Air Traffic Dataset
===================

**Implementation status:** Simplified with declared bases (coverage matrix
row *Air Traffic Dataset* — gross traffic is a global scalar on every
track; data parity impossible by construction). The old register claim
"mock + global CAGR, no route resolution" is partly stale: ``traffic_model``
is a real, URL-shareable input with three tracks (``legacy_anchor`` |
``cascade_series`` | ``route_resolved`` — DL-022/DL-045), and the opt-in
``route_resolved`` track routes CORSIA/CGI eligibility through the
routes_CORSIA state-pair filter over the synthetic-representative route
set and reports per-class flight counts (eq. 10) plus the traffic
conservation identities. Since DL-165 every ``LCAResult`` carries
``provenance.trafficBasis`` — the DL-139 §6 five facts (root year, root
value, growth rate, COVID treatment, recovery assumption) built from the
run's **actual** inputs, plus a verbatim statement of which dataset the
track runs on — so "no route resolution" (or its scope) is said by the
engine, not remembered by the UI. Gross remains the same global
:math:`y(t)` on all tracks (numerically bit-identical, asserted by test).
**Source:** ``src/engine/routes.ts:SYNTHETIC_ROUTES``,
``src/engine/routes.ts:SYNTHETIC_AIRCRAFT_TYPES``,
``src/engine/traffic-dataset-provenance.ts`` (DL-165),
``src/engine/traffic-basis-registry.ts`` (DL-139 named bases)
**Test evidence:** ``tests/future-aircraft.spec.ts`` (dataset consumers),
``tests/cascade-parity.spec.ts`` (frozen aggregate baseline),
``tests/traffic-dataset-provenance.spec.ts`` (provenance + bit-identity),
``tests/suite-p-adjudication.spec.ts`` (named bases)

The synthetic 24-route dataset
------------------------------

CASCADE resolves traffic per route from a licensed Cirium schedule extract.
HyFlux cannot and does not reproduce that dataset; instead it ships a
transparent synthetic stand-in, ``SYNTHETIC_ROUTES``, used by the
route-level Future Aircraft module (:doc:`aircraft`). Every value is a
plausible, first-principles choice — distances are plausible great-circle
values for the named city pairs (rounded), flight counts are plausible
schedule densities — and none is a schedule-database extract.

The dataset spans four size classes and three distance bands:

.. list-table::
   :header-rows: 1
   :widths: 22 20 18 40

   * - Segment
     - Routes
     - Distance band
     - Eligible classes
   * - Regional short-haul
     - 8
     - 400–1,000 km
     - ``regional``
   * - Single-aisle medium-haul
     - 8
     - 880–4,800 km
     - ``singleAisle``
   * - Widebody long-haul
     - 4
     - 5,550–11,020 km
     - ``widebody``
   * - Dedicated freighter
     - 3
     - 4,920–6,970 km
     - ``freighter``
   * - Combi (pax + freight)
     - 1 (AMS–JFK)
     - 5,850 km
     - ``widebody``, ``freighter``

Each route record carries: stable id, origin/destination codes, great-circle
distance :math:`d^*` (km), the set of eligible size classes, annual flights
:math:`n_{in}(rt, 2019)`, and the typical existing-aircraft class. Ten
synthetic existing Jet-A aircraft types (40–300 seats; 50.4 t and 68.1 t
freighters) complete the dataset. Seat and payload capacities follow the
publicly disclosed CASCADE size-class capacities (80 / 180 / 300 seats;
50.4 / 68.1 t freighters — the only public elements of Boeing's reference
aircraft, per ``constants.ts:SIZE_CLASSES``).

Traffic at this layer is **flat**: :math:`n_{in}(rt,t) = n_{in}(rt,2019)` —
no traffic growth is applied at route level; growth lives in the aggregate
engine (:doc:`trafficGrowth`).

Benchmark panel and governed Cirium snapshots
---------------------------------------------

A corridor panel provides gCO₂e/ASK benchmarks for major corridors. Two
sources exist, both display/benchmark-only:

* the labelled mock set (``src/data/cirium-mock.ts``, sampled from IATA
  WATS 2024 / ICAO figures :cite:`iata_wats`), and
* an optional **governed live Cirium corridor snapshot** (DL-153,
  ``src/data/cirium-live.ts``): real Cirium airport geometry with
  engine-derived (Method B) CO₂/ASK. The snapshot file is gitignored
  (licence gate — ``retentionConfirmed: false`` forbids redistribution,
  DL-133 §11); builds without it fall back to the mock set. The Sky
  Emissions product is not in the trial licence, so no value may be
  described as Cirium-observed emissions, and the scope note travels with
  every traffic-basis declaration (``CIRIUM_LIVE_SCOPE_NOTE``): a
  10-corridor benchmark is never a traffic basis and no engine figure
  derives from it. Test evidence: ``tests/cirium-live.spec.ts``.

Separately, DL-133 §11 defines a **versioned offline route-traffic
snapshot schema** with coverage validation and provenance
(``src/data/cirium-snapshot.ts``) and a route-level baseline builder that
puts HyFlux and CASCADE on the same functional units — corrected flown
distance :math:`d^* + 51` nmi, no imputation, schedule-vs-actual basis
kept explicit (``src/engine/route-traffic-baseline.ts``,
``tests/suite-g-cirium-baseline.spec.ts``). It ships **no** Cirium data:
every export path refuses row-level data until the licence position is
explicitly confirmed.

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

* **Fidelity vs Cirium.** CASCADE's route dataset is proprietary; the
  synthetic set is order-of-magnitude plausible only. **Data parity must
  never be claimed** (Phase-0 report §7/§12).
* **Aggregation.** The aggregate engine (``cascade.ts``) uses a single
  global baseline (0.92 GtCO₂ in 2019) with no route, operator or aircraft
  dimension in the production wedges. The aircraft-type × route flight
  tensor and its folds exist standalone (``fleet-tensor.ts``, DL-137,
  ``tests/suite-n-corsia-tensor.spec.ts``) with the existing-aircraft axis
  honestly lumped; the operator axis exists only in the synthetic CORSIA
  operator block. Real per-type/operator resolution is licence-blocked
  (Cirium).
* **Flat route traffic in the wedge path.** Route flight counts do not
  feed growth; growth is modelled in the aggregate engine. On the opt-in
  ``route_resolved`` track, eq. 10 per-class flight counts
  :math:`n(rt,t) = n_{2019}(rt)\,y(t)` and the traffic conservation
  identities are computed and **reported** (DL-165) — reporting only, not
  a route-level gross substitution.
* **Combi handling.** The one dual-eligible route (AMS–JFK) is the only
  place passenger and freighter classes interact; fleet composition beyond
  class eligibility is not modelled.