Dreaming of Hypermaps

The hypermap would be ideally suited to do what is known in the data trade as fusion: taking different data sets and combining them to produce something new and, one would hope, more informative. Fusion is the essence of mapmaking. A cartographer creates a paper map from survey data and aerial photographs. Information is added from other maps — the cartographer uses a pen or a mouse to draw in roads and county boundaries, and writes or types names such as "Wyre Piddle" next to a beautiful English village that sits next to a small stream. Information is cross-checked (the cartographer compares the scattered height data from the survey with the contour lines drawn from stereoscopic pairs of aerial photographs) and updated (a reservoir's shoreline is redrawn because the dam is now higher than it was). The hypermap should do all this as well, while combining a high-resolution photograph's intense feeling of reality with a cartographer's knowledge of the geographical significance of the landscape. Some maps already approach this — on my office wall hangs an image of Los Angeles as seen from space. The green, forested mountains, the gray-brown urban area, the brilliant white dry lake beds in the red-brown Mojave desert, the San Andreas fault, and a lot of other physical features are shown in absorbing detail. But major roads and place names are also marked, allowing us to get our bearings. We can relate what we see anew to what we already know, and thereby create knowledge from data.

The personal GIS already exists, but without the use of remote servers and the possibility of junk mail: there's a CD-ROM called Street Atlas, which gets the most use of the dozen or so CDROMs that my wife and I have bought since getting a CD-capable machine. (Street Atlas even has the scale-doubling feature, though I would like to point out that I thought of it before I got the CD!) In one sense, these little plastic disks hold a great deal of data: one of today's CDs can hold enough novels to read one a week for ten years, and soon it will be possible to pack a lifetime's supply into an alluring plastic rainbow. But for storing images or geographic data the CD-ROM seems much smaller — more like a single drawer than a library — simply because a page of image takes up a lot more bits of data than a page of text. Consequently, Street Atlas's maps do indeed cover the whole United States, but the streets are really just named geometric lines, all the same width. And there's no indication of any texture to the landscape: no forests, ruins, salt marshes, glaciers, battlegrounds, or wind farms.

There's another, more compelling reason why a physical data repository — even a roomful of CDs — cannot be adequate for the full efflorescence of the hypermap. The problem is that a single physical object is tangibly limited; no matter how much data it holds, it is obviously finite in extent. In contrast, a networked hypermap can explore a potential infinity of data and can go anywhere in the world to get it. True, the amount of digital geographic data in the world is also finite, but the difference is that it is growing. The hypermap concept carries with it the idea that as data is newly minted from cartographers and orbiting satellites, then all the world's hypermaps immediately gain the new depth. Thanks to airplanes and satellites, the quantity of geographic data is increasing even faster than the violent increase in computer speed that's changing society so quickly. When up-to-date maps of essentially infinite detail are combined with cellular phones and high-accuracy geographic positioning systems, can the concept of an expedition into the wilderness survive.

The Internet is providing this kind of three dimensional experience today, meaning that all you need is a fancy computer, rather than having to know the right people in addition to having a fancy computer. At the moment, one can tour, among other places, an Italian castle, Jerusalem's city hall, and the city of Berlin. The new protocol that enables this to occur is called Virtual Reality Modeling Language, or VRML; when you download a VRML document, your computer opens a 3D "browser" that reads the file and allows you to explore the space encoded within it — turning, twisting, accelerating, panning, and zooming at your whim. You might even meet representations of other people, known in the trade as avatars. (If I am ever virtually represented in a 3D space, I would like my avatar to be the boot token from the Monopoly game.) And VRML isn't just for architectural touring or city planning; it's a method of transmitting any kind of three-dimensional data for interactive exploration Other Net sites allow the user to wander about within molecular structures, the fruit fly's nervous system, and galaxies, to name a few.

And the data aren't just digital photographs, but the outputs of other sensors that have nothing to do with visible light and work instead in infrared or microwave frequencies. There's a strong analogy here to astronomy, which was confined to optical observations until the arrival of radio telescopes. Today, a burgeoning family of telescopes observes the entire electromagnetic spectrum, neutrinos, and soon even gravity waves. The invisible emissions captured by these instruments have provided a new view of the universe, revealing it to be a violent and capricious place, in sharp contrast to the quintessentially perfect "music of the spheres" of the medieval imagination. In the same way, the wide availability of high-resolution geographic data will change our view of Earth, making it at the same time more familiar, more mundane, more complex, and more precious.

In order to see so much so clearly, there is of course a price to pay — the raw data from the satellite are not directly visible, but need to be processed by a supercomputer before becoming intelligible. Such a project has been under way at Caltech and JPL for two years now, and it's a massive endeavor involving many people. We feed Intel and Cray parallel supercomputers with tapes of raw data and receive multichannel color images in exchange. Every pixel in a color image conventionally represents three data channels, encoded in the colors red, green, and blue, which correspond to the three kinds of receptors in the human eye. But SAR takes data at eight or more channels, leaving a choice of how to throttle the flow down to only three. Such filtering choices can be made to emphasize different aspects of the terrain; for example, to identify types of trees or the composition of volcanic lava, or to gauge the quality of potential ski slopes.

For the armchair explorer, part of the exhilaration of this remote-sensing data is that it has not been processed and digested by a human, only by a computer. There may be the ruins of a hidden city, barely visible without contrast enhancement, or a lake forming where none was known before. "Traveling" by means of SAR data carries the possibility of discovery, much like that offered the patient comet-watchers, roving the sky with binoculars. By comparison, making the trip by paper map will feel like looking at a star atlas instead of looking at the sky. SAR data are not simple pictures to be examined, but can be reprocessed in many ways — just as statistical data can be massaged and processed to highlight, emphasize, and maybe even cheat. When we combine SAR data with conventional maps, we can see correlations and associations that were previously hidden, thereby creating knowledge, and, who knows, perhaps a scrap of what we all crave: insight.

Amazon rain forest, Brazil

Nympenburg Palace, Munich, Germany

Irrigation in the delta of the Yangtze River, China