ECMWF's free 9 km forecasts: why finer weather detail matters

A weather forecast can describe the broad pattern correctly and still miss an important difference between two nearby places. A coastal wind farm, an inland field and a town on higher ground can experience different conditions under the same passing weather system. Finer spatial detail helps us ask better questions about those differences.

That is why ECMWF's planned free 9 km forecast subset matters. In a September 18, 2026 support update, ECMWF said it expected publication within a few weeks, with two-hour latency; the exact date remained unconfirmed. The scientific opportunity is to make more of the detail in ECMWF's forecasts accessible to people building weather-dependent models and decisions.

A clearer view of an existing forecasting system

This announcement concerns access to finer forecast output. It does not mean ECMWF is only now beginning to predict the atmosphere at 9 km.

ECMWF brought its medium-range ensemble to approximately 9 km in 2023, matching the high-resolution forecast. Its scientific account of that upgrade explains how finer resolution, together with changes to the model and its use of observations, improved forecasts of surface weather and tropical cyclones. In particular, finer resolution helped represent the strong horizontal gradients within intense cyclones.

Those results demonstrate the scientific value of model development. They are not a measured accuracy gain for the forthcoming free subset relative to today's 0.25° output. The new access would let more users investigate the existing forecast at a finer spatial scale.

Terrain, coastlines and the conditions people experience

Mountains steer air, force it upward and influence where rain falls. Land and sea warm differently, affecting local temperature and wind. When a grid is coarse, nearby places with different elevations or surface conditions can share a similar representation in the forecast.

A finer grid can retain more of these spatial contrasts. But even 9 km leaves much unresolved: a sheltered valley, an exposed ridge or an individual convective downpour can be far smaller than the grid. ECMWF's forecast guidance on known issues explains why terrain-related rainfall enhancement and local rainfall extremes remain challenging. A grid-box rainfall value should not be interpreted as the exact amount at every rain gauge inside it.

For a user, the practical question is whether that extra spatial information changes a useful decision. Does it distinguish a coastal asset from an inland one? Does a site near complex terrain behave differently from the surrounding region? Those are questions to evaluate against local observations.

Where finer forecasts could help

These are potential applications to test when the new subset becomes available:

• Wind energy: compare forecast winds across coastal, inland and elevated sites, then assess whether the added spatial detail improves power forecasts or identification of sharp changes in generation. Turbine height, local terrain and the conversion from wind to power still matter.
• Agriculture: examine temperature and rainfall differences across a growing region when planning field work or assessing frost and water-stress exposure. A finer forecast can inform the regional picture; field observations remain important for decisions at farm scale.
• Energy demand: evaluate whether local temperature patterns better represent the areas driving heating or cooling demand, especially where coast, elevation and urban development vary within a service territory.
• Weather risk: revisit past forecast cases to see whether the finer output better distinguishes the areas exposed to strong winds or heavy rain, while retaining uncertainty about timing and location.

The benefit should be judged using the outcome of interest—power, demand, rainfall thresholds or event detection—rather than the appearance of a sharper weather map alone.

Resolution is one part of the science

Forecast skill also depends on the observations used to initialize the atmosphere and on how the model represents clouds, soil, vegetation and exchanges of heat and moisture.

ECMWF's account of its 2024 near-surface forecast improvements describes changes to temperature-observation assimilation, land-surface modelling and the representation of uncertainty. Its May 2026 IFS update also improved the representation of convective precipitation. These developments explain why a resolution number alone cannot summarize forecast quality.

For decisions sensitive to extremes, a detailed single forecast still describes only one possible evolution of the atmosphere. Compare it with observations and, where available, ensemble information rather than treating finer spacing as certainty.

What comes next

The planned release creates an opportunity to test how much finer ECMWF output helps real applications. The most informative comparisons will use the same locations, forecast lead times and decision deadlines: a forecast is useful only if it arrives before the decision it supports.

GribStream currently offers the 0.25° IFS operational forecast. We are investigating the higher-resolution subset, but it is not yet available through GribStream and no GribStream release date has been announced. We will update this article when the public release is confirmed.