Cloud computing allows individuals, firms, and institutions to manage and process large amounts of data faster than ever before. Landsat, NASA’s longest running initiative for the acquisition of Earth imagery, has generated nearly 50 trillion pixels of data by capturing one image per season, of every place on Earth, for the past 43 years. Now, “the cloud” has allowed researchers like Matthew Hansen and Sam Goward to make use of this abundant imagery data.
Every time a disturbance occurs to a forest, the growth cycle restarts and this can be seen in satellite images of Earth. The challenge, however, comes when working with lower resolution imagery, as it requires at least a 30 meter resolution to track smallscale changes to a forest via imagery. Hansen and Goward were bound to this low resolution data for quite some time. If resources allowed, these researchers would want to develop a live forest tracking system, that alerted a locale when forest destruction reached a high level, to the point of identifying the exact cause of the deforestation. Eventually, the team obtained Landsat imagery but due to the high cost, they could only obtain what they could afford.
At a conference in 2008, the University of Maryland team met Rebecca Moore, a Google developer, at a conference and realized the value in Google’s high-powered computing ability for their 700,000 Landsat scenes (more: New Detailed Maps Show Changes in Earth’s Forests). The team worked with Google to process these images to analyze them and track whether a pixel was forested or not, and aggregated this information to better understand forest growth cycle trends. Many methods were utilized but the system has the ability to measure the levels of RGB color density in each image pixel. By doing so, this allows the research team to hone in on those smallscale changes by tracking the variation in this color density over time. In total, the analysis process required 10,000 central processing units and took 1 million hours a process that would have taken 15 years on a single computer.
One of the first case studies for this method looked at the Democratic Republic of Congo and found significant deforestation between 2000 to 2010. This amounted to 5,572 teragrams of carbon lost due to slash and burn for agriculture and the need for wood as a fuel source. For nation’s like DCR, which lack any form of a forest or tree inventory, there is incredible value to those making land use and resource planning decisions. These images offer policymakers the most succinct understanding of deforestation. As this method gets more refined, the University of Maryland team hopes to expand the application of this tree inventory to other areas like tracking human health, protected nature areas, and modeling biodiversity.
Big Data Helps Scientists Dig Deeper by Holly Riebeek, Earth Observatory, NASA. March 26, 2015.