Gulf of Alaska Rockfish Sport Fish Harvest Reconstruction

Author: Phil Joy (philip.joy@alaska.gov)

Last updated: June 2026

Local Directory

The local directory will contain files that are not available in the git repository:

• Juneau: Phil's H:\Documents\Rockfish_SF_mortality\
• Anchorage S:\RTS\FishSci1_groundfish\catch_accounting\

About this repository

This repository houses the estimation procedure for harvest, release, and total sport fish mortality of rockfish in coastal Alaska developed as part of the ADF&G Statewide Rockfish Initiative. This is where the raw data is processed and formatted and houses the estimation proceedure. There are two methods/approaches for estimating rockfish harvest, release an mortality estimates. The method used to date (through 2023 estimates) has been the Howard method (Howard et al. 2020) originally conducted by running SAS code followed by excel based calculations. Those methods were transitioned to R in 2024 for reproducibility and efficiency. One limitation of the Howard approach was a decision-tree process whereby data and information was borrowed from other areas when estimates were unavailable or unreliable due to limited sample sizes. A second limitation is that the procedure could only produce estimate back through 1998 when the logbook program was initiated

The second method is under development and comprises a Bayesian version of the Howard methods, originally recommended in Howard et al. (2020), that allows for more appropriate and defensible sharing of information between areas, handles missing data in a more appropriate manor, accurately propagates uncertainty throughout the estimation procedure and thus does not rely on the decision tree approach in the original Howard methods. Furthermore, the Bayesian approach should provide sport fish harvest, catch and mortality estimates back to 1978 when the SWHS was implemented. The original Bayesian methods were developed by Adam Reimer and has been further developed by the current author. For communication purposes we will refer to Howard methods and Reimer methods to distinguish the two approaches. A summary of these efforts are available here, here, here, and here.

Data reports for dissemination:

In development...

Data Sources:

1. SWHS data: from Jake Bozzini (jake.bozzini@alaska.gov): Available in late September when SWHS estimates are finalized. Save to [data/raw_dat/YEAR/] folder.
   1. rf_byMgmtUnit_sentDATE.xlsx
   2. IPHC_YEAR_guipri_all_sentYEAR.xlsx (NOTE: This is currently in Southcentral port sample data folder and needs to be moved)
2. eLogbookdata: YEARLogbookDataDATE.csv or statewide_YEAR_DATA.csv depending on what they are calling from Kyla Buster (kayla.buster@alaska.gov) and saved to [data/raw_dat/YEAR/] folder.
3. Southcentral port sampling data from Clay Mckean (clay.mckean@alaska.gov, git repository) saved to [data/raw_dat/Species_comp_SC/]:
   1. Spcomp_guided_SENTDATE
   2. Spcomp_unguided_SENTDATE
4. Southeast port sampling data from Diana Tersteeg (diana.tersteeg@alaska.gov) and Chris Hinds (chris.hinds@alaska.gov) and saved to [data/raw_dat/Species_comp_SE/]:
   1. Species_comp_MHS_Region1_forR_YEAR_RUN_DATESTAMP.xlsx
5. Kodiak stereocamera species apportionment data from Philip Tschersich (philip.tschersich@alaska.gov) and Tyler Polum (tyler.polum@alaska.gov) and saved to [data/raw_dat/kodiak_hydro_dat.csv]
6. Release mortality data I don't think these will change but worth checking with Chris and Clay either way. This data resides here [data/raw_dat/Species_comp_SC/rf_mort_sc24] and [data/raw_dat/Species_comp_SE/Species_comp_MHS_Region1_forR_2024_RUN_TIMESTAMP.xlsx].
7. Weight data. Raw weight comes from the port sampling teams in the form of [data/raw_dat/Species_comp_SC/sample_size_rf_SC_Port_Sampling.csv] from Region 2 and [data\raw_dat\Species_comp_SE\SE_2011_2025_number of vessels with sampled RF_TIMESTAMP.xlsx'] from region 1. SE also estimates mean weights using a 4-stage sampling program that is being redeveloped at the time, but the lates data from Diana came in the form of [data/raw_dat/Species_comp_SE/SEAK_2025_avg_GF_Area_RPT_DATE.csv].
8. Interview data: This is used to inform releases for unguided anglers and comes from the port sampling programs as [data\\raw_dat\\species_comp_SE\\Spp.Comp_MHS_Region1_forR.xlsx] for Region 1 and [data/raw_dat/Species_comp_SC/sc_rf_release.csv] for Region 2.

