How does Google Earth work?
Short cuts bring the globe to your screen without crashing your computer.
The globe-imaging software Google Earth has become a cult web product since its release last June. Using the web-based tool, users can fly around the globe and zoom in on both natural features and whole worlds of information added by other users (see 'The web-wide world').
But how is it possible for you to zoom in from outer space to a point somewhere above the rooftop of your house without bringing your desktop computer to a grinding halt?
The main obstacle to a convincing three-diménsional skydive is data transfer. If one were to download over the Internet a one-metre resolution image of the entire world it would take 69 years with a 10-megabit-per-second Internet connection, and 12,400 years with a standard 56K modem.
To slash the amount of data they have to transmit across the Internet, virtual globes such as Google Earth approximate the sphere of the planet's surface with a polygon made up of flat tiles. The further away your viewpoint is from the surface, the fewer tiles are needed to create the illusion of roundness, and the lower the resolution of these tiles can be.
As you zoom in, the computer explodes each tile into smaller sub-tiles, each with higher resolution, and re-forms the polygon into a ball. The process continues as you zoom. This means that the virtual globe only has to download high-resolution data when the viewer is actively zooming towards it.
Virtual globes also use another trick to speedthings up further: a disk cache. Images for places you have already looked at are stored locally on your hard drive, so when you fly over this area again the software does not need to re-download the images, but instead quickly calls them up from your hard disk.
Google Earth was designed by Keyhole, a software company that Google bought in 2004. Many of the tricks it uses aren't unique: the tiling idea has been used by climate modellers, for example. But with good software and the speed of Google's enormous supercomputing platform, the results are impressive.
"I think Keyhole did a very good job," says Michael Goodchild, a geographer at the University of California, Santa Barbara. "There are some elegant design decisions in there."
Post a comment to this story by visiting our newsblog.