Operations
==========

**Implementation status:** Partial (coverage matrix row *Operations* — LF
wired per DL-010/017; additive ARM/FAO/ATM present; per-route resolution
absent from the production wedge). The page's printed forms — three
**separate** linear efficiency ramps, the multiplicative
:math:`e^* = (1-\Delta\eta)\,e_{in}`, and the eq. 5 RPTK flight-count
recomputation — are implemented standalone in
``src/engine/operations-full.ts`` (DL-137, Stage-1 shadow; the three
end-values are named in the RST but never valued anywhere in the corpus,
so they are required caller arguments). The production wedge remains
subtractive on an S-curve with a fixed 0.35/0.35/0.30 split — a
documented divergence from the printed linear ramps.
**Source:** ``src/engine/cascade.ts:computeLCA`` (ops wedge, ``lfAt``,
``lfFactorAt``), ``src/engine/constants.ts:OPS_AMBITION``,
``src/engine/operations-full.ts``
**Test evidence:** ``tests/load-factor.spec.ts``,
``tests/suite-m-operations-fuelburn.spec.ts``

Efficiency levers: ARM + FAO + ATM
----------------------------------

Operational fuel-burn improvement is the sum of three categories
(cf. CASCADE *Operations*):

.. math::

   \Delta\eta_{ops} = \Delta\eta_{ARM} + \Delta\eta_{FAO} + \Delta\eta_{ATM}

* **ARM** — Aircraft Retrofit & Maintenance (ambition table 0.4–2.0 %)
* **FAO** — Fleet & Airport Operations (0.8–4.0 %)
* **ATM** — Air Traffic Management (1.0–5.0 %)

The scenario may set each component explicitly (``fb_arm``, ``fb_fao``,
``fb_atm``) or a single total ``fb``, which is split by the fixed category
shares 35 % / 35 % / 30 % (``OPS_CATEGORIES``). The ops wedge in the time
series is :math:`\Delta_{ops} = G_{gross} \cdot (\Delta\eta_{ops}/100)
\cdot s(t)`, where :math:`s(t)` is the normalised adoption curve over the
model window.

Load factor
-----------

Raising passenger load factor conserves revenue-passenger traffic, so fewer
flights deliver the same payload; energy and emissions scale with the
flight count (cf. CASCADE *Operations*, load-factor equations, RPTK
conservation; DL-010/DL-017). Freighters are exempt. The aggregate energy
multiplier at year :math:`y`:

.. math::

   F_{LF}(y) = (1 - f_{freight})\,\frac{LF_{in}}{LF(y)} + f_{freight}

with :math:`LF_{in}` the baseline load factor (``lf_entry_value``, default
85 % — neutral for the frozen baseline; set 82.6 %, the IATA 2019 global
passenger value :cite:`iata_wats`, for CASCADE-reproduction scenarios) and
:math:`f_{freight}` the freighter share of fuel burn
(``lf_freight_share``, default 10 %, literature range 8–12 %). The load
factor itself ramps on an S-curve from ``lf_entry_year`` (default 2024) to
2050:

.. math::

   LF(y) = LF_{in} + (LF_{target} - LF_{in})\, s\!\left(
   \frac{y - y_{entry}}{2050 - y_{entry}}\right)

The lever enters the ops wedge as :math:`G_{gross}\cdot(1 - F_{LF}(y))`
(negative when the target load factor is below baseline — more flights) and
applies in both offset modes; it is exactly zero at defaults so the frozen
baseline is unaffected.

Test evidence (8 tests) pins: neutrality at defaults, frozen-baseline
invariance, monotonicity in both directions, the RPTK-conservation identity
(wedge = gross × pax share × (1 − LF_in/LF)), freighter exemption, ramp
behaviour, and whole-engine conservation.

