Kortex · cooling-water concentration · live from the graph
BETAevidence: inferred. US subset validated, non-US unvalidated
Diversified assets. One river.
A thermal fleet can span a map and still share one water dependency: several plants drawing cooling from the same river system fail together in the same drought. Kortex traces each plant to the dams that cool it, chains those dams along their rivers and sub-basins, and shows how much of an operator's capacity sits on a single water system; with the WRI Aqueduct stress score of that system attached.
Who is one drought away
each bar splits the operator's water-linked thermal capacity across its independent water systems; the leftmost block is the largest single system, coloured by its Aqueduct stress (0–5). Coverage of the group's thermal fleet is stated per row; under-evidenced rows (<60% coverage or <5 linked plants) are never ranked, only shown.
Method & receipt. Plant→water links are Dam-[:SUPPLIES_COOLING]→Generator edges (thermal fuels only: gas, coal, oil, nuclear; canonical operational units). A water system = dams sharing a level-6 HydroSHEDS Pfafstetter sub-basin, merged along explicit same-river chains; river transitivity is meaningful (upstream and downstream share the drought), unlike weather-correlation chaining, which the weather page rejects. A 4-digit basin rollup was tested and rejected: it merges half a continent. Stress = the maximum WRI Aqueduct v4 baseline water-stress score among the largest system's cooling dams; basin-level, not site-measured, and it does not model water rights, discharge consents or cooling technology. Coverage fails closed: figures describe plants with cooling links in the Kortex graph; groups below 60% linked capacity or 5 linked plants are flagged under-evidenced and excluded from ranking. The cooling links are proximity-derived (inferred) and validated where a reference exists: every judgeable US edge was tested against the utility's own EIA-860 cooling filing (study 8); 35% contradicted the filing (dry-cooled, ocean-cooled or groundwater-supplied plants) and are quarantined out of these figures; the river-name-mismatch class was resolved by hydrological geometry (2026-07-11): the HydroRIVERS reach test confirms 94% sit on the same mainstem as their dam (naming noise; links real); the remaining 45 links stay flagged as doubtful or unconfirmed. Non-US edges have no equivalent reference yet and remain inferred-unvalidated; read non-US rows as screening-grade. GET /v2/spatial/water-portfolio · sources.