GribStream

GribStream Blog

ECMWF IFS Cycle 50r1 upgrade targets early 2026 operations

|

ECMWF Newsletter 185 details IFS Cycle 50r1, a coupled ocean and sea-ice upgrade with NEMO4-SI3, assimilation changes, and major compute savings, planned for early 2026 operations.

ECMWF's Newsletter 185 (published October 2025) outlines IFS Cycle 50r1, a major coupled-model upgrade planned for early 2026 operations.

Major system changes

  • Ocean and sea-ice core: replace NEMO 3.4 + LIM2 with NEMO4-SI3, and make the ocean and sea-ice coupling fully active.
  • Coupled data assimilation: introduce a coupled outer-loop (CERA-style) and run a 12-hour ocean analysis alongside the atmospheric analysis.
  • Compute efficiency: use single-precision for the coupled trajectory and reduce EDA inner-loop resolution (TL399 -> TL255), delivering about 40% compute savings for coupled data assimilation.

Forecast and physics updates

  • Convection changes and reduced vertical diffusion improve the quasi-biennial oscillation (QBO).
  • Revised gravity wave drag reduces excessive stratospheric westerlies.
  • Stratospheric humidity improved via better background error covariance for humidity and temperature.
  • Surface assimilation improves use of 2-meter temperature observations.
  • Ensemble spread: scale-selective recentering and revised stochastic physics (SPP) reduce over-dispersion in 10-meter wind.
  • Glacier and wave coupling: new glacier scheme and revised wave to sea-ice coupling.

Why it matters

Major cycle upgrades can shift skill characteristics, biases, and distributions. If you run long backtests or climatologies, plan to split evaluation windows around the Cycle 50r1 transition and re-check thresholds for sensitive variables like wind, precipitation, and temperature.

What this means for GribStream users

GribStream already supports IFS Operational. As Cycle 50r1 moves toward operations and data become available via the AWS Open Data program or other distribution channels, we will align our IFS feeds and update documentation so you can compare pre- and post-upgrade behavior.

Sources