WeatherVerity
English

Source · Compare · Explain

Weather you can verify.

Forecast receipts and changes

WeatherVerity preserves selected forecast editions so a later page can show what was predicted and how it changed. These are model forecasts from the named provider, published by WeatherVerity.

Pilot storage is limited to Oslo, London and Da Nang. At most one distinct successful edition per provider, product and dataset is sampled in each six-hour UTC slot; corrections with the same retrieval time remain traceable. Retention is 35 days with a hard cap of 512 receipts per place.

Retrieval time is when WeatherVerity fetched the response. Valid time is when a value applies. Retrieval lead is valid time minus retrieval time. Model forecast horizon is valid time minus a documented provider model-run time; it stays missing when that time is not supplied. Earlier v2/method-v1 receipts used the field name forecast_horizon_seconds for retrieval lead and are retained only as legacy evidence, never silently reinterpreted.

Forecast stability compares only the same provider, product, place, units, interval and exact valid time. Forecast agreement compares providers; it does not prove accuracy. Verified accuracy requires a documented observation and remains “Collecting data”.

weatherverity-forecast-receipt-v3 · forecast-receipt-method-v2 · forecast-evolution-method-v1

Who does what

  1. WeatherVerity
    WeatherVerity publishes the page and calculates summaries, comparisons and Outdoor Scores from the displayed forecast data.
  2. Open-Meteo
    Open-Meteo supplies the forecast data and selects weather models. A model-run issue time is not supplied.
  3. GeoNames
    GeoNames supplies the place identifier and location reference.

WeatherVerity

WeatherVerity publishes forecasts, comparisons and outdoor insights with transparent sources.

Forecast providers

Underlying forecast data: Open-Meteo. Each forecast section identifies the source actually used. WeatherVerity does not produce the underlying meteorological model data.

Location data

GeoNames (CC BY 4.0). Values are formatted for readability; missing weather data is never invented.

WeatherVerity calculations

Calculated by WeatherVerity from the forecast data shown on this page.

Best and worst outdoor periods, activity scores and weather windows use one deterministic method across HTML and JSON. Today covers the whole local day. Relative results describe the best available period, not a guarantee of safe or good conditions. outdoor-score-v2. Explore outdoor scores and their method.

Homepage weather comparison

How the scores work

Cold wind lowers the comfort cap further. Temperature below 15 °C reduces General comfort. A two-hour window including such temperatures is capped at 79. Temperature, precipitation probability/weather code and wind are combined geometrically. A period overview combines 80% time-weighted mean with 20% of the weakest quarter; daylight counts double. Risk intervals remain visible. Raw scores determine rankings; 100 is unavailable with incomplete model inputs. This is a comfort indicator, not forecast accuracy or an official warning.

Limited assessment: gusts, rainfall amount, humidity and solar data are unavailable.

Country forecasts

Country pages show selected city forecasts through the same data pipeline. They do not calculate national averages or claim to represent every part of a region.

Report a forecast or page issue.

What “current” means

Current conditions are estimates from weather models, not verified readings from a station at your exact address. We show the time the estimate is valid for separately from the time we retrieved it. The precipitation field covers the current interval shown beside it.

Reading the seven-day forecast

Each row covers a local calendar day. Temperatures are in Celsius, precipitation amounts in millimetres and wind speeds in metres per second. “Precip. chance” shows the maximum precipitation probability for that day; precipitation may include snow and other forms, not just rain. It is not a percentage of the day that will be wet. Wind is the daily maximum at 10 metres above ground.

Times, sources and freshness

Times on place pages use that place’s named time zone. Current data is cached for up to five minutes, hourly forecasts for up to fifteen minutes, daily forecasts for up to thirty minutes, and rendered pages for up to one additional minute. A page kept open does not update itself; reload it for a newer response. Source retrieval timestamps remain unchanged while a response is cached. If an update fails, the shared cache can show an explicitly labelled older forecast for up to ten additional minutes for current conditions, thirty for hourly data and sixty for daily data. Current estimates stop being served at one hour past their valid time; hourly and daily fallbacks stop at a local date change. Fallback reads never extend these limits.