Estimation instructions

Due date: early October

Work flow for data processing:

1. Collect data from various sources.
   1. Species compostional data and logbook data are available anytime after the new year.
   2. Last year's SWHS data are available in late September of the following year. You should have all of the other data sources in place when the SWHS data is released so that estimates can be generated in the fall for Board of Fish meetings.
2. Process the statewide harvest survey data in [swhs_processing.R].
3. Process the logbook data in [lb_processing.R].
4. Weight and process the SC port sampling data in [SC_apportionment_calcs.R].
   • Note that the southeast group has developed code for apportionments in Region 1.
5. Process weight data in [wt_dat_processing.R].

Howard methods:

1. Generate Howard estimates in [BRF_Howard.R], [YE_Howard.R], [DSR_Howard.R], [SLOPE_Howard.R], and [PEL_Howard.R].
   • Note that DSR, SLOPE and PEL are for Southeast Region only unless otherwise requested by Region 2.
2. Howard estimates get sent to Chris and Clay to generate release mortality estimates and biomass estimates.

Bayesian methods:

1. Prepare raw data for the Bayesian model using [scripts/bayes_data_prep.R].
2. Update prior values for a parameter in the model using [pH_B4_prior_dev.R].
3. Historical estimates from 1977-2019 were generated in [scripts/run_bayes_working.R] and are due to be published in late 2026.
4. Contemporary estimates from 2020 onward are generated with [scripts/run_bayes_annual_ests.R]. Results from the model run are visualized using [markdown/RFsportmort_annual_quickplots.Rmd] which will also generate output tables suitable for dissemination in [markdown/RF_Sport_Removals_RepV.xslx].

