To coincide with the opening of the Copenhagen Climate Summit, Google.org announced a collaboration with the Carnegie Institution for Science to build an online version of the Carnegie Landsat Analysis System (CLAS). The existing CLAS system is a desktop tool that supports conversion from the raw satellite imagery, calibration, atmospheric correction, cloud masking and spectral analysis to create maps of forest cover, deforestation, and forest disturbance that can be overlaid with other geographic data. The new version of the software, called CLASLite, does all of this online.
The Google.org folks write:
What if we could offer scientists and tropical nations access to a high-performance satellite imagery-processing engine running online, in the “Google cloud”? And what if we could gather together all of the earth’s raw satellite imagery data — petabytes of historical, present and future data — and make it easily available on this platform? We decided to find out, by working with Greg and Carlos to re-implement their software online, on top of a prototype platform we’ve built that gives them easy access to terabytes of satellite imagery and thousands of computers in our data centers.
Geoprocessing in the cloud with petabytes of satellite imagery while reducing computation from days to seconds. That’s a compelling vision. The prototype, Earth Engine, is not yet available to the public, but Google has pledged to make it accessible for free to any tropical country. And while the initial target of this effort is deforestation, it seems only logical that the Earth Engine could very well be extended to cover other types of geoprocessing.
Distributing geoprocessing has been on its way for a while. Wolfram Research has been offering the server version of its Mathematica product as a way to distribute mathematical and statistical processing across many machines in a network. Brian Flood has done a fair amount of work on cloud-based geoprocessing with his Arc2Earth Cloud Services. At Azavea, we’ve designed our own DecisionTree raster processing framework to both distribute work across multiple machines/processors/cores as well as be able to run in the Amazon Web Services EC2 environment. Each of these examples is aiming at several benefits:
- Speed: desktop processing can take many minutes and even hours to complete. By distributing the work across dozens or hundreds of machines, we can get responses that are fast enough to display the results in “web time” – a second or two.
- Lower Cost: If we can acquire processing power as we need it, rather than buying and maintaining hardware and disks ourselves, we can lower the cost of computing substantially.
- Simpler UI: By complex processing to be performed on the web, we can create crafted user interfaces that focus on the needs of a particular workflow rather than requiring that someone learn the far more complex tools in a Desktop GIS.
I’m pretty excited the prospects for bringing analytical and statistical services to a much larger audience via cloud services.
