Add HRRR historical data investigation report

[aldenks]

Summary

This PR adds a comprehensive investigation report documenting the feasibility of extending the HRRR analysis dataset backward from the current start date of 2018-09-16 to 2014-10-01, providing 4+ years of additional historical weather data.

Key Findings

• Extension is feasible: The HRRR analysis dataset can be reliably extended to 2014-10-01
• Variable availability: 16 out of 20 current variables are available in the historical period
• Data quality: File structure, grid coordinates, and temporal availability are consistent across all versions examined
• Grid stability: Minor sub-pixel shift (< 0.02 pixels) between HRRRv1/v2 and HRRRv3 is negligible

Notable Observations

1. GRIB name change: Mean sea level pressure uses PRMSL in HRRRv1/v2 but MSLMA in HRRRv3 - requires version-aware lookup
2. Missing variables:
   • downward_long_wave_radiation_flux_surface (added in HRRRv3)
   • relative_humidity_2m (added in HRRRv3)
3. URL pattern: Unchanged since 2014-09-30, compatible with current implementation
4. Data gaps: Transition period (2016-08-23 to 2016-09-01) has missing hours; first operational day (2014-09-30) is incomplete

Implementation Recommendations

The report outlines a phased approach:

1. Add version-aware variable handling with available_from timestamps
2. Implement GRIB element name mapping for PRMSL/MSLMA
3. Update append_dim_start to 2014-10-01
4. Test with small backfill before production rollout

Investigation Details

• Files examined: 7 GRIB2 files spanning 2014-2018
• Dates tested: 15+ random dates across the historical period
• Investigation date: 2026-02-02

This report provides the technical foundation for implementing the historical extension in a follow-up PR.

https://claude.ai/code/session_014f4ZR6Cc5XvmSxQCQdCerk

[aldenks]

fabulous, thanks claude.

[aldenks]

Tracking implementation in #415