Data Methodology

How we source, process, and present weather data. Transparent and accountable.

Data Source

JP Weather Pro sources all weather data from the Open-Meteo Forecast API. Open-Meteo is a free, open-source weather API service that aggregates global meteorological model outputs. It does not generate original weather forecasts. Instead, it collects, harmonizes, and serves data from established meteorological agencies' numerical weather prediction (NWP) models.

The specific model inputs relevant to Japanese cities are:

Open-Meteo performs its own bias correction and ensemble blending. For full details of their processing methodology, see the Open-Meteo documentation at open-meteo.com.

API Endpoint and Parameters

We query the Open-Meteo Forecast API with the following parameters for each city:

https://api.open-meteo.com/v1/forecast
?latitude={LAT}&longitude={LON}
¤t=temperature_2m,relative_humidity_2m,apparent_temperature,
weather_code,precipitation,wind_speed_10m
&hourly=temperature_2m,precipitation_probability,weather_code
&daily=temperature_2m_max,temperature_2m_min,
precipitation_sum,uv_index_max
&timezone=Asia/Tokyo
&forecast_days=3

This returns:

All data is requested with Asia/Tokyo timezone to ensure local-time alignment. We request exactly 3 days of forecast data because model skill degrades significantly beyond this horizon for local temperature and precipitation prediction.

Update Frequency

Weather data updates every time you load or refresh a page. We do not cache weather data on our servers. Each page load triggers a fresh API call to Open-Meteo. This ensures you always see the most recent model output available.

The underlying models update on their own schedules:

When you load our page at, say, 10:30 AM JST, the "current" conditions you see are interpolated from the most recent available model run (likely the 09:00 JST MSM run). They're not live station observations. This distinction matters for accuracy assessment.

No Data Manipulation

We do not modify, smooth, correct, or "enhance" the data from Open-Meteo. What you see on our dashboard is what the API returns, translated into human-readable format. Specifically, we do not:

The only transformations we perform are:

Accuracy and Limitations

Numerical weather prediction is remarkably good but not perfect. Based on our verification against JMA AMeDAS station observations, we observe the following typical accuracy for the data we display:

Parameter Typical Accuracy (0-24h) Typical Accuracy (24-72h)
Temperature ±1.0-1.5°C ±1.5-2.5°C
Feels-like (apparent temp) ±1.5-2.0°C ±2.0-3.0°C
Humidity ±8-12% ±12-18%
Wind speed ±2-3 km/h ±3-5 km/h
Precipitation (yes/no) ~85% correct ~75% correct
Precipitation amount ±2-3mm (light rain) ±5mm+ (variable)

These accuracy figures represent our observations from comparing API outputs against JMA AMeDAS observations over a 6-month verification period. They are not guaranteed performance metrics — actual accuracy varies by season, weather pattern, and city.

The most significant limitation is spatial resolution. The JMA MSM model uses approximately 5km horizontal grid spacing. This means a single grid cell covers all of central Tokyo plus surrounding areas. Weather conditions within that 5km cell can vary significantly — Shinjuku and Odaiba (both within the same grid cell) can differ by 3-4°C in summer due to urban heat island and bay breeze effects. Our city pages discuss these microclimatic variations qualitatively, but the model cannot resolve them.

Coordinate Locations

We use city-center coordinates for API queries. These represent the approximate geographic center of each city's primary urban area, not airport locations or specific station positions:

City Latitude Longitude Location Reference
Tokyo35.6762139.6503Imperial Palace district
Osaka34.6937135.5023Central Osaka (Umeda-Namba midpoint)
Yokohama35.4437139.6380Kannai/Minato Mirai area
Nagoya35.1815136.9066Sakae/Nagoya Station area
Sapporo43.0618141.3545Odori Park district
Fukuoka33.5902130.4017Tenjin/Hakata midpoint
Kobe34.6901135.1956Sannomiya district

If you need weather data for a specific location within these cities, the readings may differ from our displayed values. As a rough guide: add 1-2°C for dense urban commercial districts, subtract 1-2°C for waterfront or park areas, and subtract 2-3°C for hillside neighborhoods.

Verification Process

Our data verification lead (Mei Tanaka) checks API outputs against JMA AMeDAS observations daily. The process:

  1. Compare the previous day's forecasted temperatures against actual AMeDAS recordings.
  2. Log discrepancies greater than 2°C as notable.
  3. Identify patterns: consistent overestimation/underestimation by city or season.
  4. Publish findings on our accuracy log page.

This verification is for transparency, not correction. We don't adjust the API data based on discrepancies — we note them so users can calibrate their expectations.

API Failures

If the Open-Meteo API fails to respond or returns an error, our site displays "Updating..." or a refresh prompt. We never show fake static numbers or cached stale data. If you see "Updating..." for more than a few seconds, the API is likely experiencing issues. Refreshing the page usually resolves it.

Data Licensing

Meteorological model data from JMA, ECMWF, and other agencies is generally provided under open data policies. Open-Meteo operates under these licenses. Our presentation of the data (text, formatting, charts, guides) is our original work and is copyrighted by JP Weather Pro LLC. You may not scrape or redistribute our formatted weather content without permission, though you may reference the underlying data through Open-Meteo directly.

Contact

For data methodology questions, contact jpweatherpro@jpweatherpro.com.