From Imagery to Data to Science

I love science. I love learning it, doing it, teaching it, and sharing it. But here’s the thing with science: it goes nowhere fast without data. That’s actually one of the things I like best about science, its course isn’t what any person or group dictates; science changes and progresses only according to the dictates of data. While I love all science, I do have my favorite corner: the geosciences. And in the geosciences, as size of research and application projects have increased, imagery has been turning out to be indispensable data for better understanding how our world works. More frequently than one might hope, we’re confronted with data that just can’t support what we’d like to do, though.

For use in science, engineering, research, and application, not all images are data. If they were, then the (in)famous site icanhascheezburger.com would be the definitive source on cats instead of a giant collection of silly pictures. It takes something more for imagery to become data. If I had location information on my snapshots, and tagged all the content in them, and aggregated them with pictures from many others, they can start to be a dataset. Also, like so many other terms in geospatial science and application, “imagery” has no official meaning or definition. For imagery to be data, it needs a certain amount of context and metadata, such as units of measure in the actual pixel values, location, coverage, sensor, and time information. No matter how nice it is a picture with nothing else about it is not data.

For it to be useful data, it may also depend on who is trying to use it for what purposes. An 8 cm pixel panchromatic image of several city blocks is probably not too helpful to a climatologist looking for initial state data; conversely, an image of global ocean salinity with half-degree cells is likely of no use for a defense analyst. One man’s data might be only a pretty picture to another. Imagery, in our field, means raster or raster-like data of known units with positional and temporal information well characterized by accompanying or embedded metadata records.

In about a month, 11 February as of the latest update, Landsat 8 will launch. For geoscientists and many others, this will probably be the biggest thing to happen in data for quite some time. The imagery, and I mean that in every sense of the word, will continue the long history of the Landsat program in checking every box required to turn imagery in to data. There is no other dataset that has the global coverage, regular repeat (as often as every two weeks), “just right” spatial resolution, and unbeatable history (carefully calibrated, characterized, and geolocated since 1972). Add it’s freely available in an online archive – there’s just no contest. The image analyst can get some great land cover, change detection, and situational awareness information. The geoscientists can get model inputs, land change and trend, vegetation maps, and more. There’s truly something for everyone in the Landsat program. Other sensors and systems can and do contribute critical data well beyond the capabilities of Landsat, but I think there’s a convincing case that Landsat is a foundation on which all others build. Here’s hoping for a successful launch, and looking forward to the flood of exciting discoveries that will come with the next generation of data coming from imagery.