Specify prospective operating-model robustness perturbations

Neutral defaults preserve the existing CRA2 simulation. Perturbations apply only to projection years, not historical observations or fitted parameters. The EM is not told which trial generated its observations.

Usage

mse_robustness(
  cpue_multiplier = 1,
  cpue_drift = 0,
  cpue_power = 1,
  cpue_sd_multiplier = 1,
  cpue_reported_sd_multiplier = 1,
  implementation_multiplier = 1,
  implementation_log_sd = 0,
  catch_reporting_multiplier = 1,
  recreational_multiplier = 1,
  other_catch_multiplier = 1,
  recruitment_multiplier = 1,
  missing_cpue_seasons = integer(),
  missing_cpue_years = NULL
)

Arguments

cpue_multiplier

Positive scalar or AW/SS vector of index multipliers.

cpue_drift

Additional proportional linear index drift per projection year.

cpue_power

Positive exponent distorting the baseline index response.

cpue_sd_multiplier

Positive scalar or AW/SS multiplier of true log noise.

cpue_reported_sd_multiplier

Positive scalar or AW/SS multiplier of uncertainty supplied to the EM, independent of true noise.

implementation_multiplier

Non-negative multiplier of commercial TACC.

implementation_log_sd

Non-negative log SD of commercial implementation error.

catch_reporting_multiplier

Positive scalar or AW/SS multiplier of projected observed catch, applied consistently to legal and total components.

recreational_multiplier

Non-negative scalar or AW/SS multiplier of recreation.

other_catch_multiplier

Non-negative multiplier of other fisheries' removals.

recruitment_multiplier

Positive multiplier of future recruitment.

missing_cpue_seasons

Seasons (1=AW, 2=SS) whose projected CPUE is unavailable.

missing_cpue_years

Optional years with unavailable CPUE; NULL means all projection years. Missing observations invoke each MP's stated failure policy.

Value

Immutable configuration consumed by om_decamod().

Details

For baseline simulated CPUE I with conditional median mu and a fixed seasonal anchor a (last historical observation), the perturbed observation is a * (mu/a)^cpue_power * (1 + cpue_drift*k) * cpue_multiplier * exp(cpue_sd_multiplier * log(I/mu)), in projection year k. This is a distortion of the baseline index relationship, not a new fitted stock model. Drift is ADDITIONAL to fitted catchability creep; it is not an absolute replacement for the historical 1% assumption. A power below one flattens the baseline index response relative to the anchor. Reported CPUE SD is controlled separately, permitting recognised or unrecognised observation deterioration.

Commercial catch is TACC times a mean-one lognormal error times implementation_multiplier. Reporting bias changes observations, not actual removals. Recruitment multipliers alter future log recruitment deviations; the historical overlap is untouched. No trial is scientifically selected by this constructor. More general biological changes require conditioned OM files.