Data provenance and acknowledgment

Meteorological observations

Temperature, relative humidity, pressure, wind speed, wind direction, optional total cloud cover, and optional measured GHI originate from SMHI MetObs corrected archives. Only accepted quality flags are used. The selected SMHI station identifier is written into the EPW comments; it is not substituted for a WMO number. Automatic recovery can use assessed nearby stations, with filled hours, validation results and corrections retained in the processing report.

Solar radiation

STRÅNG provides gridded solar radiation at the requested coordinates. The EPW header includes the required acknowledgment:

STRÅNG data used here are from the Swedish Meteorological and Hydrological Institute (SMHI), and were produced with support from the Swedish Radiation Protection Authority and the Swedish Environmental Agency.

Reanalysis fallback

Same-year ERA5 through Open-Meteo recovers required hours that primary sources, short interpolation, assessed donors and bounded daily profiles cannot supply. The report records the selected model, returned grid location/elevation, source fractions and warnings. EPW comments identify fallback use; retain the JSON sidecar for its complete details. A mixed or reanalysis-only year remains an AMY estimate for the requested dates, not proof of local measured conditions.

Acknowledge Open-Meteo and the Copernicus Climate Change Service ERA5 data when using fallback estimates. Optional ERA5 cloud cover is recorded only at hours where ERA5 replaces required meteorology or GHI; usable cloud values and other hours are preserved. reanalysis_filled_hours includes recovered cloud hours, while source_fractions describes required meteorology and GHI. The adapter uses the public noncommercial endpoint; consult Open-Meteo service access for current limits and commercial requirements. See Missing-weather recovery and audit trail for source alignment, safeguards and limitations.

Reproducibility

Archive the configuration, package and pvlib versions, generated EPW, diagnostics, and source-access date with a simulation project. Raw payload caching improves repeatability, but upstream corrected archives may legitimately be revised. Use refresh intentionally and record when data was re-fetched.

Automatic policy writes OUTPUT.epw.json by default. Set provenance_path (CLI --provenance) for another destination. Strict mode writes a receipt only when a path is supplied. Sidecar paths must differ from the EPW and have an existing parent directory.

The sidecar contains reconstruction settings, diagnostics, warnings, output SHA-256, package/source identity and hashes of consumed UTF-8 responses. Receipts belong to a single compilation: reused clients and repeated URLs do not import unrelated earlier requests. Cached and live responses follow the same receipt contract. A custom client without scoped response recording is explicitly marked incomplete.

New receipts add the top-level pvlib_version field while retaining schema version 1 and existing fields. Earlier receipts can lack this field; retain their original source hashes and study identity. Current solar calculations use NREL SPA nrel_numpy geometric zenith with delta_t=None, ASCE extraterrestrial irradiance with a 1367 W/m² solar constant, and the continuous Erbs-Driesse fallback. See Pipeline and scientific methods for the preceding-hour conventions and Bibliography and source documentation for the implementations.

The completeness and sensitivity JSON files bundled with the examples describe their recorded pre-pvlib source revisions. Their counts, hashes and load differences remain historical evidence. Regenerate EPWs in a new directory when adopting pvlib and preserve the new receipt with its consumer results.

EPW export and sidecar export are individually atomic. The sidecar is written after the validated EPW. Under automatic policy, a sidecar filesystem error returns the valid EPW with a structured provenance_write_failed warning. Strict policy retains a reported sidecar error. A valid EPW alone does not establish that its audit trail was successfully saved.

Target elevation and derived pressure

SMHI MetObs parameter 9 is sea-level-reduced pressure (QFF), not station pressure. The compiler estimates surface pressure at target_elevation_m using the inverse SMHI reduction, with hourly observed temperature and target latitude. The result records pressure_method=smhi_qff_inverse_v1 and resolved target coordinates/elevation separately from result.station. This conversion applies only to usable SMHI QFF samples. ERA5 supplies surface pressure, which is converted from hPa to Pa and never passed through the QFF reduction again.

This estimate uses target elevation as barometer height and nearby observed temperature as representative of the target. It is not a direct pressure observation at the building; temperature inversions and station separation can introduce uncertainty. Original API payloads remain unchanged in the cache.

Reference: https://www.smhi.se/kunskapsbanken/meteorologi/lufttryck/hur-mats-lufttryck

Generated pressure values intentionally differ from releases that treated QFF as station pressure. Regenerate EPWs and update simulation identities rather than modifying an existing dataset in place.