The selected API does not provide a model-run issue time with these responses. We label that as unavailable. The field generationtime_ms is the API’s processing duration, not the time a forecast was issued.

Backup weather

When configured, MET Norway is a secondary source if Open-Meteo cannot supply validated data. The affected sections identify the provider. MET wind is a 10-minute mean, and its time points do not necessarily supply equivalent daily extrema, probabilities or condition codes. A backup daily table therefore labels sampled ranges explicitly and leaves unsupported fields missing. This is failover, not a verified multi-model consensus.

Backup attribution: MET Norway, CC BY 4.0. Changes include normalization, local-date grouping and sampled summaries. Retrieval, source product update and model run are distinct times; no model run is invented. MET data definitions.

What this release can compare

Open-Meteo is the primary provider and selects suitable weather models. A configured backup replaces unavailable sections; it does not combine providers into a consensus forecast. See the Open-Meteo documentation for model selection and variable definitions.

On Oslo, London and Da Nang pages, choose “Compare temperature forecasts” to compare Open-Meteo and MET Norway for the same requested location and exact valid time, from the current hour to the end of the local day. The comparison reuses the primary hourly forecast and loads MET through a shared cache when requested. Values from an out-of-date source are excluded. Missing hours stay missing; we do not interpolate them. Browser time selection uses data already in the page and does not fetch another forecast.

Only temperature in Celsius is compared. Forecast differences do not measure observed accuracy. These providers may share model inputs and are not confirmed to be independent. Each source has its own retrieval time; model run times are not supplied. There is no blended forecast, winner or Confidence score. Selecting comparison does not activate automatic backup weather.

Your location on the homepage

The location card appears as soon as the page opens. On the live site we can use an approximate city from your internet connection. This can be inaccurate with VPNs or mobile networks; change the place whenever needed. Choosing “Set as my location” remembers only that public place ID in this browser until you clear site data. We do not request GPS access. Local previews require a place selection. Personal location responses are not stored in shared page caches.

Favorite places

Save up to five places using “Add to favorites” on a place page. Your favorites appear on the homepage and place pages in this browser. Saving or removing a favorite does not change “Your location”. Favorites are shortcuts: weather is fetched when you open a place, not for every saved place in the background. Remove a favorite with its remove button, or clear site data to remove all saved preferences.

Attribution and use

Weather data: Open-Meteo. Place data: GeoNames. Attribution licence: CC BY 4.0. Presentation changes include unit selection, rounding and condition labels.

When data is unavailable

Missing individual values are marked “Not provided”. If a service is unavailable or its response cannot be validated, that section displays an error. WeatherVerity does not provide an official severe-weather warning service; consult your local meteorological authority for alerts.

Preferences on this device

Language, units and favorites stay on this device. Recent places are saved only if you enable them. No account needed.

Clear history
How this works

Cold wind lowers the comfort cap further. Temperature below 15 °C reduces General comfort. A two-hour window including such temperatures is capped at 79. Temperature, precipitation probability/weather code and wind are combined geometrically. A period overview combines 80% time-weighted mean with 20% of the weakest quarter; daylight counts double. Risk intervals remain visible. Raw scores determine rankings; 100 is unavailable with incomplete model inputs. This is a comfort indicator, not forecast accuracy or an official warning.

Forecast intervals must cover the entire comparison period. When source hours cross its boundaries, only the overlapping duration counts, with double weight for daylight. The original hourly scores and endpoint estimates are retained; no new weather values are interpolated. Boundary estimates may therefore include an endpoint just outside the period. The highest and lowest scored intervals can be shorter than one hour at a boundary.

Temperature bounds apply to the entire period, not its average. Mostly clear requires at least 75% of complete daylight intervals to be clear.

The shared hourly data covers three local calendar days. Weekend and seven-day comparisons appear only with complete coverage. UV, humidity, gusts, snowfall amounts, maps and travel times are not included.

Calculated by WeatherVerity. Underlying forecast data: Open-Meteo · CC BY 4.0 · Sources & method