Treatment transitions

A treatment is a transition, not a reset to zero. The P4.3 model applies a method- and season-specific top-kill and cover reduction, changes density through resprouting, records a regrowth delay, and runs the P4.2 growth model for the post-delay trajectory. The efficacy parameters are priors; P4.4 calibrates them.

Input

A TreatmentRequest carries:

  • the current state distribution and functional group;

  • the method and season;

  • operational intensity;

  • treated area for cost;

  • optional regrowth horizon;

  • a growth environment.

A TreatmentConfig carries residual floors, season factors, growth rules, discretization rules, and the maximum regrowth horizon.

Immediate transition

For one treatment:

efficiency = min(1, base_efficacy * season_factor * intensity)
height_after = max(min_height, height - height * efficiency * top_kill)
cover_after = max(min_cover, cover - cover * efficiency)
density_after = density * (1 + resprout_factor * efficiency)

The residual floor prevents a zero reset. A density increase records resprouting. The immediate age is zero because the treatment is a disturbance. Immediate uncertainty grows with efficiency.

The immediate class is recomputed from the new height and cover. A residual that does not match a class becomes partial.

Post-delay trajectory

When regrowth_years exceeds the method’s regrowth_delay_years, the model grows the immediate state for the remaining years with the supplied GrowthEnvironment. A blocked growth trajectory leaves the immediate state in place and records the reason.

from brushline.vegetation.growth import GrowthEnvironment
from brushline.vegetation.state_space import StateDistribution
from brushline.vegetation.treatment import (
    TreatmentConfig,
    TreatmentRequest,
    apply_treatment,
    validate_treatment_result,
)

current = StateDistribution(
    class_probabilities={"shrub": 0.8, "unknown": 0.2},
    height_m=2.0,
    height_sd=0.2,
    cover_pct=30.0,
    cover_sd=5.0,
    density_per_ha=1000.0,
    age_years=3,
    confidence=0.8,
)
result = apply_treatment(
    TreatmentRequest(
        segment_id="s1",
        functional_group="shrub",
        current=current,
        method="mowing",
        season="spring",
        intensity=1.0,
        quantity_ha=2.0,
        regrowth_years=6,
        environment=GrowthEnvironment(),
    ),
    TreatmentConfig(),
)
assert validate_treatment_result(result).is_valid

Methods and priors

Method

Efficacy

Selectivity

Top-kill

Resprout

Delay

Cost/ha

manual

0.55

0.90

0.70

0.50

1 year

1,200

mechanical

0.75

0.70

0.90

0.60

1 year

900

mowing

0.65

0.80

0.80

0.70

1 year

500

chemical

0.80

0.60

0.60

0.30

2 years

700

grazing

0.35

0.40

0.40

0.40

0 years

300

prescribed_fire

0.70

0.30

0.70

0.60

2 years

1,100 + 2,500 fixed

other

0.30

0.50

0.30

0.50

1 year

0

Season factors are spring 1.0, summer 0.9, fall 0.8, winter 0.6, and any 1.0. These are priors and must be calibrated in P4.4.

Cost

The cost is:

cost = cost_per_ha * quantity_ha + fixed_cost

The result records the cost and currency. It is an estimate hook, not a bid.

Output

TreatmentResult carries:

  • previous, immediate, and trajectory distributions;

  • immediate and trajectory state classes;

  • height and cover reduction, density change, and effective efficiency;

  • cost, delay, horizon, environment, and reason;

  • model card id.

to_vegetation_state(state_year) uses the trajectory if present, otherwise the immediate state.

Named errors

Prefix

Meaning

TREATMENT-STATUS

Unknown result status.

TREATMENT-NEXT

An applied result has no immediate state, a blocked result has one, or a trajectory has no class.

TREATMENT-HEIGHT, TREATMENT-COVER

Cutting increased height/cover or a reduction is negative.

TREATMENT-RESIDUAL

Height was reset to zero or cover is negative.

TREATMENT-EFFICIENCY

Efficiency is outside 0 to 1.

TREATMENT-COST

Cost is negative.

TREATMENT-REASON

A blocked result has no reason.

Limits

  • The parameters are priors, not calibrated values.

  • The model does not simulate chemical mode of action, fire behaviour, or grazing preference.

  • The cost is a hook for a later bid build-up, not a contractor price.

  • The model does not decide whether a treatment is safe or permitted.