The model
Decision units
A segment is the decision unit. A segment carries a corridor position, length, width, area, site and vegetation attributes, treatment history, access context, and observation history.
Spans and treatment units are runtime aggregates. They let tactical and operational plans group segments for access, contracting, and scheduling without changing the underlying decision unit.
Vegetation state
A vegetation state is a distribution over interpretable classes plus continuous summaries. The state is not a single height or a single cover percentage. Growth changes the distribution; treatment applies a method- and season-specific transition.
The low-vegetation model is a Brushline state-transition core, not a generic forest growth-and-yield model. Its parameters, evidence, uncertainty, applicability domain, and out-of-domain rejection are documented in Low-vegetation model card and Low-vegetation growth.
Treatments
A treatment is a transition with method-specific efficacy, selectivity, top-kill, resprouting, regrowth delay, and cost. It is not a reset to zero. The transition and calibration contracts are in Treatment transitions and Treatment calibration.
Planning scales
Strategic plans choose where and when to treat over a multi-year horizon, subject to budget, capacity, service, and risk objectives.
Tactical plans turn selected segment actions into treatment units, contractor demand, access windows, and multi-year budgets.
Operational plans turn tactical demand into work orders, crew and machine assignments, and a shift calendar.
The exact stateful MILP is suitable for small instances. Larger instances use Fresh Solutions heuristics and cluster repair. Exactness, heuristic status, and global optimality are different claims; see Stateful strategic MILP and Stateful heuristic.
Objective and risk
The prototype reports cost, crew-hours, service, and a length-weighted mean risk. The risk metric is a model quantity, not an asset-owner risk standard. A lower number does not certify clearance, reliability, fire behaviour, or safety.
What this model does not do
It does not replace a field survey or a LiDAR/field stem map.
It does not certify access, outage, clearance, or safety constraints.
It does not claim that a synthetic calibration is a field calibration.
It does not claim global optimality for the heuristic paths.