---

Data, R code, and analysis supplement to:

Title: Environmental conditions and herbivore biomass determine coral reef benthic community composition: implications for quantitative baselines.

Authors: James P.W. Robinson, Ivor D. Williams, Lauren A. Yeager, Jana M. McPherson, Jeanette Clark, Thomas A. Oliver, Julia K. Baum

Coral Reefs, 2018. We are open access here!

---

The following R packages were used to analyse data:

install.packages(c("brt", "gbm", "spdep"))

---

data/CREP_predictor_df.csv are the response and predictor covariates for all results

scripts/autocovariates.R estimates autocovariates to reduce spatial autocorrelation

scripts/BRT_optimize.R fits and optimises BRTs for the full dataset

scripts/BRT_optimize_decoupled.R fits and optimizes BRTs for uninhabited and inhabited island datasets (decoupling analysis)

Be aware that BRT optimization functions are essentially loops of boosted models. These can be computationally intensive and will eat your computer if you aren't careful. Please refer to Elith and Leathwick (2017) for details on BRT approaches used here.

---

In the results folder, Rdata files contain optimum BRTs for each analysis. File names indicate the response variable and fitted dataset:

• Response variable
  • log10ratiopq = reef-builder index
  • reefbuilderpq = calcifier cover
  • fleshypq = fleshy algal cover
• Dataset
  • pop = inhabited islands
  • nopop = uninhabited islands

---

In the data/ folder, CREP_predictor_df.csv contains all response and explanatory covariates analyzed. Autocovariate csvs contain all estimate autocovariates fitted in BRTs. Predictor dataframe column names are:

• site.id = site name by island (letters) and site (numbers)
• lat = latitude
• lon = longitude
• source = dataset
• unique.site = unique survey in space and time
• ISLAND = island
• STATE = uninhabited or inhabited island
• COUNTRY = island archipelago
• min_SST = minimum SST (celsius)
• prod = net primary productivity (mg C m-2 day-1)
• complexity = structural complexity
• IslType = island type (atoll, low or high island)
• wave = wave energy (KWhr m-1)
• arag = aragonite saturation state
• depth = site depth (metres)
• cropper = cropper biomass (kg/ha)
• browser = browser biomass (kg/ha)
• scraper.excavator = scraper and excavator biomass (kg/ha)
• total_herb = total herbivore biomass (kg/ha)
• ratiopq = untransformed reef-builder index
• fleshypq = fleshy algal cover (proportion of 1)
• reefbuilderpq = reef-builder cover (proportion of 1)
• log10ratiopq = log10 reef-builder index (main response variable)