Skip to contents

Iteratively Optimise an MPA Network

Source: R/optimize_mpa.R
optimize_mpa.Rd

Uses a stochastic importance-resampling (SIR) algorithm to identify an MPA network that maximises a weighted combination of biodiversity and economic objectives. At each iteration the function evaluates the marginal value of toggling individual patches in or out of the MPA, then expands or contracts the network by the patch(es) with the highest marginal objective value.

Usage

optimize_mpa(
  fauna,
  fleets,
  starting_conditions = NA,
  alpha = 0.33,
  max_prop_mpa = 1,
  resolution,
  prop_sampled = 0.2,
  max_delta = 2,
  workers = 6,
  bio_objective = "max_ssb",
  econ_objective = "yield",
  work_backwards = TRUE,
  patches_at_a_time = 1
)

Arguments

fauna

Named list of fauna objects from create_critter.

fleets

Named list of fleet objects from create_fleet, tuned with tune_fleets.

starting_conditions

Initial population state, typically sim[[length(sim)]] from a prior simmar run. Passed as initial_conditions to each inner simmar call.

alpha

Numeric in [0, 1]. Weight on biodiversity (SSB/SSB0) in the composite objective. alpha = 1 ignores economics; alpha = 0 ignores biology. Default 0.33.

max_prop_mpa

Numeric in (0, 1]. Maximum proportion of patches that can be designated as MPA. Setting this below 1 forces work_backwards = FALSE automatically. Default 1.

resolution

Integer scalar or length-2 vector c(nx, ny). Must match the resolution of fauna and fleets.

prop_sampled

Numeric in (0, 1]. Proportion of candidate patches to evaluate at each iteration. Lower values speed up computation but increase stochasticity. Default 0.2.

max_delta

Numeric. Unused; kept for backwards compatibility.

workers

Integer. Number of parallel workers for future::multisession. Default 6.

bio_objective

Character. Biodiversity sub-objective:

"max_ssb"

(default) Sum of SSB/SSB0 across species; favours absolute biodiversity gains.

"min_loss"

Count of species that do not decline relative to baseline; favours preventing any losses.

econ_objective

Character. Economic sub-objective used to evaluate each candidate patch:

"yield"

(default) Total catch across species.

"revenue"

Total revenue.

"profits"

Total profit.

work_backwards

Logical. If TRUE (default and recommended), start at full MPA coverage and iteratively remove patches. Set to FALSE to grow the network from empty.

patches_at_a_time

Integer. Number of patches to toggle per iteration. Default 1.

Value

A named list:

outcomes

Tibble of biodiversity and economic outcomes at each MPA size level.

objective

Tibble of the composite objective value (obj) at each MPA size, plus component scores (ssb, econ, yield).

mpa_network

Tibble mapping each iteration (proportion protected) to the corresponding MPA location data frame.

Details

Algorithm

When work_backwards = TRUE (recommended), the algorithm starts with 100\ improves the weighted objective. This "reverse greedy" approach tends to produce more stable networks than the forward greedy (start empty, add patches) because it avoids early lock-in to suboptimal configurations.

The composite objective at each iteration is: $$obj = \alpha \cdot \tilde{B} + (1 - \alpha) \cdot \tilde{E}$$ where \(\tilde{B}\) and \(\tilde{E}\) are percent-rank-scaled biodiversity and economic outcomes, respectively. Scaling by percent ranks means \(\alpha\) strictly controls the weight given to biodiversity relative to economic rank, not their absolute magnitudes.

Parallel evaluation of candidate patches uses furrr::future_map; set workers to control the number of parallel sessions.

See also

simmar, create_critter, create_fleet