Wind Study

Wind Study makes weather a collaborator in a digital drawing. Built with artist Jitish Kallat, it uses atmospheric data to steer virtual flames across a curling mathematical line. Choose somewhere on the globe and the wind gets a say in the resulting mark.

What is Wind Study?

The starting pattern is a Hilbert curve: a line that folds repeatedly through a square. Google’s experiment connects that geometry to wind conditions from around the planet, so the drawing has both a strict template and an unruly partner. The documented sequence moves from picking a location to lighting the pattern with a digital match, then saving the wind-shaped result. Kallat’s physical Wind Study drawings use actual burning to record the air’s movement. This browser version translates the idea into simulated fire rather than asking visitors to handle a flame. It offers an unusual way to think about weather: as something that can leave a picture behind, rather than just a number beside a cloud icon.

What you can do there

• Choose a planetary location
• Ignite a digital curve
• Save a wind-shaped drawing

Why we picked it

Weather maps normally help you decide whether to bring a coat. Here, the same broad subject becomes a drawing material. The connection to Kallat’s physical practice gives the wind-driven marks a clear artistic reason to exist.

How to get the most from it

Begin with a place you know, follow the digital ignition sequence and let the result develop. Then choose a distant location and compare the marks. Use the saving option to keep an example; treat the comparison as an art exercise rather than a weather measurement.

Good to know

The experiment uses WebGL graphics and simulated flames. Google identifies NOAA/NCEP atmospheric data as its input and describes a six-hour refresh cycle. The output is an artistic interpretation of that data; it is not a forecast or a guide to lighting physical drawings.

Who made it, and when?

Jitish Kallat with Google Arts & Culture Lab is the credited creator or organization. The earliest supported launch date we found is February 3, 2026.