Repository Directory

1. [data/]: Data folder:
   1. [data/raw_dat/]: Despite the name, this folder contains both the raw data as described above, and processed data and calculations generated during the course of the analysis. Files in these folders that are not included in the source list above are generated in the process of running the scripts.
      1. [data/raw_dat/YEAR/]: The year folders will contain the raw SWHS and logbook data for the year as well as files that will be generated by the code each year.
      2. [data/raw_dat/Species_comp_SE/]: This folder contains the updated port sampling data as generated by Chris and Diana each year. Chris has developed code to calculate the apportionment numbers so this file contains those results.
      3. [data/raw_dat/Species_comp_SC/]: This folder contains the updated port sampling data as generated by Clay each year. Other files in this folder are generated when running the code [SC_apportionment_calcs.R] code.
      4. [data/raw_dat/kodiak_hydro_dat.csv]: This is data from the hydroacoustic survey in Kodiak used to augment species apportionment in Kodiak pelagics.
      5. [data/raw_dat/Archive/]: This folder contains old data used prior to transitioning to R in 2024.
      6. [data/raw_dat/logbook_harvest_thruYEAR.csv] and [data/raw_dat/logbook_release_thruYEAR.csv] are compiled logbook data generated when running the code.
      7. [data/raw_dat/SWHS_1977_1995_rf_estimates_sent20241107.xlsx] are the pre-1996 rockfish removal estimates from the SWHS.
   2. [data/bayes_dat/]: This folder contains the processed data that is prepared for the Bayesian model.
2. [scripts/]: Scripts for running the analysis
   • Data processing:
     1. [scripts/swhs_processing.R]: Processes and archives the swhs data.
     2. [scripts/lb_processing.R]: Processes and archives the logbook data and adds it to the swhs data.
     3. [scripts/SC_apportionment_calcs.R]: Apportions the raw Region 2 port sampling data to generate species apportionment estimates. Note that Region 1 began doing this task in house in 2024.
     4. [scripts/wt_dat_processing.R]: Processes weight data for use in the Bayesian model.
     5. [scripts/bayes_data_prep.R]: Prepares raw data for the Bayesian model.
     6. [scripts/pH_B4_prior_dev.R]: Code for developing and updating the priors used to inform unguided releases based on SWHS data.
   • Howard estimates: Produces rockfish harvest and release estimates using the Howard methods.
     1. [scripts/BRF_Howard.R]: Black rockfish estimates.
     2. [scripts/YE_Howard.R]: Yelloweye rockfish estimates.
     3. [scripts/DSR_Howard.R]: Demersal shelf rockfish assemblage estimates (Southeast region only).
     4. [scripts/SLOPE_Howard.R]: Slope rockfish assemblage estimates (Southeast region only).
     5. [scripts/PEL_Howard.R]: Pelagic rockfish assemblage estimates (Southeast region only).
   • Bayesian estimates: Produces rockfish harvest and release estimates using the Bayesian methods. Under development
     1. [scripts/bayes_data_param_load.R]: This script provides functions for formatting and preparing data for use in the Bayes model.
     2. [scripts/functions.R]: Functions for use in Bayesian scripts..
     3. [scripts/run_bayes_working.R]: This is the script used for running the historical model between 1977 and 2019.
     4. [scripts/run_bayes_annual_ests.R]: This script is used for generating annual estimates from 2020 onward.
     5. [scripts/functions.R]: Functions for use in the bayes_est.R script.
     6. [scripts/bayes_data_check.R]: This script was developed to compare data preparation from 2 different scripts.
     7. [scripts/bayes_model_comps.R]: This was code use in model development to compare the output from different models.
     8. [scripts/mkdwn_functions.R]: This is code to make functions available for rendering output in Markdown.
     9. [scripts/re_check_code.R]: This is code used in model development to make sure that random effects are functioning properly in the model.
     10. [scripts/logit_prior_plotting.R]: This is rough code for understanding how logit curves in the Bayes model function when developing the model.
     11. [scripts/slope_wt_exam.R]: This code was used to help the SE region determine whether slope rockfish needed to be split out into two sizes. Spoiler alert: they do not.
     12. [scripts/bayes_retro.R]: This code was developed to look at retrospective patterns in the Bayesian estimates. This was used in model development but not included in the pubilcation.
     13. [scripts/bayes_prior_development.R] is just rough script for looking at prior distributions.
     14. [scripts/SE_lb_processing.R]: This script has information of assigning SWHS areas to logbook data. It is deprecated but left in here in case those geographical relationships need to be referenced at some point.
     15. [scripts/bayes_dev/] is a folder containing deprecated code used in developing the Bayesian model. This will be scrapped at some point.
3. [models/]: This contains jags models for the Bayesian Reimer model.
4. [figures/]: This folder contains figures produced in the analysis.
5. [output/]: This folder contains files and results for dissemination.
   1. [output/reports/]: This folder contains area specific reports for dissemination.
6. [markdown]: This contains text files and rmarkdown files.

Old readme notes below while develoiping this repo:

Replicated an excel based version of the harvest reconstruction prepared by Katie Howard. Currently this repository contains a Bayesian version of the same estimates although the objective was to add earlier SWHS data to the analysis to extend the time series of estimates.

The working directory for this repository is located at S:\RTS\Reimer\RockfishSportMort.

The file/folder structure for this analysis is as follows:

• data/: R data sets created by readdata.R.

• data-raw/: raw data for the analysis.

• functions/: helper functions for the analysis.

• markdown/: RMarkdown document and html output associated with this analysis.

• models/: Jags code associated with this analysis.

• posts/: Posterior samples from the model. These files are large and not available on GItHub (stored in the network location indicated above). Posterior samples from earlier commits follow the naming convention postH_{8digitsha}.rds

• scripts/: R code to run analysis.