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:
- JMA MSM (Meso-Scale Model): Japan Meteorological Agency's regional model. Horizontal resolution of approximately 5km. Covers the East Asian region including all of Japan. Updates every 3 hours (00, 03, 06, 09, 12, 15, 18, 21 JST). Forecast range: 33 hours (longer-range forecasts use GSM).
- JMA GSM (Global Spectral Model): JMA's global model. Coarser resolution (approximately 20km over Japan) but longer forecast range (11 days). GSM provides the 3-day forecasts we display when MSM doesn't cover the full period.
- ECMWF IFS (Integrated Forecasting System): The European Centre for Medium-Range Weather Forecasts' global model. Widely considered the most accurate global NWP model. Open-Meteo blends ECMWF data with JMA models for ensemble forecasting.
- GFS (Global Forecast System): NOAA's global model. Used as an additional ensemble input.
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:
?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:
- Current conditions: Temperature at 2m, relative humidity, apparent temperature (feels-like), WMO weather code, precipitation amount, and wind speed at 10m height.
- Hourly forecast (24 hours): Temperature, precipitation probability, and weather code for each hour.
- Daily forecast (3 days): Maximum and minimum temperature, total precipitation, and maximum UV index.
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:
- JMA MSM: Every 3 hours (8 runs per day)
- JMA GSM: Every 6 hours (4 runs per day)
- ECMWF IFS: Twice daily (00Z and 12Z runs)
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:
- Add artificial "feels like" corrections beyond what the API provides.
- Smooth temperature curves to make them look prettier.
- Round temperatures differently to make conditions seem milder.
- Suppress precipitation probability to make forecasts seem more favorable.
- Apply any city-specific calibration factors.
The only transformations we perform are:
- Rounding temperatures to the nearest whole degree for display.
- Converting weather codes (WMO standard numerical codes) to human-readable text ("Clear sky", "Moderate rain", etc.).
- Converting ISO timestamps to local Japan Standard Time.
- Building SVG polylines from hourly temperature arrays for the 24-hour chart.
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 |
|---|---|---|---|
| Tokyo | 35.6762 | 139.6503 | Imperial Palace district |
| Osaka | 34.6937 | 135.5023 | Central Osaka (Umeda-Namba midpoint) |
| Yokohama | 35.4437 | 139.6380 | Kannai/Minato Mirai area |
| Nagoya | 35.1815 | 136.9066 | Sakae/Nagoya Station area |
| Sapporo | 43.0618 | 141.3545 | Odori Park district |
| Fukuoka | 33.5902 | 130.4017 | Tenjin/Hakata midpoint |
| Kobe | 34.6901 | 135.1956 | Sannomiya 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:
- Compare the previous day's forecasted temperatures against actual AMeDAS recordings.
- Log discrepancies greater than 2°C as notable.
- Identify patterns: consistent overestimation/underestimation by city or season.
- 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.