Methodology and data limitations
Updated: October 6, 2026. WeatherRecall publishes educational synthesis of archived weather. The explanation below reflects the current retrieval and aggregation code, not a claim of certified forecasts, validated prediction skill, or personalized safety advice.
Calendar-date retrieval
The main search requests the selected month and day in previous years from Open-Meteo's historical archive, using the chosen coordinates and timezone=auto. Daily and hourly variables support temperature, rainfall, wind, humidity, cloud, pressure, and time-of-day summaries. The request does not fix a models parameter; provider availability and defaults matter. Reanalysis is a model-based reconstruction informed by observations, not a direct measurement at every coordinate.
Nearby coordinates and weighting
The main lookup finds catalog stations within 25 km, requests Open-Meteo history at their coordinates, and blends successful responses with the target response. It does not thereby retrieve the stations' original sensor observations. An “Open-Meteo + Stations” interface label should be understood as this coordinate-based augmentation, not proof of an agency observation feed.
The target receives a virtual distance of 0.1 km, producing weight 10. Nearby points receive inverse-distance weights, with a 0.1 km substitute for zero distance. Numeric fields use weighted averages of available numeric values. The target category and sunrise/sunset are retained in the daily blended result. This implementation strongly favours the target; it is not an externally validated correction model.
Historical frequency versus prediction
A condition frequency counts qualifying entries divided by the retrieved sample. Hypothetical example: eight qualifying entries out of ten gives 80%. It does not establish an 80% forecast probability for a future date. Small samples, interannual variability, changing climate, and missing entries can materially alter interpretation. State the actual denominator and period rather than reporting an unexplained percentage.
Scores and aggregates
Weather categories and 0–10 scores are application-defined summaries, not official comfort or safety standards. The planner's qualifying-year share uses its score threshold; temperature, wind, and venue suitability must be considered separately. Different tool pages use their own aggregation logic and should not be assumed to share an identical sample or all main-search fields.
The monthly climate tool summarises archived daily data over ten completed years and counts precipitation days using its 0.5 mm threshold. City reports provide a multi-year monthly view. Annual reports compare a selected year with a preceding-ten-year baseline for supported fields. These shorter or custom baselines are not automatically formal 30-year climate normals. Global records rank saved lookup summaries, not a complete global census.
Checks and limits
The main fetch checks response status, a daily time array, numeric minimum and maximum temperature, and a usable weather code. Failed main requests return no result. Nearby failed responses are excluded. Those checks are not evidence of a universal physical-outlier filter, systematic station validation, or a confidence interval. Inspect missing fields and returned years before drawing conclusions; do not invent replacement observations.
Grid resolution, coastal boundaries, elevation differences, urban exposure, and convective showers can limit local representation. Daily totals hide timing; averages hide extremes. A sunny profile is not measured sunshine duration. For reproducibility, keep coordinates, dates, units, source notes, retrieval date, and the sample size.
Source references and corrections
Consult Open-Meteo historical API documentation, ECMWF ERA5 documentation, and NOAA Climate Data Online for independent station-based checks. NOAA is a verification reference, not a claimed feed for every site query.
See source provenance and editorial standards. Send reproducible discrepancies to [email protected]. For consequential decisions, use current official warnings and appropriate local expertise.