Skip to content

Latest commit

 

History

History
94 lines (80 loc) · 5.88 KB

README.md

File metadata and controls

94 lines (80 loc) · 5.88 KB
Raw

Seabird assemblages are linked to the major western boundary current off eastern Australia

This .scripts/README guide you through the scripts used in Daudt et al. (2024).

There will be two sessions. Jump straight in modelling if you are interested in the results of the paper. The full pipeline explains the logic behind each script and what it does, before getting the the modelling steps.

General workflow and formatation

The scripts are named following a numerical order, so its is easy to follow (numbers relate to script names):

Fig 1. Scripts workflow

All scripts were developed using RStudio IDE and, for the sake of facility, I structured them into 'code sections'. So, if you use RStudio, you can follow the document outline easily.

Below, I provide a 'pseudo-code' underlying what each script does.

Full pipeline

NOTE: Seabird raw data can not be publicly available at the moment, so scripts '01'--'07' are here for the transparency of the process. Environmental data are all publicly available through the links provided within the manuscript (see Table 1 in the main text). The gridded, tidied-up data after script '06' and '07' are available under ./data_out/, which allows you to reproduce the main results.

  • '00': R environment
    • Sets up R environment by calling renv::restore()
  • '01': Merge and tidy up raw dataset
    • Merge two raw datasets into one
    • Clean columns, standardise some data and create season column
    • Standardise taxonomy (e.g. common names to scientific names, subspecies to species), and clean taxa that won't be used (e.g. egrets, shorebirds)
    • Brief summary of data (e.g. no. voyages, no. records, no. taxa)
  • '02': Create the grids on which data will be aggregated
    • Read and subset only East Australia data
    • Transform it into spatial objects
    • Create the grids (0.5--2° latitude)
  • '03': Extracts environmental data
    • Bathymetry (bat), slope (slope), distance from the coast (dist_coast)
    • Calculate and extract Eddy Kinetic Energy (eke)
    • Extract Chlorophyll-a (chl8) and its log10 (log10_chl8)
    • Crop original Sea Surface Temperature (sst) files and extract values
    • Extract Sea Surface Salinity (sss) and Mixed Layer Depth (mld)
    • Create and extract SST gradient (sst_grad)
    • Create and extract climatic data for SST gradient (clim_sst_grad), Eddy Kinetic Energy mean (clim_eke_mean), Eddy Kinetic Energy standard deviation (clim_eke_sd)
  • '04': Aggregate seabird and environmental data by grids, and create 'final' wide-format dataset (to be used in analyses)
    • Aggregate data by grids
    • Based on the code section "EDA - how many species by grid/season and grid size", we decided on 1° latitude grids would be best
    • Wrangling and tidy up the dataset for future analyses
  • '05': Exploratory data analyses
    • General summary (e.g. no. voyages, no. records, no. taxa, no. individuals recorded)
    • No. of individuals by [species/]grid/season
    • No. of occurrences, Frequency of occurrence, Numeric frequency of species, by season
    • Based on the above, get a vector with species names to remove from models
    • Maps with seasonal sampling, no. of species (species richness), no. of birds
  • '06': Prepares data to fit RCP models
    • Split data by season, scale environmental data (by season) and check correlation between them
  • '07': Creates average climatic seasonal, static layers for all environmental (data for prediction)
    • Create SST layers
    • Create SSS and MLD layers
    • Create CHL layers
    • Crop and extract data for each season
    • Note: 'clim_eke_sd', 'clim_eke_mean' (consequently, 'eke'), and 'clim_sst_grad' (consequently, 'sst_grad') were already created under script '03'
  • '08_1': Fit Bernoulli RCP models
    • See below
  • '08_2': Fit Negative Binomial RCP models
    • See below
  • '09': Prepare Figures for publication
    • Use {patchwork} to put together figures for publication

Modelling

  • '08_1': Fit Bernoulli RCP models
    • Transform abundance data into binary 0/1 (absence/presence)
    • Choose the number of groups (RCPs) based on ecomix::regional_mix.multifit()
    • Fix number of RCPs and run ecomix::regional_mix()
    • Choose the best model based on the optimal covariates
    • Model diagnostics
    • Bootstrap the best model to get uncertainty in model parameters
    • Results: plot species profiles
    • Results: predict and map point-predictions
    • Results: predict and map uncertainty
    • Results: partial plots relating RCP probability agains environmental data
    • Results: diversity rarefaction curves based on {iNEXT}
  • '08_2': Fit Negative Binomial RCP models
    • Choose the number of groups (RCPs) based on ecomix::regional_mix.multifit()
    • Fix number of RCPs and run ecomix::regional_mix()
    • Choose the best model based on the optimal covariates
    • Model diagnostics
    • Bootstrap the best model to get uncertainty in model parameters
    • Results: plot species profiles
    • Results: predict and map point-predictions
    • Results: predict and map uncertainty
    • Results: partial plots relating RCP probability agains environmental data

Manuscript

After running the above steps, you will get all the results needed to compile the manuscript. The source RMarkdown files are in ./ms_PROOCE. Note, however, as explained in the main README, the higher quality Figure 1 from the manuscript was built outside R (although a lower quality version is available through the codes).