GribStream Blog
ECMWF AIFS operational and ensemble now on GribStream
AIFS operational and ensemble datasets are live on GribStream, providing AI-based global forecasts with deterministic and probabilistic outputs through stable dataset codes.
On July 31, 2025, GribStream added ECMWF AIFS Operational and AIFS Ensemble datasets. This brings AI-driven global forecasts into the same API you already use for physics-based numerical models.
The GribStream dataset codes are:
- AIFS Operational:
aifsoper - AIFS Ensemble:
aifsenfo
Current note: this post records the July 2025 GribStream dataset launch. ECMWF later moved AIFS Single and AIFS ENS to v2 at the 2026-05-12 06 UTC run. GribStream keeps that history in the same dataset timelines; see AIFS v2 fields are now visible on GribStream for the May 2026 catalog update.
About AIFS
AIFS is ECMWF's AI-based global forecasting system. It is operational guidance, not only an experimental chart product. ECMWF implemented AIFS Single operationally in February 2025, and implemented the ensemble version, AIFS ENS, on July 1, 2025.
The public open-data subset that GribStream serves uses:
-
00/06/12/18 UTC cycles
-
6-hourly steps out to 360 h (15 days)
-
0.25 deg global latitude/longitude output
-
AIFS Oper provides a single best-estimate forecast.
-
AIFS Ensemble provides a 51-member ensemble (50 perturbed + 1 control) for probabilistic workflows.
The ensemble member structure matters. It lets you evaluate AI guidance not just as one deterministic forecast, but as a distribution of possible outcomes.
Why it matters
AI forecasts are now operationally relevant, and ensembles let you quantify uncertainty instead of relying on a single scenario. With AIFS in GribStream, you can:
- Compare AI vs. physics guidance (AIFS vs. IFS).
- Build percentiles and probability of exceedance products.
- Run backtests using
asOftime travel for consistent evaluation. - Track behavior across AIFS version boundaries without changing GribStream dataset codes.
The practical workflow is simple: use aifsoper when you need a single AI forecast, and use aifsenfo when you need ensemble spread, percentiles, or event probabilities.
Links
Sources
- ECMWF implementation page for AIFS Single v1: https://confluence.ecmwf.int/display/FCST/Implementation%20of%20AIFS%20Single%20v1
- ECMWF news on AIFS ENS becoming operational: https://www.ecmwf.int/en/about/media-centre/news/2025/ecmwfs-ensemble-ai-forecasts-become-operational
- ECMWF implementation page for AIFS ENS v1: https://confluence.ecmwf.int/display/FCST/Implementation%2Bof%2BAIFS%2BENS%2Bv1
- AIFS v2 fields on GribStream: /blog/aifs-v2-parameters-now-on-gribstream