WeatherPerch Intelligence
WeatherPerch Data & Methodology
WeatherPerch combines weather forecasts, environmental observations, historical climate information, and other relevant datasets from multiple sources.
These inputs are normalized and processed through WeatherPerch’s intelligence layer to produce activity scores, timing recommendations, historical comparisons, condition assessments, and plain-language explanations.
WeatherPerch’s intelligence layer is designed to answer a different question from a traditional weather forecast: not simply “What will the weather be?” but “What does the weather mean for what I want to do?” It applies transparent, documented scoring rules to third-party forecast data — it does not generate its own weather forecasts.
Data sources, availability, and methodology may vary by location and feature. Third-party data is used and attributed in accordance with the applicable licences and terms of each provider. WeatherPerch does not represent third-party data providers as endorsing or sponsoring WeatherPerch.
Data sources
WeatherPerch combines forecast, observation, air-quality, marine, climate, and official-alert sources according to location and feature coverage. Source-specific licence notices and attribution are listed below.
Activity decisions
Running, hiking, beach, wedding, and gardening pages evaluate rain probability, wind, temperature, daylight, air quality, frost, and relevant alerts. Scores summarize forecast inputs; they are not guarantees or medical, safety, or professional advice. Official warnings always take priority.
Intelligence reports
Anomaly and ranking reports compare a location with an explicitly stated historical baseline. Each report should state its coverage, timestamp, sample size, source, formula, limitations, and whether it uses forecast or observed data. Forecast confidence will only be published after enough forecast-versus-observation history exists.
Updates and limitations
Forecasts change as providers update their models. Coverage, resolution, historical depth, and alert availability vary by location. A missing signal is not treated as proof of safe or normal conditions.
How scores are calculated
Each activity has documented preferred ranges and response curves for temperature, wind, precipitation, humidity, UV, daylight, and other relevant variables, plus modeled interactions between them (for example, humidity making heat feel worse). These weights and curves are configured by WeatherPerch, not learned from unreviewed user data, and are versioned so a given recommendation can always be traced back to the exact model configuration that produced it.
Severe-weather hazards (official alerts, extreme heat or cold, high wind, flooding, poor air quality, cyclones, and snow or ice where relevant) are checked separately from the comfort-based score and can override an otherwise favorable rating. A hazard override is never softened into positive language.
Confidence and uncertainty
Confidence reflects how complete and current the underlying data is, how far in advance a forecast was made, and whether relevant signals — such as forecast disagreement between models or missing inputs — reduce how much weight a recommendation should carry. Confidence is not lowered automatically just because a forecast extends further into the future; it is lowered only when the available evidence genuinely supports less certainty. WeatherPerch does not present a recommendation with unwarranted precision, and withholds a confident recommendation when the underlying data is insufficient.
Historical context
Where a location has enough years of historical observations for a given date or season, WeatherPerch compares current conditions against that location's own historical distribution — not a fixed global threshold or an arbitrary "yesterday" comparison — to describe how typical or unusual conditions are. This comparison is only shown once the sample size is sufficient to support the claim; otherwise WeatherPerch says so rather than presenting a comparison it cannot back up.
Verification status
WeatherPerch's scoring, explanation, and historical-comparison logic is covered by automated tests that check score bounds, hazard precedence, confidence behavior, and wording consistency. As of this writing, that verification has been run locally and in continuous integration; independently measured, real-world forecast-versus-outcome accuracy for WeatherPerch's own recommendations has not yet accumulated enough history to publish, and no such accuracy claim is made until it can be backed by real data.
Attribution and licences
WeatherPerch's own work is its scoring, decision, recommendation, and explanation logic — not the underlying weather data itself, which remains the property of the providers below under their own terms. This list reflects the data sources this build of WeatherPerch is configured to call; a source appearing here does not imply WeatherPerch has completed every provider's commercial or legal review, and some sources listed are not yet enabled in production.
- Weather data by Open-Meteo.com — CC BY 4.0 data licence
- Weather data by OpenWeather — Plan-specific OpenWeather licence
- Weather data by WeatherAPI.com — WeatherAPI.com subscription terms
- Weather Data Provided by Visual Crossing — Visual Crossing commercial subscription licence
- Source: NOAA / National Weather Service — U.S. public-domain government information unless otherwise marked
- Source: U.S. Geological Survey — U.S. public domain for USGS-produced data unless otherwise marked
- Source: Federal Aviation Administration — U.S. government public information; endpoint-specific conditions must be checked
- Source: NASA POWER — NASA Earthdata open-data guidance; dataset-specific restrictions control
- © OpenStreetMap contributors — Open Database Licence for map data; tile usage policy also applies