GraphCast and AIFS catch light rain in Luxembourg's flood events

For three Luxembourg flash floods, GraphCast and AIFS detected light rain and forecast temperature better than a regional model that assimilates observations. The regional model won on heavy rain.

  • Not rated an organization

    The forecasts were run for past floods, and no warning service is reported to have used them.

Depth runs from 1, a convenience or time saved, to 5, a life or a livelihood preserved.

Status
Reported and intended benefits.
Evidence

foundstudiedreportedclaimed

An independent peer-reviewed study supports the benefit.

Compared with

Regional WRF model that assimilates observations with 3D-VAR. Measured.

Without the system
CSI 0.389 for the assimilated WRF at 1 mm
With it
CSI 0.424 for GraphCast at 1 mm
The system’s part
All systems drew their initial and boundary information from ERA5 and were checked at the same independent stations. Their effective starting states differed, since WRF was cycled and corrected by observations.
Limits
Not yet checked.
Costs and harms
Not yet checked.
Later
Not yet checked.

Download data Cite Suggest a correction

Benefit type
Improved Task Accuracy
System
GraphCast, AIFS Single 1.0 and FuXi
Made by
Google DeepMind (GraphCast), ECMWF (AIFS) and Fudan University with the Shanghai AI Laboratory (FuXi)
Where
Luxembourg and the Greater Region (flash floods of 2016, 2018 and 2021)
Stage
Tested in a laboratory
Date
, when the source was published
Kind
Aggregate result
Broad benefit
Estimating the Quantifiable Characteristics of Products, Events, or Information
Last revised
31 October 2026

Sources

  1. Peer-reviewed study, by authors with no tie to the system.

    Rehman HU, Teferle FNR. Physics vs. AI in weather prediction: evaluating GraphCast, AIFS, and FuXi against an observation-corrected WRF model for flash floods. Geoscientific Model Development 19, 9831-9855 (2026).

    At the 1 mm threshold AIFS achieves the highest detection rate (POD 0.835) and GraphCast the highest net categorical skill (CSI 0.424, against 0.415 for AIFS and 0.389 for the assimilated WRF), and AIFS and GraphCast reduce temperature RMSE by 19.0 % and 17.6 % relative to assimilated WRF

Details

AI system

Workflow

The AI models and WRF drew initial and boundary data from ERA5 for about 90 simulated days each. Forecasts were scored at stations screened to be independent of the assimilated observations.

Follow-up

Conditions for replication

Forecasts of the heavy rain that causes floods need the regional model, which scored better at high thresholds. The event must lie outside the AI model's training data for a fair test.

Entry details

Sector

Weather forecasting / flash floods.

Purpose

Google DeepMind built GraphCast, ECMWF built AIFS and Fudan University with the Shanghai AI Laboratory built FuXi to forecast global weather with machine learning.

Not yet checked: model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.

Cite this record

Peter H. Vartanian. (2026). GraphCast and AIFS catch light rain in Luxembourg's flood events (AIBD-0766). Artificial Intelligence Benefits Database, edition of 31 October 2026. https://petervartanian.xyz/aibd/records/AIBD-0766.html (archived at https://doi.org/10.7910/DVN/AADSII, version 4.0)

BibTeX Data Field definitions