R/hcr_management_fmort_pred.R

In einarhjorleifsson/mac: Miscalaneous set of functions related fish stock assessment work

#' @title F-based Harvest Control Rule
#' 
#' @description The F-based rule is the conventional ICES decision rule. Here it
#' is implemented such that the TAC next year is calculated from the true
#' stock in numbers based on a fishing mortality that includes observation error.
#' 
#' If the Btrigger is set in the rule (Btrigger > 0) then linear reductions of 
#' fishing mortality is done relative to observed spawning stock biomass (i.e.
#' that includes observation errrors).
#' 
#' @export
#' 
#' @param y XXX
#' @param h XXX
#' @param ctr Control file
#' @note XXX
#'

hcr_management_fmort_pred <- function(y, h, ctr)
{
  hrate <- ctr$HRATE[h]
  # SHORT TERM PREDICTION MODEL
  # All values in this block are actually true numbers
  # Use last three years
  # Special case: Need to take into account that the short term mean
  # can only be counted calculated in simulation year 3.
  # I.e.:
  #       when y=2, only have one historical year
  #       when y=3 have two years of history
  #       when y>3 have three years of history
  y1 <- ifelse(y < 4,1,y-3)
  
  bWa  <- X$bW[,c(y1:(y-1)), h,]
  cWa  <- X$cW[,c(y1:(y-1)), h,]
  sWa  <- X$sW[,c(y1:(y-1)), h,]
  selF <- X$selF[,c(y1:(y-1)), h,]
  selD <- X$selD[,c(y1:(y-1)), h,]   # Not used for mackerel
  Ma   <- X$M[,c(y1:(y-1)), h,]      # Actually a constant
  pF   <- X$pF[,c(y1:(y-1)), h,]
  pM   <- X$pM[,c(y1:(y-1)), h,]
  mat  <- X$mat[,c(y1:(y-1)), h,]
  
  selB <- X$selB[,y + ctr$delay,h,] # This stuff is fixed
  # Only take mean if there are more than 1 year
  if(y > 2) {
    bWa  <- apply(bWa,c(1,3),mean)
    cWa  <- apply(cWa,c(1,3),mean)
    sWa  <- apply(sWa,c(1,3),mean)
    selF <- apply(selF,c(1,3),mean)
    selD <- apply(selD,c(1,3),mean)
    Ma   <- apply(Ma,c(1,3),mean)
    pF   <- apply(pF,c(1,3),mean)
    pM   <- apply(pM,c(1,3),mean)
    mat  <- apply(mat,c(1,3),mean)
  }
  
  # True stock in numbers
  Na <- X$N[,y + ctr$delay,h,]
  
  
  Fa <- t(hrate * t(selF))
  
  
  # RECRUITS short term predictions
  # 1. Geometric mean assumptions applied to age 0
  Na[1,] <- ctr$r_mean
  # 2. Geometric mean assumptions applied to age 1 if the delay is 1, i.e.
  #    if we are using the stock in numbers in the advisory year
  #    Since pF (and pM) may vary need to use the short term prediction values
  if(ctr$delay > 0) {    
    Na[2,] <- ctr$r_mean * exp(-(Fa[1,]+Ma[1,]))
  }
  
  # Predicted biomass based only on true short term mean and recruitment assumption
  bio <- colSums(Na * bWa * selB)
  ssb <- colSums(Na * exp( -( pM * Ma + pF * Fa)) * mat * sWa)
  
  # ASSESSMENT ERROR
  assError <- X$assError[y + ctr$delay, h,]
  
  bio_hat   <- bio   * ctr$a_bias * exp(assError)
  ssb_hat   <- ssb   * ctr$a_bias * exp(assError)
  hrate_hat <- hrate * ctr$a_bias * exp(assError)
  
  # adjust harvest rate according to the trigger
  i <- ssb_hat < ctr$b_trigger  # Note this operates on the ssb with error
  hrate_hat[i] <- hrate_hat[i] * ssb_hat[i]/ctr$b_trigger 
  
  
  Fa_hat <- t(hrate_hat * t(selF))
  Da_hat <- t(hrate_hat * t(selD))
  
  tac_next_year <- colSums(Na * Fa_hat/(Fa_hat + Da_hat + Ma + 1e-05) * 
                             (1 - exp(-(Fa_hat + Da_hat + Ma))) * cWa)
  
  # 20% TAC constraint
  #tac_next_year <- ctr$h_alpha * tac_this_year + (1 - ctr$h_alpha) * tac_next_year
  
  # TAC buffer
  if(ctr$h_alpha > 0) {
    tac_this_year <- X$TAC[y,h,] # This years TAC
    i <- tac_next_year > tac_this_year * (1 + ctr$h_alpha)
    if(any(i)) tac_next_year[i] <- tac_this_year[i] * (1 + ctr$h_alpha)
    i <- tac_next_year < tac_this_year * (1 - ctr$h_alpha)
    if(any(i)) tac_next_year[i] <- tac_this_year[i] * (1 - ctr$h_alpha)
  }
  
  X$TAC[y+1,h,] <<- tac_next_year
  
  if(ctr$h_beta > 0) {
    # now need to add in the constraint that the F does not deviate by more than
    #  10% from the target
    # Calculate the current Fmultiplier
    xF <- hcr_TAC_to_Fmult(y+1,h)
    # Now find out which of the xF deviate by more than 0.9 and 1.1 of the hrate
    i <- xF > hrate_hat * (1 + ctr$h_beta)
    if(any(i)) hrate_hat[i] <- hrate_hat[i] * (1 + ctr$h_beta)
    i <- xF < hrate_hat * (1 - ctr$h_beta)
    if(any(i)) hrate_hat[i] <- hrate_hat[i] * (1 - ctr$h_beta)
    # now we have adjusted the realized harvest rate the 10% F-constraint
    # Need to update the whole calucation for the TAC
    
    # NEED TO DOUBLE CHECK NEXT STEP, THIS WAS ALREADY FIDDLED WITH ABOVE
    #  No need to worry if NO Btrigger
    #i <- ssb < ctr$b_trigger  
    #hrate[i] <- hrate[i] * ssb[i]/ctr$b_trigger  
    Fa_hat <- t(hrate_hat * t(selF))
    Da_hat <- t(hrate_hat * t(selD))
    
    tac_next_year <- colSums(Na * Fa_hat/(Fa_hat + Da_hat + Ma + 1e-05) * 
                               (1 - exp(-(Fa_hat + Da_hat + Ma))) * cWa)
    X$TAC[y+1,h,] <<- tac_next_year
  }
}