How weecology/everglades-forecasting works
This repository forecasts Everglades wading-bird populations by preparing configured species and spatial data, fitting selected models across sliding time windows, and evaluating numeric and ordinal forecast skill. It also documents forecast visualizations and a latest-year forecast-versus-actual view. The README supports the main workflow and model/evaluation relationships; source excerpts were unavailable, so uncertain implementation wiring and peripheral scripts are omitted.
Main entry points: evaluation.R Sliding-window evaluation · main.R Pipeline runner · config.yml Run configuration · data_functions.R Data preparation
Components 15
Forecast workflow
- Run configuration
config.yml - Pipeline runner
main.R - Data preparation
data_functions.R
Dynamic GAM models
- Long-term baseline
models/mvgam_baseline.R - AR smooth covariates
models/mvgam_ar.R - AR linear covariates
models/mvgam_ar_exog.R - AR interactions
models/mvgam_ar_exog_plus.R - Species-specific effects
models/mvgam_species_specific.R - Trait-weighted effects
models/mvgam_trait.R - Trait and guild effects
models/mvgam_trait2.R
Time-series models
- Fable model set
models/fable_models.R
Evaluation and outputs
- Sliding-window evaluation
evaluation.R - Skill visualizations
plotting.R - Latest-year plots
final_year_plots.R
Other components
- Forecast user
Connections 14
- Forecast user → Pipeline runner: starts run
- Pipeline runner → Run configuration: reads settings
- Pipeline runner → Data preparation: prepares data
- Pipeline runner → Sliding-window evaluation: runs windows
- Sliding-window evaluation → Long-term baseline: fits selected
- Sliding-window evaluation → AR smooth covariates: fits selected
- Sliding-window evaluation → AR linear covariates: fits selected
- Sliding-window evaluation → AR interactions: fits selected
- Sliding-window evaluation → Species-specific effects: fits selected
- Sliding-window evaluation → Trait-weighted effects: fits selected
- Sliding-window evaluation → Trait and guild effects: fits selected
- Sliding-window evaluation → Fable model set: fits selected
- Sliding-window evaluation → Skill visualizations: summarizes skill
- Pipeline runner → Latest-year plots: generates forecasts