Time series analyses for GBR resilience using ADRIA - CoralBlox reef ecosystem modelling suite.
- Clustering reef timeseries to identify what characteristics define resilient clusters.
- Assessing the characteristics that influence positive carbonate budget maintainence into the future.
Full recreation of this project requires that users have Julia v1.11.5 and Quarto installed. Set up the project in the usual Julian way by instantiating the project.
Start Julia from the project root (the main project directory, where the README file sits).
julia --project=.Followed by:
] instantiateThe above will setup the project and install all dependencies.
Note that the specific version of Rasters.jl and Makie.jl currently has an error preventing precompilation of the Makie extension for Rasters.jl (RastersMakieExt.jl). An error message will appear, but can be safely ignored as we do not rely on this extension for any visualizations.
All scripts are expected to be run from the src directory.
The current directory can be changed without leaving Julia by hitting semi-colon.
; cd srcScripts are run with the include method, for example:
include("ADRIA/HighResCoralStress/1_GCM_model_runs.jl")See further notes under the "Analysis scripts" section below.
Repository structure:
ADRIA-reef-indicators/
├─ src/ # Analysis scripts
├─ outputs/ # ADRIA results sets and analysis / figure outputs
├─ data/
├─ paper/ # Reproducible quarto manuscript via the rendering of .qmd file
├─ .gitignore
├─ config.toml # configuration file for project
├─ Project.toml # Julia project spec
├─ Manifest.toml # Julia project spec
└─ README.md # this file
The ADRIA Domain data package GBR_2024_10_15_HighResCoralStress is required to run ADRIA
for this project.
The path to this domain folder should be specified in the config.toml file.
The following structure and files should be maintained in the outputs, figs and data folders:
ADRIA-reef-indicators
│───outputs
│ └───ADRIA_results
│ └───HighResCoralStress
│ ├───processed_model_outputs # Folder containing scenario-median timeseries arrays for each GCM
│ │ ├───median_cover_ACCESS-CM2.nc
│ │ ├───median_cover_ACCESS-ESM1-5.nc
│ │ ├───median_cover_EC-Earth3-Veg.nc
│ │ ├───median_cover_GFDL-CM4.nc
│ │ └───median_cover_NorESM2-MM.nc
│ ├───HighResCoralStress_ADRIA_scens_ACCESS-CM2.csv # CSV files containing the ADRIA-CoralBlox input parameter scenarios used in 1_.jl for each GCM
│ ├───HighResCoralStress_ADRIA_scens_ACCESS-ESM1-5.csv
│ ├───HighResCoralStress_ADRIA_scens_EC-Earth3-Veg.csv
│ ├───HighResCoralStress_ADRIA_scens_GFDL-CM4.csv
│ ├───HighResCoralStress_ADRIA_scens_NorESM2-MM.csv
│ ├───clustered_reefs_carbonate.gpkg # Reef spatial data with cluster timeseries information attached (before context layers are attached in 3_.jl).
│ └───analysis_context_layers_carbonate.gpkg # Geopackage containing the required timeseries clustering, connectivity and carbonate budget values for each reef
│───figs # Directory directly containing general GCM-wide and paper-methods figures, as well as GCM subdirectories
│ ├───ACCESS-CM2
│ │ ├───bioregion
│ │ ├───gbr
│ │ └───management_area
│ ├───ACCESS-ESM1-5
│ │ ├───bioregion
│ │ ├───gbr
│ │ └───management_area
│ ├───EC-Earth3-Veg
│ │ ├───bioregion
│ │ ├───gbr
│ │ └───management_area
│ ├───GFDL-CM4
│ │ ├───bioregion
│ │ ├───gbr
│ │ └───management_area
│ └───NorESM2-MM
│ ├───bioregion
│ ├───gbr
│ └───management_area
└───data
├─ GBRMPA_Management_Areas.gpkg # File containing polygons of GBRMPA management areas
├─ GBRMPA_Reef_Features.gpkg # File containing polygons of GBRMPA reef feature data
└─ GBRMPA_Reefal_Bioregions.gpkg # File containing polygons of GBRMPA reefal bioregions.
Additionally, if available, ADRIA Result Sets should be located in the outputs/ADRIA_Results/HighResCoralStress/ folder.
Note that scripts 1, 2 and 3 are only required to be run if creating model outputs from scratch.
Script 1 takes > 6h to run on 6 threads, therefore we recommend accessing the preprocessed
outputs and running from script 2_.jl onwards.
