Re-simulate at the calibrated parameter vector for goodness-of-fit plots

Runs n_reps replicate Dynamic Fire simulations at the calibrated parameter vector and returns their per-replicate fire statistics plus the loss against the observed targets, so a calibration report can plot the goodness-of-fit that calibrate_dynamic_fire() does not retain.

Usage

run_calibration_validation(
  scenario_template,
  best_params,
  observed_targets_path,
  cfg,
  n_reps = 20L,
  scratch_root,
  base_seed = 99999L
)

Arguments

scenario_template

Character path to scenario.txt inside the calibration scenario directory.

best_params

Named numeric vector of calibrated parameters (the calibrate_dynamic_fire() result's best_params).

observed_targets_path

Character path to the saved observed-targets .rds (the save_observed_fire_targets() output); reattached to the return value so the report has a single self-contained input.

cfg

The calibration_config list (for sim_years, weights).

n_reps

Integer number of replicate simulations at best_params.

scratch_root

Docker-visible host directory for the warm pool bind-mount (distinct from the main calibration scratch so concurrent runs do not collide).

base_seed

Base seed for the validation reps (base_seed + i per rep).

Value

A list with reps, best_params, observed, pixel_area_ha, sim_years, n_reps, loss, pool_image, and pool_digest.

See also

Other Dynamic Fire calibration helpers: apply_calibrated_damage_age(), apply_calibrated_hi_prop(), apply_calibrated_ignprob(), apply_calibrated_num_fires(), bc_fuel_code_to_base(), build_calibration_scenario_template(), build_calibration_spinup_scenario(), calibrate_dynamic_fire(), calibration_par_names(), dedup_community_snapshot(), default_severity_prior_sturtevant2009(), landis_overstory_mortality_share(), loss_from_stats(), observed_fire_sizes(), parse_dynamic_fire_logs(), patch_fire_config(), run_calibration_spinup(), save_observed_fire_targets(), sim_landis(), sim_mock(), sim_r_reimpl()