docs/china-logistics-corridors.mdx
The China logistics control towers are transparent source compositions over reviewed geographic corridors. They report evidence availability and freshness by family. They do not generate an opaque logistics-risk score, forecast disruption, or infer activity from source presence alone.
| Corridor ID | Name | Reviewed scope |
|---|---|---|
china-yangtze-river-delta | Yangtze River Delta | Shanghai, Jiangsu, and Zhejiang ports, aviation, and advanced-manufacturing nodes |
china-greater-bay-area | Greater Bay Area | Guangdong, Hong Kong, and Macao-facing ports, airports, hazards, and industrial gateways |
china-bohai-rim | Bohai Rim | Beijing-Tianjin, Hebei, Shandong, and Liaoning maritime-industrial gateways |
china-western-land-sea-corridor | Western Land-Sea Corridor | Inland manufacturing hubs, Xinjiang land crossings, and Guangxi maritime gateways |
Each corridor has a reviewed polygon, node registry, node owner, and source selector. A source record contributes only when its selector or reviewed geography maps it to that corridor. National inputs are clearly marked as national and can apply to all four corridors.
| Family ID | Source contract | Scope and observation |
|---|---|---|
port | IMF PortWatch | Reviewed China/Hong Kong port IDs; calls, DWT measures, trend deltas, and anomaly flag when published |
aviation | AviationStack | Reviewed PEK, PVG, CAN, SZX, CTU, KMG, URC, and HKG provider coverage |
hazard | Western Pacific cyclone sources and Hong Kong Observatory | Events inside a corridor boundary plus explicit HKO warning coverage for the Greater Bay Area |
power_energy | WorldMonitor energy spine | National China energy dependency coverage, latest source observation time, and the observed year-over-year oil-product demand change with its own period end when published |
strategic_industry | UN Comtrade | Annual strategic-product observations for reporter 156 |
trade | UN Comtrade and Shanghai Shipping Exchange CCFI | Annual reporter-156 trade observations, the current China Containerized Freight Index, and its period-over-period change — the exchange's own percentage, or the change between two period levels it published — with the prior period it was measured against |
IMF PortWatch and AviationStack are node-scoped. Hazard observations are regional. WorldMonitor energy spine, UN Comtrade, and Shanghai Shipping Exchange inputs are national. The output keeps those scopes visible rather than presenting national evidence as a measured node value.
Every source signal carries:
Each family condition is available, partial, stale, or unavailable.
A corridor is:
available only when all six conditions are available;stale when every condition is available or stale and at least one is
stale;unavailable when all six are unavailable; orpartial for every mixed or provenance-invalid state.An absent configured port or airport becomes an explicit unavailable signal. A healthy feed with zero hazards can establish feed coverage, but it does not create a hazard event. Energy-spine coverage flags do not establish an activity direction; only an explicitly published demand change does, and it is dated by its own period end rather than by the family's latest source timestamp. A current CCFI level without a prior comparison does not become a freight change.
The published JODI demand change is a same-month year-over-year percentage. Its
basket sums TOTDEMO for the available GASOLINE, GASDIES, JETKERO,
RESFUEL, and LPG secondary products. At least three products must be
available in both comparison months with the identical basket; a change beyond
±50% is refused as an implausible reporting or unit error. The source signal
keeps the product list and current/prior kbd totals in provenance, while the
directional value is labeled % change.
The composer evaluates source timestamps at request time:
| Input | Transport budget | Content handling |
|---|---|---|
| PortWatch | 36 hours | Individual observations stale after 72 hours |
| AviationStack | 90 minutes | Individual observations stale after 180 minutes |
| Western Pacific hazards / HKO | 9 hours | Observations stale after 9 hours |
| Energy spine | 48 hours | Year/month/day precision gets a corresponding periodic budget |
| UN Comtrade | 48 hours | Annual observations use a 550-day budget |
| CCFI | 7 hours | Index content stale after 28 days |
Transport freshness proves that collection ran. Content freshness proves that the observation itself is timely. Neither substitutes for the other.
GET /api/supply-chain/v1/get-china-corridor-control-towers reads the
canonical Redis source seeds in one batch and isolates missing dependencies.
It validates every composed provenance envelope before returning the wire
payload.
Healthy aggregates use a five-minute server cache. An all-unavailable result is not positively cached, so a recovery can appear immediately. The client circuit breaker deliberately retains no last-good aggregate that could hide a current fail-closed response. The panel refresh interval is 15 minutes.
The response exposes upstreamUnavailable: true only when every corridor is
unavailable. Partial corridor evidence remains inspectable rather than being
collapsed into total outage or false health.
This detailed control-tower response is a REST/UI surface. MCP has no dedicated
corridor-control-tower tool; Pro MCP receives only its bounded contribution to
get_china_decision_signals.
The China Activity Nowcast can use a corridor-family breadth change only when a comparable prior corridor snapshot exists. The control-tower response itself remains a source composition and does not publish a directional economic conclusion.