Process scenario ensembles ========================== Purpose ------- ``ws4.ecosystem.ensemble`` runs the actual process engine outside the optimizer: fire, succession, salvage, regeneration, and the provisional volume bridge, one executor per scenario seed. It produces per-year, account-like metrics and optionally risk-aggregates them with the existing scenario layer. This is a process-driven ensemble, not the P181 account-overlay optimizer path. Metrics ------- Each year records: - ``fire.count``, ``fire.burned_area_ha``, ``fire.mortality_area_ha``; - ``succession.count``, ``salvage.count``, ``regeneration.count``; - ``state.biomass_t``, ``state.fuel_t``, ``state.snag_t``; - ``bridge.volume_m3``. Usage ----- .. code-block:: python from ws4.ecosystem import ( ProcessEnsembleConfig, run_process_ensemble, ScenarioSet, ) result = run_process_ensemble( state, edges, ProcessEnsembleConfig(horizon_years=20, fire=fire_config), scenarios, targets, ) result.write("runtime/process-ensemble") Artifacts are ``process_values.csv``, ``process_aggregated.csv``, and ``process_summary.json``. Fire-regime report ------------------ ``ProcessEnsembleResult.fire_regime_summaries`` applies the P188 event-log summary to each scenario's replayable log. It returns a mapping of scenario id to a fire-regime risk summary, including annual burned area, burn probability, FRI, escaped fires, and size quantiles. The layer is distributional; it is not an observed-perimeter comparison or a calibration step. ``ProcessEnsembleResult.write_fire_regime`` writes the same summaries to ``fire_regime.json`` with top-level schema ``ws4.process-fire-regime.v1``. Limits ------ - The process state is standalone; there is no management-schedule bridge yet. - A harvest schedule does not change fuel before the process runs. - The volume bridge remains provisional. - The optimizer uses the P181 account overlay in its inner loop; P183 is an outside-loop process ensemble.