaluate ability to i) classify stock status relative to target and limit biomass and exploitation rate reference points; ii) identify trends, and iii) prediction rebuilding horizons.
Set quick=FALSE to match the slide figures (100
iterations, full \(F\) grid).
lhPar() / lhEql() give an
FLBRP. Convert to an FLStock and project with
ffwd() and recruitment deviations.
#> An object of class "FLPar"
#> quant
#> refpt harvest ssb eb yield
#> msy 4.38e-01 1.69e+02 2.54e+02 8.05e+01
#> crash 3.24e+00 4.45e-06 1.26e-05 1.69e-05
#> units: NA
Project a grid of constant \(F\)
multipliers. Process error is the log residual after catch
(location="after"). SD of that series is the PE scale at
each \(F\).
Figure 1. Process-error scale (SD of log surplus-production residuals) against constant \(F/F_{\mathrm{MSY}}\).
PE scale increases with \(F\): i.e. the surplus-production residual is larger when the stock is fished harder.
Trajectories that ramp up from \(0.01F_{\mathrm{MSY}}\) to \(2F_{\mathrm{MSY}}\); and declines from \(2F_{\mathrm{MSY}}\) to \(0.01F_{\mathrm{MSY}}\); noise is based on same recruitment residuals.
Figure 2. Operating-model trajectories under increasing and declining \(F/F_{\mathrm{MSY}}\).
Year-wise SD of log PE (after a burn-in):
Figure 3. SD of log process error after burn-in, for increasing and declining \(F\).
Compares PE and \(B/B_{\mathrm{MSY}}\) for an example
iteration
Figure 4. Four-panel process-error diagnostics for a single increasing-\(F\) realisation: time, status versus \(B/B_{\mathrm{MSY}}\), ACF, and histogram.
The status panel is the empirical counterpart of
JABBA’s proc.dyn curve: note PE SD is larger at low \(B/B_{\mathrm{MSY}}\).
Grid for \(k\), \(M\), \(\sigma_R\)
Take one realisation (it) of the increasing-\(F\) stock, use catch and the biomass index
from FLStock and fit JABBA is fitted to the
exploitable-biomass index. \(F_{\mathrm{MSY}}\), \(B_{\mathrm{MSY}}\) and \(B_0\) from FLBRP.
quick=TRUE uses JABBA’s short MCMC.
plotJabbaTs() draws the posterior median with 95%
credible-interval ribbons. OM overlays (dashed) are spawning biomass and
exploitable biomass, each relative to its own \(B_{\mathrm{MSY}}\) from FLBRP
(refpts SSB and refptsEB()).
Figure 5. Process-error diagnostics for the JABBA fit with constant process-error variance.
Figure 6. JABBA posterior medians and 95% credible intervals (constant PE).
Figure 7. JABBA \(B/B_{\mathrm{MSY}}\) (median and 95% CI) with spawning biomass and exploitable biomass (dashed) relative to its own \(B_{\mathrm{MSY}}\).
Optional: same iteration with proc.dyn=TRUE.
```
Figure 8. Process-error diagnostics for the JABBA
fit with depletion-dependent PE (proc.dyn).
Figure 9. JABBA posterior medians and 95% credible intervals (depletion-dependent PE).
Figure 10. Depletion-dependent JABBA \(B/B_{\mathrm{MSY}}\) with OM spawning biomass and exploitable biomass (dashed).
Figure 11. Constant versus depletion-dependent JABBA \(B/B_{\mathrm{MSY}}\) (median and 95% CI) and OM spawning and exploitable biomass (dashed).
Figure 12. Same \(B/B_{\mathrm{MSY}}\) comparison as the previous figure, restricted to the recent period (year \(> 100\)).
Figure 13. Constant versus depletion-dependent JABBA \(B/B_{\mathrm{MSY}}\), \(F/F_{\mathrm{MSY}}\) and process error, with 95% CIs and OM biomass overlays.