Contrail avoidance
------------------

**Implementation status:** Extended — CASCADE *application* parity. The RST
set publishes **no** contrail formulation (verified by full-text grep of
the RST set, 2026-07-18); the CASCADE app exposes the strategy with exactly
the inputs modelled here
**Source:** ``src/engine/contrails.ts`` (pure module — **not** imported by
``cascade.ts``), UI in ``src/components/modals/OperationsModal.tsx``
(Contrails tab)
**Test evidence:** ``tests/contrails.spec.ts`` (20 tests)

Avoidance ramps from 0 at ``ca_start_year`` (default 2025) to the target
share of contrail forcing at ``ca_target_year`` (default 2050), then holds
— the same hold-not-stretch convention as the other levers. Any
``CurveShape`` may be selected via ``curve_ca``; the default ``'linear'``
matches the RST default for operations improvements:

.. math::

   r(y) = s\!\left(\frac{y - y_{start}}{y_{target} - y_{start}}\right)
   \in [0, 1], \qquad r = 1 \text{ held beyond } y_{target}

Prediction-success (knowledge) quality scales effectiveness — ``none /
partial / perfect`` = ×0.0 / ×0.6 / ×1.0. These factors are **documented
HyFlux fallback assumptions**: the RST set publishes no numeric
effectiveness factors for the app's knowledge modes (grep-verified):

.. math::

   X_{avoided}(y) = \operatorname{clamp}_{[0,1]}\!\Big(
   \frac{X_{target}}{100}\; r(y)\; k_{knowledge}\Big)

Rerouted flights burn ``ca_fuel_penalty`` % more fuel; the fleet-average
extra burn is the per-flight penalty times the avoided fraction
(conservative — in reality ~10 % of flights cause ~80 % of contrail
warming, so the true rerouted share is smaller), and is netted against the
avoided forcing as a CO₂ cost (fuel-burn % ≈ CO₂ % for the same fuel):

.. math::

   \Delta_{net}\ [\text{frac. of fleet CO}_2] =
   X_{avoided}\cdot R_{CO_2eq} - \frac{X_{avoided}\cdot p_{fuel}}{100},
   \qquad R_{CO_2eq} = \tfrac{2}{3}

``CONTRAIL_CO2EQ_RATIO`` (:math:`R_{CO_2eq}`) expresses contrail-cirrus
forcing as a fraction of aviation CO₂ forcing, anchored to the
``nonco2.ts`` envelope: Jet-A central ERF multiplier 2.0 ⇒ non-CO₂ ≈
1.0×CO₂, of which contrail-cirrus is ~2/3 (Lee et al. 2021 ordering
:cite:`lee2021`). Conceptual, literature-approximate, editable — same
status as ``nonco2.ts``. The net fraction is signed and can go negative
when the fuel penalty dominates.

Defaults are **lever OFF** (``ca_target_share = 0`` ⇒ all outputs exactly
zero), so every existing scenario result is bit-identical. The module is
deliberately pure and is not yet wired into the aggregate sweep — a future
integration multiplies the net fraction by fleet CO₂ to obtain GtCO₂e; the
OperationsModal tab (knowledge picker, start/target years, target share,
fuel penalty, curve shape) exposes the parameters and URL-persists them
today.

Known deviations / limitations
------------------------------

* Per-route operations resolution (fleet-assignment deltas, direct-routing
  and taxi savings per route) is not modelled; all levers are aggregate.
* The contrail lever is an aggregate, display-level quantity: it is not
  wired into the aggregate sweep, and both the knowledge factors (0 / 0.6 /
  1.0) and :math:`R_{CO_2eq} = 2/3` are documented fallback/conceptual
  assumptions, not CASCADE-published values.
* The ambition-table ``AMBITION.ops`` (3–21 %) is treated as ARM+FAO+ATM
  only; load factor is a separate multiplicative lever.
* Curve-shape parameters for the individual ARM/FAO/ATM levers
  (``curve_arm`` etc.) are URL-persisted but the engine currently applies
  one shared adoption curve to the summed total.
* The custom **pulse curve** (CASCADE *Operations*, Custom Curves → Pulse)
  is implemented exactly as ``curves.ts:pulseAt`` and documented under
  :doc:`extensions`; UI exposure remains deferred because the
  ``CurveEditor`` param model does not map onto the pulse parameter set
  (see :doc:`aircraft`).