The known provincial WRF archive tile 0307 contains the requested temperature, pressure and water-vapor fields. We recovered 24 hourly states using 27 guarded partial requests: 1,769,472 new bytes (1.6875 MiB), within the frozen 16 MiB limit. Existing geopotential, terrain, coordinates and vertical metadata were reused. Every retrieved block retains its request headers, byte range, ETag and SHA256; the 1.80 GB annual file was not downloaded.
This dataset covers 100 native columns, parent I31–40/J71–80, and 33 mass levels, at 2023-06-01 09:00 through 2023-06-02 08:00 UTC. The tile was already selected for a diagnostic neighbourhood before its thermal values were inspected. It does not establish coverage of the whole community. The archive retains 34 vertical faces but omits the upper seven mass levels present in the 40-level M3D file. The earlier 08:00 UTC initialization state is outside this acquisition.
The archive field T is perturbation potential temperature, not air temperature in kelvin. The retained CALWRF source gives pressure in hPa as (P + PB) × 0.01, then air temperature as (T + 300) × (pressure_hPa / 1000)^(287.04 / 1004). We preserve its single-precision operation order and store a separately labelled binary64 arithmetic comparison. The WRF registry distinguishes this T from the separate moist/dry THM field; no extra moisture correction was applied to T.
All 79,200 values per field reproduce the retained printed tokens for pressure, temperature, water-vapor mixing ratio and reused mass-level height. All fields are finite; pressure, temperature and humidity remain within the released CALWRF QA bounds, so this check requires no repair or interpolation.
| Field | Reconstructed range in this tile and interval | Retained M3D printing step | Largest absolute residual |
|---|---|---|---|
| Air temperature | 226.02173–292.12674 K | 0.1 K | 0.05 K exactly |
| Pressure | 314.69836–1017.38409 hPa | 1 hPa | 0.5 hPa |
| Water-vapor mixing ratio | 0.045852–9.060075 g/kg | 0.01 g/kg | 0.004999924 g/kg |
| Mass-level geopotential height | 11.32216–8944.98047 m | 1 m | 0.5 m |
The first floating-point comparison flagged four temperature residuals as 0.05000000000001137 K. Exact decimal checks show all four are 0.05 K ties, for example 283.75 versus the printed 283.8. The initial flags and full locations remain in the audit; no tolerance was widened. Evaluating the same air-temperature formula in binary64 changes the result by at most 0.000041092 K here. That is an arithmetic difference, not an accuracy estimate.
The practical gain is recovery of information lost by M3D decimal printing. It enables a separately declared precision check on this preselected neighbourhood. It does not improve the physical grid resolution, validate the weather model, establish June production settings, or justify a thermal/stability change. The annual file's global attributes carry initial-run December/January metadata. Agreement with the retained M3D supports the field identity and conversion, without proving the provider's executable or compiler options. These are atmospheric profiles; they are not skin temperature or sea-surface temperature and have not been adopted as a water boundary. No new model run, model-input change or Site edit was made.
Download the compact profiles, data schema, checks, provenance and declared acquisition protocol. Local acquisition and request logs retain the individual range retrieval evidence. All paths in the machine-readable artifacts are relative to the repository.