Loading Results

Loading ADRIA Results

Results produced by ADRIA.run_scenarios are stored on disk and can be reloaded at any time using their path:

rs = ADRIA.load_results("path/to/result_set")

The returned rs is an ADRIAResultSet that gives access to everything needed for analysis and visualization.

Key fields

Field	Description
rs.name	Name of the result set (derived from the domain name and run time)
rs.RCP	RCP scenario string (e.g. "45")
rs.inputs	DataFrame of scenario inputs used for the run
rs.model_spec	DataFrame describing all model parameters and their bounds
rs.outcomes	Dict of named outcome arrays (coral cover, shelter volume, etc.)
rs.ranks	Location ranking log; dims (timesteps, locations, intervention, scenarios)
rs.seed_log	Seeding deployment log; dims (timesteps, locations, species, scenarios)
rs.shading_log	Fogging/shading log; dims (timesteps, locations, intervention, scenarios) where intervention is ["fog", "shade"]
rs.mc_log	Moving-coral (assisted migration) log
rs.coral_dhw_tol_log	Per-location DHW tolerance trajectories (only populated when log_dhw_tols = true in config.toml)
rs.coral_cover_log	Raw coral cover for all size classes (only populated when log_cover = true in config.toml)
rs.loc_ids	Location identifiers
rs.loc_area	Location areas (m2)
rs.loc_centroids	Location centroid coordinates
rs.loc_data	DataFrame with spatial attributes for each location
rs.dhw_stats	Summary statistics for the DHW projections used
rs.wave_stats	Summary statistics for the wave stress projections used
rs.connectivity_data	Connectivity matrix data

Accessing outcomes

Individual outcome arrays can be extracted directly from rs.outcomes or via the ADRIA.metrics.* functions (preferred):

# Via the metrics API (recommended)
tac = ADRIA.metrics.total_absolute_cover(rs)
rsv = ADRIA.metrics.relative_shelter_volume(rs)

# Directly from the outcomes dict
rs.outcomes[:relative_cover]

See Running scenarios and the Metrics page for more detail on available metrics and result set properties.

Loading ReefModEngine Results

Results from ReefModEngine.jl can be loaded with the load_results function.

rs = ADRIA.load_results(RMEResultSet, "<path to data dir>")

Expected data directory structure:

data_dir
|
+---con_bin
|       CONNECT_ACRO_2010_11.bin
|       CONNECT_ACRO_2011_12.bin
|       ...
|
+---id
|       id_list_2023_03_30.csv
|
+---region
|       reefmod_gbr.gpkg
|
+---results
        results.nc
        scenarios.csv

To reduce duplication of geospatial and connectivity data, the data directory and results directory can be supplied separately to avoid keeping copies for each result set analysed.

rs = ADRIA.load_results(RMEResultSet, "<path to data dir>", "<path to results dir>")

Loading C~scape Results

Results from C~scape can be loaded with the load_results function.

The first argument is always the C~scape data package directory — the folder holding ScenarioID.csv plus the connectivity/, site_data/ and initial_cover/ subdirectories (see the tree below).

rs = ADRIA.load_results(
    CScapeResultSet, "<path to C~scape data package>";
    result_dir="<path to result NetCDF directory>",
    result_files=["NetCDF_Scn_140001.nc", "NetCDF_Scn_142162.nc"],
    show_progress=true
)

All keyword arguments are optional:

• Omit result_dir and the NetCDFs are read from the data package's own results/ subdirectory. Set it to point at NetCDFs kept outside the data package.

• Omit result_files and every NetCDF_Scn_* file in result_dir is loaded. Set it to a list of NetCDF paths to load only those; result_dir is then ignored.

• show_progress (default true) toggles the progress bar shown while outcomes are computed.

Expected C~scape data package structure (the directory passed as the first argument):

cscape_data_package
|   ScenarioID.csv
|
+---connectivity
|       connectivity.csv
|
+---site_data
|       geospatial_data.gpkg
|
+---initial_cover
|       initial_cover.csv
|
+---results (optional)
        NetCDF_Scn_140001.nc
        NetCDF_Scn_140002.nc
        ...

C~scape data package

Most of these files can be sourced from the RRAP data store (published dataset names in italics below). They are usually distributed as R .Rdata objects or write.table text and need light reformatting into the CSV / GeoPackage layout shown above.

File	Contents	Data store source
ScenarioID.csv	One row per scenario: input parameters, intervention settings and the datasets each run used. The ID column matches the NetCDF_Scn_<ID> result files.	not yet published
connectivity/connectivity.csv	Larval connectivity matrix between locations, with reef_siteid row and column labels.	Spatial inputs - Moore cluster 2022 v2 (MEAN_all_Connectivity_MooreReef_cluster_221019.Rdata)
site_data/*.gpkg	Location polygons and their spatial attributes (reef_siteid, k, area, depth, ...). The first .gpkg found in the folder is used.	Spatial inputs - Moore cluster 2022 v2 (MooreReefCluster_Polygon_Geometry.Rdata)
initial_cover/initial_cover.csv	Initial coral cover per location and functional group.	Coral Cover Initialisation data inputs - C~scape - Counterfactuals Mar 2024

C~scape model outputs

The results directory holds one NetCDF per scenario. Files must contain the NetCDF_Scn_<ID> prefix to be discovered automatically.

The full model output set is large (~100 GB). RRAP M&DS publishes instructions for downloading it via the AWS CLI on the Model Outputs data store page (download tab).

Accessing C~scape outcomes

Only relative cover is loaded automatically. All other outcomes are computed on demand via the ADRIA.metrics.* functions and cached in rs.outcomes:

settlers = ADRIA.metrics.total_settlers(rs)
rs.outcomes[:total_settlers]  # now cached

---

This page was generated using Literate.jl.