Projections

[marcnadon]

Here is the gist of how the projections should be done:

1) Run the full base case SS model with bootstrapped catches. 2) Specify in this model run that we want to project until 2028 using a fixed catch of X metric tons in all years. 3) Re-run this for catches ranging from 0 to MSY metric tons, with an appropriately low catch increment. 4) Use the MVLN function within each run to save key metrics: SSB, F, Fmsy, SSBmsst (also save the fixed Catch that was used). 5) Save the MVLN output into an R list that collates all the model run with different fixed projected catches.

Key things to do:

• [x] Update the Build_all_SS function so that the MVLN results are saved in a specific folder when "bootrap_n > 1"
• [x] Update the Build_all_SS function to include options to run projections (n_projYear, Fixed_Catch, etc.)
• [x] Build a new wrapper R script to run projections. This script will 1) define the catch values to be used for each species (APRU <- seq(0,1.7,0.1)), 2) Generate a list for a single species where each object is a fixed catch level, 3) call the parlapply or lapply function on this "catch" list, 4) within that function the build_all_SS function will be called on each catch level in the list, in turn, and create a temp folder with the MVLN output, 5) still within that function, process that MVLN output to keep SSB, F, Fmsy, SSBmsst (0.9 * SSBmsy) in each projected year (2022-2028), 6) save this output to a list. 7) the output of the lapply function will be that list
• [x] Create script to process the lapply outputted list object, convert it to a data.table (rbindlist function) and save it to a csv file
• [x] Process either that csv file or the outputted data to generate the risk tables and figures associated with projections.

[MOshima-PIFSC]

Changes I have made so far:

• [x] Update the Build_all_SS function so that the MVLN results are saved in a specific folder when "bootrap_n > 1":

  • MVLN outputs are saved in the bootstrap folder with the other files from all the bootstrap runs (data_boot, report_boot, etc)
  • Full output of mvln is saved as mvln_draws.RData (can be brought into R using load(file.path(dir, "mvln_draws.RData))). This includes 5000 mc draws for each bootstrap for all years of the model, including projection years. Columns include: year, run, type, iter, stock, harvest, SSB, F, Recr, and Catch.
  • Projection only mvln output is saved as mv_projections.RData in the same directory. This includes 5000 mc draws for each year and bootstrap. Columns include:year, run, stock, harvest, SSB, F, and Catch. An additional column was added SSB.SSBmsst where SSBmsst = (SSB/stock)*0.9. (Kept the run column for plotting purposes, Catch shows what the fixed catch was for that run).
  • MVLN only runs when N_boot > 0

• [x] Update the Build_all_SS function to include options to run projections (n_projYear, Fixed_Catch, etc.)

  • added Fixed_forecatch to allow you to input the fixed catch for projections. A single number or vector of numbers that is the same length as the number of years to be forecasted can be used. This is only used if Nforeyrs > 1.
  • I adjusted the first year for caps and allocations so that if you are running forecast then it will be after the forecasted period so that the Fixed_forecatch will not be overwritten (implemented because of the warning message).
  • added F_report_basis which changes the denominator to use when reporting F std. We were reporting raw F values but if you change it to 2 this will report F/FMSY.

[marcnadon]

The projection function is in a file named "Run_projections" and is operational. It generates projections from bootstrap x fixed catches and generates the appropriate figs and table.