Seasonal Forecast API

This API provides data from the ECMWF EC46 and SEAS5 seasonal forecast models. SEAS5 generates forecasts up to 7 months in advance, EC46 provides forecasts for up to 46 days ahead. Both models are ensemble forecast systems with 51 members, each initialized with slightly different starting conditions. This approach allows the models to estimate not only the most likely future weather but also the range of possible outcomes and their associated uncertainties.

Forecasts from EC46 for the next 46 days are updated every day at around 20:30 GMT+0. Seasonal SEAS5 forecast update every month on the 5th.

For short-term forecasting consider using the 15-days ECMWF forecast

For a detailed list of all available weather variables please refer to the general Weather Forecast API. Only notable remarks are listed below

• 6-hourly resolution: Forecast data is provided at a native 6-hour interval. While it can be interpolated to 3-hourly or 1-hourly steps, this does not increase forecast accuracy. However, using 1-hourly resolution may be practical for applications such as solar PV modeling.
• Data is available for 51 ensemble members, which can be useful for applications such as running pest or energy production models using each member individually and averaging the results. Although EC46 generates 100 ensemble members, providing 51 offers a good balance between data volume and accuracy and is sufficient for most use cases.
• Solar Radiation: SEAS5 only contains global solar radiation data and does not offer direct or diffuse solar radiation. Open-Meteo applies the separation model from Razo, Müller Witwer to calculate direct radiation from shortwave solar radiation. EC46 used for the first 46 days directly models direct and diffuse solar radiation.
• Weekly and monthly data are computed directly by the ECMWF SEAS5 and EC46 models. Monthly data is available from SEAS5 for up to 7 months ahead, based on 51 ensemble members, while weekly data is available from EC46 for up to 6 weeks ahead, using 100 ensemble members.
• Anomalies are calculated by comparing forecast values against a long-term model climatology (a baseline of past simulations from the same forecasting system). ECMWF runs past forecasts (typically covering the last 20–30+ years) using the same model configuration. These hindcasts build a model-consistent climatology. For each forecast, the anomaly is obtained by subtracting the model climatology from the forecast value: Anomaly = Forecast - Model Climatology
• The Extreme Forecast Index (EFI) highlights potential extreme events in temperature or precipitation relative to the expected climate. Rather than showing absolute values, the EFI expresses how unusual a forecast is compared to the model climate. For temperature, EFI values near +1 indicate a high likelihood of much warmer-than-normal conditions, while values near –1 suggest much colder-than-normal conditions. For precipitation, EFI values close to +1 point to much wetter-than-normal conditions, and those near –1 indicate much drier-than-normal conditions. More information in the ECMWF documentation.
• The Shift of Tails (SOT) is a complementary metric to the Extreme Forecast Index (EFI) that indicates how extreme an event could potentially become. While the EFI is limited by the historical extremes of the model climate, the SOT examines the outer tails of the forecast distribution — capturing rare, exceptional events that may exceed anything previously simulated by the model. The SOT is computed for the 10th and 90th percentiles across all 100 EC46 ensemble members. More information in the ECMWF documentation.
• Bias correction: The dataset is currently provided without bias adjustment. Future versions of the API may include bias-corrected and downscaled outputs for improved local accuracy. Parameters such as anomalies, EFI, and SOT are expressed relative to the model climate, which inherently reduces some systematic biases.