Visualizing machine learning

Sketching became our common language with machine learning scientists to accurately visualize ECMWF’s complex AI-driven weather forecasting system.
Client

ECMWF

Project TypeExplainers
A compilation of the final visualizations created for the European Centre for Medium-Range Weather Forecasts, sat side by side on a plain green background.

If we’ve worked together, you’ve probably heard us say, "You can’t design what you can’t understand." This was especially true for our machine-learning project with the European Centre for Medium-Range Weather Forecasts (ECMWF).

ECMWF has pioneered using machine learning to model the world’s weather and climate. Our brief was to create a set of static visualizations explaining their latest technological feat: the AI Integrated Forecasting System (AIFS).

Working closely with ECMWF’s experts, we ran a series of sketch sessions. These allowed us to capture the key details about how this sophisticated AI-driven weather forecasting system works – then distil them into clear, easy-to-understand visuals.

The visual language stayed true to the complexity of AIFS and ECMWF's machine learning advancements, while being simple enough to spark conversations about these innovations with Member States, policymakers, and journalists.

One machine learning scientist told us that we’d ”done the impossible” by explaining the super-complicated process of training an entire weather forecasting system in a single image.

This reinforced our belief that if you invest the time in understanding information, you can design it in ways that others can understand too.

An animation showing an initial sketch of a visualization and a revised more refined version of the sketch
A final visualization of a conventional integrated forecasting system (IFS), showing 4 stages: Data assimilation, initial conditions, modelling and predictions
A final visualization of a hybrid integrated forecasting system (IFS), with machine learning, showing 4 stages: Data assimilation, initial conditions, modelling and predictions
A final visualization of an AI forecasting system (AIFS), with machine learning used in segments, showing 4 stages: Data assimilation, initial conditions, modelling and predictions
An animation showing an initial sketch of a visualization and a revised sketch of the same visualization
A final visualization of the process to train an AIFS machine learning (ML) model, showing 3 parts: Training data feeding into a training loop producing and ML model

Get in touch:hello@beyondwordsstudio.com

beyond words logo
Follow us
TwitterInstagramLinkedInYouTube
Find us
Beyond Words Studio UK 91 Brick Lane London, E1 6QL
Beyond Words Studio US Florida, USA
Beyond Words Studio EU Dublin, Ireland
Copyright Beyond Words Studio Ltd. 2024