If median cover timeseries and analysis_context_layers_carbonate.gpkg are already available
in the relevant output structure, then scripts 4 and 5 can be run immediately.
If figures are already available and analysis_context_layers_carbonate.gpkg and the GBR-wide
domain package are available, then the paper.qmd document can be rendered immediately.
Workflow:
flowchart LR;
A[Start] --> B{All publication figures produced?};
B -->|Yes| C[Render paper .qmd file];
B -->|No| D{median_cover_GCM timeseries and analysis_context_layers_carbonate available?};
D -->|Yes| E[Run scripts 4 and 5];
E --> C;
D -->|No| F[Run scripts 1, 2 and 3];
F --> E;
E --> C;
1_GCM_model_runs.jl: Script to run ADRIA-CoralBlox with the required ADRIA Domain and produce processed median coral cover outputs.2_timeseries_clustering_absolute.jl: Perform timeseries clustering analysis on reef-area-scaled absolute coral cover. Analysis is performed at bioregion, management area and GBR wide scales. Cluster assignments for each reef are saved asanalysis_context_layers_carbonate.gpkg3_collating_context_layers.jl: Attach the required values for each reef for analysis factors including connectivity metrics, reef DHW levels and perform carbonate budget analysis for a range of carbonate budget live coral cover thresholds.4_analysis_plots_clustering_carbonate.jl: Create required results for paper using timeseries data and analysis factors collated in3_*.jl. Additionally, calculate the proportion of reefs that change cluster assignment across the GCM levels and the proportion of reefs with depths of 1-10m that occur in low and medium coral cover clusters.5_publication_plots.jl: Create extra required plots for publication such as a context map and diagram of GCM Equilibium Climate Sensitivity values.6_quantifying_cluster_differences.jl: Quantify the differences between cluster timeseries across management areas and GCMs. (Does not need to be run as it is include()ed in the .qmd file).7_rf_analysis.jl: Perform Random Forest analysis on bioregion scale clustering assignments using reef-characteristics as predictors such as depth and connectivity, producing plots for the paper.
GBR-wide ADRIADomain data with HighResCoralStress DWHs: https://registry.mds.gbrrestoration.org/item/102.100.100/708080GBRMPA management areas: https://geohub-gbrmpa.hub.arcgis.com/datasets/management-areasGBRMPA reef features: https://geohub-gbrmpa.hub.arcgis.com/datasets/GBRMPA::great-barrier-reef-featuresGBRMPA reefal bioregions: https://geohub-gbrmpa.hub.arcgis.com/datasets/GBRMPA::reefal-marine-bioregions
ADRIA Result Set outputs from script 1 (100GB total size!): https://registry.mds.gbrrestoration.org/item/102.100.100/708082pregenerated intermediate outputs from scripts 2 & 3: https://registry.mds.gbrrestoration.org/item/102.100.100/708366pregenerated figure outputs: GitHub /figs/ folder
Using GBR-wide ADRIA Domain.
- Initial coral cover data from RME
- DHW timeseries data from HighResCoralStress.org (Dixon, A.M, Forster, P.M., Heron, S.F., Stoner, A.M.K., Beger, M. (2021). Future loss of local-scale thermal refugia in coral reef ecosystems. PLOS Climate, 1(2), e0000004, doi: 10.1371/journal.pclm.0000004)
- Reef spatial data from RME
- GBR1 connectivity data
Timeseris clustering analyses are performed at bioregion, management area and GBR-wide scales. Only bioregion and management area analyses are carried forward into sections of the paper as findings are the same across the spatial scales.
Timeseries are clustered based on their complexity invariant distance (CID). The CID corrects the Euclidean distance between pairs of timeseries (which accounts for magnitude differences) with the complexity of each timeseries (obtained through a measure of their variability) to ensure reefs are clustered based on their underlying temporal dynamics rather than simply their average coral cover. Reefs are clustered into 3 clusters based on these values, with clusters being assigned a low, medium or high value based on their median-cluster coral cover from 2030-2060.
Reef health can be assessed by estimating the carbonate budget status. A positive carbonate budget means that carbonate accumulation processes outweigh erosion and corals can build the skeletal structures important to ecosystem functioning. We can estimate the number of years a reef is predicted to have a positive carbonate budget but assuming a live-coral-cover threshold above which a reef has a positive carbonate budget. As this threshold is uncertain we explore a range of values and compare the results in analyses.