v10.0

v10.0 - Zenodo archival & manuscript sent for review

What's Changed

• Upgrade SimpleSDMLayers to v0.6.0 and DataFrames to v1.1.1 by @gabrieldansereau in #29
• Temporary fix for the land cover layers dimensions by @gabrieldansereau in #30
• Load landcover data with geotiff & Copernicus type by @gabrieldansereau in #31
• Upgrade to SimpleSDMLayers v0.7.0 by @gabrieldansereau in #32
• BlueStyle by @gabrieldansereau in #34
• Housekeeping by @gabrieldansereau in #35
• Code edits for thesis corrections by @gabrieldansereau in #36

Full Changelog: v9.0...v10.0

v9.0

v9.0 - Preprint & M.Sc thesis release

Main features

• This version matches the ms repo v1.0 with the preprint on EcoEvoRxiv and the thesis repo v1.0 with the initial thesis submission.
• Also contains the content used in my M.Sc seminar on March 26th
• Added difference, residuals, autocorrelation & rare species analyses
• Started using absolute LCBD values instead of relative ones & reformatted figures to display scientific notation

Merged pull-requests

• Upgrade to Julia 1.6.0 & remove module (#25)
• Additional analyses (#26)
• Analysis feedback (#27)
• Cleaning up figures (#28)

Other

• Added GLM.jl to Project environment and made some GLM attempts
• Added step-by-step scaling medians figures
• Upgraded to Julia v1.6.1 & upgraded packages
• Removed outdated module calls
• Added total beta diversity on plots using custom rectangle() function
• Created functions to scale the colorbar on heatmaps
• Added difference, residuals, autocorrelation & rare species analyses
• Added broom, MASS, betareg, SpatialPack to required R packages
• Tweaked rlimit to avoid memory overloads
• Added script to export all values mentioned in the manuscript
• Added subtitles on figures
• Archived unused figures & code

v8.0

v8.0 - Pre MS feedback version

Main features

• Moved the manuscript in development to gabrieldansereau/ms_betadiversity_hotspots
• This version of the codes matches the ms repo v0.1 sent to Tim & Pierre for feedback
• Includes the changes made for the manuscript and for my presentation at QCBS2020

Merged pull-requests

• Set up renv for R package management (#20)
• Upgrade to DataFrames v0.22 & CSV v0.8 (#21)
• QCBS 2020 presentation (#22)
• Clean up manuscript from this repo (#23)
• Test SimpleSDMLayers v0.3.6 (#24)

Others

• Upgraded to Julia v1.5, SimpleSDMLayer v0.3.6
• Started using renv for R package management similar to Julia
• Removed Julia parallelization
• Tested tidymodels with RandomForests
• Updated article figures, citations format, draft before moving to the ms repo
• Updated RandomForest figures (but not used)
• Added VS Code Workspace
• Added a bunch of figures for QCBS 2020
• Added Makefile entry for article pdf
• Minor tweaks to figures: colorbar label, margins, added medians, total beta,
• Added scaling medians figure

v7.0

v7.0 - Upgrade to SimpleSDMLayers v3.0

Main features

• Upgraded to SimpleSDMLayers v3.0, which required major rewrites to allow type Nothing and changed the climate data used to WorldClim 2.1. (#18)
• Changed the default branch from master to main
• Started working on an article draft (#17)

v6.1

v6.1 - BARTs

Main features

• Implemented BARTs (Bayesian additive regression trees) models in R with package embarcadero (#15)
  • Full-scale, all species distribution models
  • Also QC-scale, single species attempts while implementing models
  • Results for all previous analyses (richness, LCBD, relationship, subareas, moving-windows)
• Implemented direct BART models on species richness & LCBD value (#16)
  • Same workflow for BART & RF
  • Separate model (R) & visualization (Julia) scripts
• Implemented functions to apply operations directly on SimpleSDMLayer elements (e.g. max, min, + , -, mean)

Others

• Switched to tidyverse syntax & style in R
• RData files moved to separate folder
• Attempt at using a module for library scripts (implemented, but then reverted)
• Started using VS Code as primary editor for both Julia and R (required some minor style changes)
• Renamed exploration scripts

v6.0

v6.0 - Moving windows & major restructuring

Main features

• Major repository restructuring (see #13 and #14 for details)
• Implemented moving windows algorithm
• Uploaded JLD2 data as ZIP archive
• Updates of pretty much everything

v5.2

v5.2 - Subareas & richness RF predictions

Subarea analysis

• Performed LCBD & relationship analyses on Northeast (NE) and Southwest (SW) subareas
• Performed analyses on expanding subareas (result in GIFs)
• Compared 3 scales to show the change in the relationship when calculating LCBDs at larger scales
• Created library script for analysis functions

Richness Random Forests predictions

• Attempt at predicting species richness directly with Random Forest based on observed species richness

Others

• Renamed repository from BioClim to betadiversity-hotspots
• Specified file paths with joinpath()
• Minor changes to BIOCLIM functions
• Attempt to apply BIOCLIM on training data
• Tweaked default plot() recipe to have result similar to plotSDM(), but faster

v5.1

v5.1 - Evaluation attempt & minor fixes

Model evaluation attempts

• Train Random Forests in Julia
• Partition data into training & validation sets, stratified by presence-absence
• Evaluate a few performance metrics
• Attempt plotting & calculating AUC

Minor fixes

• Now using PNG as default figure format
• Updated SimpleSDMLayers version

v5.0

v5.0 - Random Forests

• Trained Random Forests in R to predict presence-absence
• Applied analyses on RF results

v4.1

v4.1 - Global landcover data, BIOCLIM update & quantiles

Global landcover data

• Now using global landcover data, downloaded from Zenodo
• Implementing 10 min resolution for prediction, 5 min for training (for real this time)
• Data download & coarsening through Bash scripts

BIOCLIM function update

• Fixed problem with model returning no predictions when one variable had zero land cover at all observed sites
• Updated inner function syntax

Quantiles

• Built option to view results as quantiles instead of raw/relative scores
• Added quantiles figures in separate folder for now, might become main versions at some point

Others

• Briefly explored weighted endemism metric
• Minor fixes