Writing a custom brood function

2026-08-13

Default hatchery assumptions

Integrated hatchery programs may balance taking sufficient brood to reach release targets while minimizing gene flow of traits that favor juvenile freshwater survival in the hatchery environment. To minimize this gene flow, some proportion of the brood should be natural-origin.

Important decisions for hatchery planning must be made as the system becomes more hatchery-dominant, i.e., as a larger proportion of the population is hatchery-origin.

If there is sufficient return, does the hatchery prioritize production or the natural-ness of the population?

In salmonMSE, the various default settings for hatchery production roughly follow those from the All-H Analyzer. Users specify the target releases and the target pNOB (the proportion of brood that is natural-origin). Importantly, these settings prioritize hatchery production at the cost of PNI management (PNI is a metric of natural vs. hatchery influence on the population, higher values indicate higher natural influence). If there is sufficient return but few natural-origin fish, then brood is continually taken to achieve target releases and the realized pNOB will be lower than the target.

An additional assumption of the default settings is that broodtake is non-selective with respect to age.

Custom brood function

Alterations of the brood mechanics can be achieved with custom brood rule functionality in salmonMSE.

This feature allows the user to write a function in R that implements flexible rules that determine the brood composition given the return and relaxes the two assumptions of (1) prioritizing hatchery production at all costs and (2) non-selective brood with age.

Additional creativity in the brood rule is also possible, for example, turning off hatchery production when the return is too low (however, this could compound problems if survival in the hatchery is higher than in the natural environment!).

In the projection, once the brood is calculated from applying the custom function, the brood numbers is also reduced proportionally to ensure that the production target is not exceeded (in other words, the target is also the limit). This allows the user to focus on rules that determine the appropriate brood composition based on inputs (the return) rather than the outputs.

Writing the function

The custom function requires four input arguments, values of which are provided by salmonMSE to the function during the projection:

• NO a matrix with rows for age and columns for life cycle groups (frequently 1) for natural-origin return available for brood (after escapement from marine fisheries and en-route return survival)
• HO a matrix with rows for age and columns for release strategies for hatchery-origin return available for brood
• stray a matrix with rows for age and columns for release strategies for strays
• m a numeric for the mark-rate of hatchery-origin return

The output from the function needs to be a named list containing the following entries:

• NOB a matrix, same dimensions as NO, for the natural-origin brood
• HOB_marked a matrix, same dimensions as HO, for the hatchery-origin brood that is marked
• HOB_unmarked a matrix, same dimensions as HO, for the natural-origin brood that is un-marked
• HOB_stray a matrix, same dimensions as stray, for the brood taken from strays

The hatchery-origin brood calculations is explicit regarding marked and stray status to facilitate hatchery programs based on mark rate and the presence/absence of strays.

The minimum acceptable function looks like this (this is a function for no hatchery production):

brood_function <- function(NO, HO, stray, m) {
  NOB <- array(0, dim(NO))
  HOB_marked <- HOB_unmarked <- array(0, dim(HO))
  HOB_stray <- array(0, dim(stray))
  
  output <- list(
    NOB = NOB, 
    HOB_marked = HOB_marked, 
    HOB_unmarked = HOB_unmarked, 
    HOB_stray = HOB_stray
  )
  return(output)
}

The function can then be passed to the f_brood slot in the Hatchery object of the operating model:

Hatchery <- new("Hatchery")
Hatchery@f_brood <- brood_function

SOM <- new("SOM", Hatchery = Hatchery)
SMSE <- salmonMSE(SOM)

Example 1

The first example describes a hatchery program where:

• Brood is preferentially selected among age 4 and age 5 (any younger fish may be too small to be selected as brood)
• Up to 70 percent of the return in those age classes can be taken for brood
• Brood is not selective regarding origin (for example, no marking is done)

brood_function <- function(NO, HO, stray, m) {

  NOB <- array(0, dim(NO))
  HOB_marked <- HOB_unmarked <- array(0, dim(HO))
  HOB_stray <- array(0, dim(stray))
  
  n_age <- nrow(NOB)
  
  ptake <- rep(0, n_age)
  ptake[4:5] <- 0.7
  
  NOB[] <- ptake * NO
  HOB_unmarked[] <- ptake * (1 - m) * HO
  HOB_marked[] <- ptake * m * HO
  HOB_stray[] <- ptake * stray

  output <- list(
    NOB = NOB, 
    HOB_marked = HOB_marked,
    HOB_unmarked = HOB_unmarked, 
    HOB_stray = HOB_stray
  )
  return(output)
}

Example 2

Here’s a more complex example where:

1. No brood if the return is less than 500
2. The brood does not exceed 33 percent of the return
3. No more than 50 percent of the natural-origin return is used as brood
4. Natural-origin fish always at least 75 percent of the broodtake (pNOB = 0.75)
5. The mark rate is assumed to be 100 percent so that we can achieve rules 3 and 4
6. Strays are negligible

A function that implements this program looks like this:

brood_function <- function(NO, HO, stray, m) {

  NOB <- array(0, dim(NO))
  HOB_marked <- HOB_unmarked <- array(0, dim(HO))
  HOB_stray <- array(0, dim(stray))
  
  if (sum(NO, HO, stray) >= 500 && m == 1) { ## Rules 1 and 5
    
    max_brood <- 0.33 * sum(NO, HO, stray) ## Rule 2
    NOB_max <- 0.5 * sum(NO) ## Rule 3
    
    ## Rule 4
    NOB_total <- min(max_brood, NOB_max)
    
    # pNOB = NOB/(NOB + HOB) -> HOB = (NOB - pNOB * NOB)/pNOB
    pNOB <- 0.75
    HOB_total1 <- (NOB_total - pNOB * NOB_total)/pNOB
    HOB_total2 <- max(0, max_brood - NOB_max)
    HOB_total <- min(HOB_total1, HOB_total2)
    
    # This code distributes the brood proportionally among the various age classes in the return
    if (sum(NO)) {
      ptake_NOB <- NOB_total/sum(NO)
      NOB[] <- ptake_NOB * NO
    }
    
    if (sum(HO)) {
      ptake_HOB <- HOB_total/sum(HO)
      HOB_marked[] <- ptake_HOB * HO
    }
  }
  
  output <- list(
    NOB = NOB, 
    HOB_marked = HOB_marked,
    HOB_unmarked = HOB_unmarked, 
    HOB_stray = HOB_stray
  )
  return(output)
}