AIFS -- ECMWF's data-driven forecasting system

TL;DR

AIFS is ECMWF's data-driven forecasting system that combines graph neural networks with a transformer-based processor to forecast atmospheric states. Trained on ERA5 reanalysis and operational analyses, it demonstrates competitive skill relative to IFS, especially in tropospheric and surface fields, and shows improved tropical cyclone tracks. The approach emphasizes modularity, large-scale parallelism, and rollout training to enable high-resolution, near-real-time forecasts, with open-data aspirations and plans for probabilistic extensions. Practical impact includes faster, data-driven forecasts that supplement physics-based NWP and enable ensemble-based probabilistic forecasting in the future.

Abstract

Machine learning-based weather forecasting models have quickly emerged as a promising methodology for accurate medium-range global weather forecasting. Here, we introduce the Artificial Intelligence Forecasting System (AIFS), a data driven forecast model developed by the European Centre for Medium-Range Weather Forecasts (ECMWF). AIFS is based on a graph neural network (GNN) encoder and decoder, and a sliding window transformer processor, and is trained on ECMWF's ERA5 re-analysis and ECMWF's operational numerical weather prediction (NWP) analyses. It has a flexible and modular design and supports several levels of parallelism to enable training on high-resolution input data. AIFS forecast skill is assessed by comparing its forecasts to NWP analyses and direct observational data. We show that AIFS produces highly skilled forecasts for upper-air variables, surface weather parameters and tropical cyclone tracks. AIFS is run four times daily alongside ECMWF's physics-based NWP model and forecasts are available to the public under ECMWF's open data policy.

Figures (9)

• Figure 1: Example visualisation of an AIFS encoder (left) and decoder (right) graph. ERA5 grid points are shown as small black circles, processor grid points as larger grey circles. Edges are shown as black lines. Only a small region of the globe is plotted for visibility.
• Figure 2: Attention windows (grid points highlighted in blue) for different grid points (red). The grey grid points show an example of how far information can travel within 6 layers of the processor (AIFS has 16 processor layers in total). For illustration purposes, a lower-resolution grid is shown than that used in the AIFS.
• Figure 3: AIFS encoder / decoder and processor block schematics: GNN block (left), processor block (right). The GNN block uses a multi-head graph transformer convolution operation to update the nodes and the edges of the processor, whereas the pre-norm transformer block relies on multi-head self-attention (vaswani2023attention).
• Figure 4: Northern Hemisphere ACC (anomaly correlation) of geopotential at 500 hPa for AIFS (blue) and IFS (red) and ERA5 (black) forecasts for different years: 2 day forecasts (top left), 6 day forecasts (top right) and 10 day forecasts (bottom). Forecasts are initialised on 00 UTC each day and shown is a 30 day running mean. Insets show a zoomed-in view.
• Figure 5: Scorecard comparing forecast scores of AIFS versus IFS (2022). Forecasts are initialised on 00 and 12 UTC. Shown are relative score changes as function of lead time (day 1 to 10) for northern extra-tropics (n.hem), southern extra-tropics (s.hem), tropics and Europe. Blue colours mark score improvements and red colours score degradations. Purple colours indicate an increased in standard deviation of forecast anomaly, while green colours indicate a reduction. Framed rectangles indicate 95$\%$ significance level. Variables are geopotential (z), temperature (t), wind speed (ff), mean sea level pressure (msl), 2 m temperature (2t), 10 m wind speed (10ff) and 24 hr total precipitation (tp). Numbers behind variable abbreviations indicate variables on pressure levels (e.g., 500 hPa), and suffix indicates verification against IFS NWP analyses (an) or radiosonde and SYNOP observations (ob). Scores shown are anomaly correlation (ccaf), SEEPS (seeps, for precipitation), RMSE (rmsef) and standard deviation of forecast anomaly (sdaf, see text for more explanation).