Limited Conservation Benefits From the First Marine Protected Area in the High Seas

Data and code for our evaluation of the world's first large-scale spatial closure in areas beyond national jurisdiction.

Authors: Juan Carlos Villaseñor-Derbez, John Lynham

Repository structure

.
├── Makefile                 # rebuilds the whole pipeline (see below)
├── .Rprofile                # activates renv, then sources scripts/00_config.R
├── renv.lock                # pinned package versions (see Requirements)
├── renv/                    # renv infrastructure; library/ is gitignored
├── dag.png                  # dependency graph of the build
├── PNA_HS_closure_prereg.pdf
├── data/
│   ├── raw/                 # never modified by our code
│   ├── processed/           # built by scripts/01_processing
│   └── output/
├── scripts/
│   ├── 00_config.R          # colors, labels, table/figure helpers
│   ├── 01_processing/       # raw data  --> analysis panels
│   ├── 02_analysis/         # panels    --> tables, event studies, summaries
│   └── 03_content/          # panels    --> maps & descriptive figures
├── content/
│   ├── tab/                 # LaTeX regression tables
│   ├── img/                 # figures (png)
│   └── summaries/           # LaTeX summary statistics quoted in the text
├── docs/                    # standalone slide decks / handouts
└── manuscript/              # git submodule --> Overleaf project

The data/raw folder holds datasets that have not been touched by our code. The scripts folder holds all code, split by objective (process data, analyze data, or create content). Everything those scripts produce is exported to content/. The DAG below shows how the pieces connect:

manuscript/ is a submodule

manuscript/ is a git submodule pointing at the Overleaf project. It's only purpose was to simplify the updating of figures and tables into the text, and bears to role in the reproducibility pipeline.

Data sources

All raw data are publicly available. Large vector files are tracked with Git Large-File Storage LFS — install it before cloning, or run git lfs pull afterwards.

Catch and effort

From the WCPFC Public Domain Aggregated Catch/Effort data download page, downloaded Oct 31, 2025. Both cover 1950–2023 for the WCPFC Convention Area.

Fishery	File	Resolution
Tuna purse seine	WCPFC_S_PUBLIC_BY_1x1_MM_5	1°×1° grid, year and month
Tuna longline	WCPFC_L_PUBLIC_BY_YR_MON_3	5°×5° grid, year and month

Spatial layers

Layer	Folder	Source
Exclusive Economic Zones (v12, 2023-10-25)	data/raw/World_EEZ_v12_20231025_gpkg	Marine Regions
High seas (v2, 2024-10-10)	data/raw/World_High_Seas_v2_20241010_gpkg	Marine Regions
Seamounts (v2)	data/raw/YessonEtAl2019-Seamounts-V2	Yesson et al. 2019

The WCPFC Convention Area polygon is not downloaded. It it is constructed from the coordinates in Article 3 of the Convention by scripts/01_processing/02_build_WCPFC_convention_area.R.

data/raw/fish_pics/ holds the species and gear icons (SVG) used to annotate the main figures.

Requirements

Analysis was run with R 4.6.1. Package versions are snapshotted with renv: renv.lock records the exact version of all 170 packages in the dependency tree, resolved against the Posit Package Manager CRAN mirror.

.Rprofile sources renv/activate.R, so opening the project in R activates the project library automatically. To install it:

renv::restore()

That is the only setup step. renv/library/ is gitignored — the lockfile is the source of truth, and renv::restore() rebuilds the library from it.

Scripts still load packages with pacman::p_load(), which now resolves against the renv library rather than your system library.

Every script resolves paths with here::here(), so scripts run correctly from anywhere in the project. .Rprofile also sources scripts/00_config.R, which defines the project-wide color scheme, species labels, modelsummary defaults, and the save_table() / es_save() helpers used by every analysis script. If you run a script outside RStudio, source it yourself.

Reproducing the analysis

The Makefile drives the reproducibility pipeline:

data/raw        --> data/processed           (scripts/01_processing)
data/processed  --> content/tab, content/img (scripts/02_analysis)
data/processed  --> content/img              (scripts/03_content)

make                # build everything
make processing     # build raw data --> processed panels only
make analysis       # build regression tables and event-study figures only
make content        # build maps and descriptive figures only
make h1             # build a single hypothesis (also: h2, h3,
                    #   h2_robustness, h3_robustness)
make clean          # remove generated tables and figures

# Optional
make dag            # regenerate dag.png (needs make2graph and graphviz)

What each script does

scripts/01_processing/ — raw data to analysis panels

Script	Output
01_make_PNA_hs_pocket.R	PNA_high_seas_pockets.gpkg, PNA_eezs.gpkg
02_build_WCPFC_convention_area.R	WCPFC_convention_area.gpkg
03_clean_wcpfc_ps_data.R	wcpfc_ps_annual.rds
04_clean_wcpfc_ll_data.R	wcpfc_ll_annual.rds
05_build_h1_panel.R	h1_panel.rds — assigns treated/control cells for H1
06_build_h2_panel.R	h2_panel.rds — assigns treated/control cells for H2
07_build_h3_panels.R	h3_ps_panel.rds, h3_ll_panel.rds — near/far cells

scripts/02_analysis/ — panels to models, tables, and event studies

Script	Output
00_basic_stats.R	scratch script for descriptive tonnage / bigeye mortality numbers quoted in the text; saves nothing, so it is not wired into make
01_estimate_h1.R	tab/h1_reg*.tex, img/h1_*_es.png, img/h1_main_figure.png, summaries/h1_summaries.tex
02_estimate_h2.R	tab/h2_reg*.tex, img/h2_*_es.png, img/h2_main_figure.png, img/h2_coefplot_*.png, img/h2_effort_ts.png, summaries/h2_summaries.tex
02_estimate_h2_robustness_by_pocket.R	H2 for bigeye split at 152.5°E (HSP1 vs. HSP2) → h2_rob_pocket_bet_ll.tex
03_estimate_h3.R	tab/h3_reg*.tex, img/h3_*_es.png, img/h3_main_figure.png, summaries/h3_summaries.tex
03_estimate_h3_robustness_by_pocket.R	H3 for skipjack split at 152.5°E → h3_rob_pocket_skj_ps.tex + event study

scripts/03_content/ — panels to maps and descriptive figures

Script	Output
01_make_HS_pocket_map.R	fig_HS_pocket_map.png, fig_h1_map.png, fig_h2_map.png, fig_h3_ps_map.png, fig_h3_ll_map.png, fig_seamount_density.png
03_fig_ps_catch_inside.R	fig_ps_catch_inside.png — pre-closure purse seine catch composition inside the pockets
04_fig_fishing_in_hs_pockets.R	diagnostic that rasterizes purse seine presence to check which grid cells count as fully covered by a pocket; draws to the device and saves nothing, so it is run by hand rather than through make

Funding

J.C.V.D. received funding from the Pew Charitable Trusts. The funders had no say in the design or execution of the research.