OFEL-ENSOLL

ENSO Effects on Pacific Tuna Distribution and Abundance

Code for the manuscript submitted.

Overview

This repository contains the R analysis pipeline for examining how ENSO (El Niño–Southern Oscillation) affects the spatial distribution and relative abundance of four Pacific tuna species using longline fishery data from 1993 to 2023.

Species studied:

Code	Species	Spatial domain
ALB_N	Albacore (Thunnus alalunga)	North Pacific
ALB_S	Albacore (Thunnus alalunga)	South Pacific
BET	Bigeye tuna (Thunnus obesus)	Pacific-wide
YFT	Yellowfin tuna (Thunnus albacares)	Pacific-wide

Pipeline

The scripts are designed to be run sequentially. Each script documents its inputs, outputs, and dependencies in its file header.

Data preparation

Script	Description
GCB_Data.R	Read and clean WCPFC + IATTC longline data; aggregate to 5°×5° monthly grid; compute nominal CPUE per 1 000 hooks; merge environmental layers; download MEI.v2 and CPC ONI v5; assign strict ENSO phase; split into species-specific datasets
GCB_envData.R	Environmental data visualisation (depth profiles, correlation matrices, spatial maps); defines shared theme_gcb()

Species distribution modelling

Script	Description
GCB_Model.R	XGBoost Tweedie species distribution models (hyperparameter tuning via mlr3, spatial block CV, final model fit and prediction)
GCB_Model_VIF.R	VIF-based stepwise feature elimination with trace logging
GCB_RFECV.R	Recursive feature elimination with cross-validation (RFECV) score plots
GCB_Model_instance.R	Tuning instance diagnostic plots
GCB_Model_Dia_Pl.R	Model diagnostic plots (observed vs predicted, residuals, Q-Q)
GCB_SHAP.R	SHAP value computation and visualisation (importance bars + dependence plots)

ENSO–abundance analysis

Script	Description
GCB_RA.R	Relative abundance vs ENSO index analysis module (correlation, linear vs GAM, lag screening, GLS AR(1) phase contrasts, relative change vs Neutral with 95% CI)
GCB_RA_Pl.R	Three-panel figure: abundance–MEI relationship, ENSO phase violin plots, and relative change from Neutral → Fig 1
GCB_RA_CCF.R	Cross-correlation analysis (abundance vs MEI) → Fig A (CCF)

Centre of gravity analysis

Script	Description
GCB_COG-A.R	Centre of gravity (COG Lon/Lat) computation, GLS candidate models with harmonised AICc selection
GCB_COG-A_Pl.R	Three-panel COG figures (COG–MEI relationship, ENSO phase contrast, shift from Neutral) → Fig 2 (Lon), Fig 3 (Lat)

Spatial mapping

Script	Description
GCB_Spatial_Pl.R	ENSO phase spatial maps (Δ relative abundance % vs Neutral)
GCB_Spatial_Lonlat_Pl.R	Longitudinal and latitudinal profiles of abundance change, combined with spatial maps → Fig 4

Data exploration

Script	Description
GCB_Data_Pl.R	Spatial distribution and time-series plots for nominal CPUE → Fig S1

Data sources

• Longline catch and effort: WCPFC public domain data and IATTC public longline data
• ENSO indices: NOAA MEI.v2 and CPC ONI v5
• Environmental data: Ocean temperature, salinity, dissolved oxygen, chlorophyll, net primary productivity, SSH, and MLD at multiple depth layers (from Copernicus Marine Service)

Requirements

The analysis was developed with R ≥ 4.3. Key packages include:

• Modelling: xgboost, mlr3, mlr3tuning, mlr3learners, shapviz, mgcv, nlme
• Data wrangling: tidyverse, lubridate
• Visualisation: ggplot2, patchwork, colorspace, metR, maps
• Utilities: car (VIF), rvest (ONI web scraping), ragg + svglite (plot export)

License

This project is licensed under the MIT License. See LICENSE for details.