wayve adapter reads a capping-inversion structure the windIO schema does not define

[bjarketol]

Summary

wifa/wayve_api.py::flow_io_abl reads the capping inversion from a nested block in the wind resource:

wind_resource:
  thermal_stratification:
    capping_inversion:
      ABL_height: {dims: [time], data: [...]}
      dH:         {dims: [time], data: [...]}
      dtheta:     {dims: [time], data: [...]}
      lapse_rate: {dims: [time], data: [...]}

(see wayve_api.py lines ~761-764 for the scalar/Nieuwstadt branch and ~815-820 for the vertical-profile branch on current main).

The windIO schema (windIO/schemas/plant/energy_resource.yaml, EUFLOW/windIO) instead declares flat keys at the wind-resource level:

• ABL_height
• capping_inversion_thickness (m)
• capping_inversion_strength (K)
• capping_inversion_height (m)
• lapse_rate (K/m)

thermal_stratification / capping_inversion as a nested object appears nowhere in the schemas, and the flat capping_inversion_* keys are read nowhere in the wayve adapter. So:

• a windIO file that is schema-conformant (flat keys) silently gets WIFA's hard-coded defaults (H=1500 m, dH=100 m, dtheta=5 K, lapse_rate=2e-3 K/m) or the internal ci_fitting fallback;
• the input the adapter actually consumes (nested block) does not validate against the schema. This currently goes unnoticed because the wayve path loads the YAML without schema validation.

There is also a naming/semantics mismatch to resolve while at it: the schema's capping_inversion_height vs the adapter's ABL_height (both meaning the inversion-center height h), _thickness vs dH, _strength vs dtheta.

Suggested resolution

Either

1. teach the adapter to read the schema's flat capping_inversion_* / lapse_rate keys (possibly keeping the nested spelling as a deprecated alias), or
2. change the windIO schema (EUFLOW/windIO) to define the nested thermal_stratification.capping_inversion object and drop/deprecate the flat keys,

and in either case enable schema validation on the wayve input path so the mismatch cannot silently recur.

Context

Found while grounding the FLOW model-chain Twin Groves wayve example in per-state ERA5-derived capping inversions (flow_model_chain, gitlab.windenergy.dtu.dk:eu-flow/wp5/flow_model_chain). We currently write the nested spelling because that is what the adapter reads, and document the non-conformance in the example. Happy to contribute a PR for either direction once maintainers pick one.

🤖 Generated with Claude Code

[bjarketol]

One more wind-resource key in the same category: the wayve adapter now reads an optional per-state air_density from the wind resource (bjarketol/WIFA@1fff183, defaults to 1.225 kg/m³ when absent — it linearly scales the reported turbine power; the APM velocity solution is density-independent). air_density is also not declared in the windIO wind-resource schema, so whichever direction is chosen for the capping-inversion keys, it would be good to add air_density to the schema at the same time.

🤖 Generated with Claude Code