Package index • taupatch

Running a study

A study is a config file plus data. Point-and-click through the Shiny app or drive the same pipeline from YAML; both take the same route.

compare_runs(): Is the gap between two model runs real?
generate_config(): Write a taupatch run config
load_config(): Load and validate a taupatch run config
pipeline_stages(): The stages a run passes through, and how far along each one is
run_taupatch(): Run the full taupatch pipeline
run_taupatch_app(): Launch the taupatch Shiny app
save_config(): Write a config list to a YAML file
single_stages(): Individually resolved life stages

Zooplankton data

Reading plankton counts and getting them into one shape. A patch is defined by a species-and-stage abundance threshold, so which stages a column holds matters as much as the number in it.

abundance_units(): The count and the unit a suffix encodes
available_species(): Species a formatted database could model
available_stages(): List the life stages a database holds for a species
default_species_catalog(): Default species catalog
format_zoop_data(): Build a station database from a raw zooplankton export
is_raw_export(): Whether a header is a raw export rather than a station database
label_patch(): Label high-abundance patches
load_zoop_data(): Load zooplankton station data
measurement_columns(): In-situ measurement columns the raw export carries
raw_abundance_suffix(): The abundance suffix a raw export uses
species_catalog_from(): A species catalog for the taxa in a raw export
split_dates(): Split a date column into year, month, and day
stage_suffix_pattern(): The suffix marking a life stage on a taxon column
station_summary(): Summary of a station dataset
taxon_shorthand(): Conventional shorthand for a taxon name
validate_columns(): Check the config's declared columns against the CSV header
zoop_taxa(): Taxon columns in a raw zooplankton export

Covariates

Fetching environmental fields and joining them to stations, including the derived layers that come from derivoce.

add_derivoce_covariates(): Add the configured derived covariates to the covariate grid
apply_prejoin_steps(): Apply the configured pre-join steps to the fetched products
attach_covariates(): Attach covariates to zooplankton stations
bathymetry_covariates(): Static seafloor covariates
climate_index_covariates(): Climate indices selectable as covariates
copernicus_client(): Locate the Copernicus Marine client
copernicus_covariates(): Catalog of selectable environmental covariates
covariate_grid(): Build the covariate grid for one month
covariate_info(): Covariate reference table
covariate_monthly_means(): Average each covariate over the study area, per month and year
covariate_transforms(): Covariate transformations available in the recipe
default_copernicus_datasets(): Default Copernicus datasets
derivoce_choices(): Derived covariates offerable for a given covariate selection
derivoce_covariates(): Derived covariates computed from the covariate grid
derivoce_required_inputs(): Covariates a set of derived choices needs fetching
derivoce_steps_for(): Config steps for a set of chosen derived covariates
fetch_covariates(): Fetch environmental covariates
generate_mock_covariates(): Generate synthetic environmental covariates
plot_covariate_annual(): Plot a covariate's annual mean over the record
plot_covariate_heatmap(): Plot a month-by-year heatmap of a covariate
plot_covariate_map(): Map one covariate for one month
plot_covariate_seasonal(): Plot a covariate's seasonal cycle, one line per year
prejoin_steps(): Per-covariate steps applied before products are joined

Fitting and projecting

The models themselves, single or ensembled, and the projection of a fitted model onto a new grid.

ensemble_rules(): Ways an ensemble can combine its members
ensemble_settings(): Multi-algorithm ensemble settings
fit_patch_ensemble(): Fit an ensemble of model types on the same data
fit_patch_model(): Fit a patch habitat suitability model
gam_bases(): Spline bases mgcv offers for a smooth
gam_methods(): Smoothing parameter estimation methods mgcv offers
gam_smooth_terms(): Smooth terms of a fitted GAM
glm_coefficients(): Coefficients of a fitted logistic regression
model_engine_fit(): The underlying engine object from a fitted model
model_types(): Model types a run can fit
partial_effects(): Partial effect of each predictor on patch probability
power_curve(): How the comparison would improve with more stations
print(<taupatch_ensemble>): Print an ensemble
project_patch_model(): Project a fitted model to monthly habitat suitability maps

Uncertainty and validation

What the model does not know. Novelty and overlap surfaces say where a projection is extrapolating rather than interpolating.

bootstrap_evaluation(): Bootstrap intervals for the evaluation metrics
jackknife_covariates(): Test every covariate by leaving it out, in parallel
jackknife_dropped(): Which covariates a jackknife would drop
jackknife_settings(): Covariate jackknife settings
novelty_surface(): How far outside the training data each cell sits
thin_covariates(): Thin a covariate grid to a bounded number of cells
uncertainty_settings(): Projection uncertainty settings

Plots and output

Figures for inspecting each stage, and the raster stack written at the end.

plot_calibration(): Plot probability calibration
plot_gam_smooths(): Plot a GAM's fitted smooths, with fancyfx
plot_glm_coefficients(): Plot logistic regression coefficients with their intervals
plot_importance(): Plot variable importance
plot_partial_effects(): Plot partial effects, one panel per predictor
plot_pr_curve(): Plot the precision-recall curve
plot_projection(): Plot a monthly habitat suitability projection
plot_projection_uncertainty(): Plot how far a monthly projection can be trusted
plot_roc_curve(): Plot cross-validated ROC curves
plot_station_map(): Map the stations, coloured by abundance
plot_station_series(): Abundance over the record
plot_threshold_performance(): Plot performance across classification thresholds
projection_map(): Build a leaflet map of a projection GeoTIFF
write_suitability_stack(): Write the monthly projections as one multi-layer raster

Test fixtures

Mock data, so an example or a test can run without a data extract.

generate_mock_zoop_data(): Generate a synthetic zooplankton database