ABLAQ-R05: When weather changes a courtyard comparison
Version 1.0.0 | Published 13 October 2026 | ABLAQ Studios
https://ablaqstudios.com/weather-weighted-solar-access

This release tests an idealized courtyard model using historical external solar-radiation
forcing. The data cover 2005-2023; ABLAQ's analysis and publication occurred in 2026.
The model coordinate (32.3 N, 36.2 E) is not a claimed field site or institutional address.

REPRODUCE
Download and extract weather-research.v1.zip into an empty directory.
The archive contains source at its root, the original forcing under inputs/,
and the reference numerical outputs under published/.
Tested environment: CPython 3.14.7, numpy 2.5.4, pandas 3.0.6, pvlib 0.16.1.
Install the pinned dependencies in an isolated Python environment:

python3 -m pip install -r requirements-weather.v1.txt
python3 -B weather-study.v1.py --inputs inputs --output reproduced
python3 -B verify-weather.v1.py --inputs inputs --results reproduced --output independent-audit.json
python3 -B compare-weather.v1.py --expected published --actual reproduced

The output folder must not already exist. The scientific run makes no network requests.
On the reference environment a fresh process reproduced every computational output byte
for byte. The comparison command reports that exact criterion; other platforms may need
inspection of last-digit floating-point differences rather than an assumption of equivalence.
The separate verifier uses polygon clipping and scalar accumulation without importing the
primary study code. It shares pvlib's SPA ephemeris. This is internal numerical verification,
not independent replication by another researcher, external peer review or field validation.

FILES
weather-seasonal.v1.csv: all 3,024 source/season/year/geometry rows, with component totals.
weather-aggregate.v1.csv: 162 exposure-weighted source/season/geometry summaries.
weather-pairwise.v1.csv: 36 paired elongated-minus-square comparisons by source, season and height.
weather-sensitivity.v1.csv: 162 2023 numerical/time-model comparisons.
weather-summary.v1.json: structured results and paired-year summaries.
weather-verification.v1.json: input census, source anomalies, numerical checks and sensitivities.
weather-independent-audit.v1.json: separately expressed internal audit.
weather-computation-manifest.v1.json: SHA-256 digests of regenerated results and primary code.
weather-input-provenance.v1.json: source requests, original response digests and acquisition dates.
weather-data-dictionary.v1.json: variable meanings and units.
weather-protocol.v1.json: analysis design and the subsequent documented data-quality extension.
LICENSE-weather.v1.txt: original-material and third-party terms.

IMPORTANT INTERPRETATION
Winter is December-February, labelled by its January year, with complete winters 2006-2023.
Annual and June-August summer values cover 2005-2023.
SARAH3 is sampled at minute 09; ERA5 hourly means are centred at minute 30.
ERA5 uses a cosine-weighted within-hour shading average under constant within-hour DNI.
Irradiance is horizontal already; do not apply an additional cosine to the source beam term.
All aggregate fractions are ratios of exposure sums, not unweighted means of percentages.
The model excludes reflections and has no thermal, comfort or occupied-building calculation.
The original radiation products are kept separately. Their difference is not a confidence interval.
The angular sky profiles are prescribed sensitivity scenarios, not inferred weather states.
424 unflagged SARAH3 zero-radiation records occur above 10 degrees geometric solar elevation.
The primary calculation retains those values. A post-inspection diagnostic removes the same
220 affected UTC dates from both products; no source value is imputed or silently corrected.

Scientific accountability: Osama Khaled Ahmed Abdel-Qader, Scientific Research Lead.
Institutional author and publisher: ABLAQ Studios.
Questions or corrections: osama.khaled@ablaqstudios.com.