Since its launch on Feb. 11, 2013, scientists have been working to understand Landsat 8 Earth observation satellite data. Some have been calibrating the data—checking them against ground observations and matching them to the rest of the 42-year-long Landsat record. At the same time, the broader science community has been learning to use the new data.
There are many ways to produce a composite map. David Roy, a co-leader of the USGS-NASA Landsat science team and researcher at South Dakota State University, breaks each Landsat scene into its smallest component, the picture element or “pixel.” For Landsat, every full-color pixel corresponds to a 30- by 30-meter area, a chunk of land about the size of a baseball diamond. Those pixels are accurately matched to a map, or “geolocated.”
Using a computer algorithm, all 11 billion pixels in the contiguous United States (most of the lower 48—a small section of Maine is missing) are examined to identify the most cloud-free pixels. The best pixels are assembled into a single, composite image, so a pixel acquired on the first of the month might be next to one acquired a few weeks later. To make the map manageable, the image is reduced so it has a resolution of 300 meters per pixel.
The strips in the accompanying image are a result of the way Landsat 8 operates. Like its predecessors, Landsat 8 collects data in 185-kilometerwide strips called swaths or paths. Each orbit follows a predetermined ground track so the same path is imaged each time an orbit is repeated. It takes 233 paths and 16 days to cover all of the land on Earth. This means that every land surface has the potential to be imaged once every 16 days, giving Roy two or three opportunities to get a cloud-free view of each pixel in the United States in a month. Some northern latitudes are imaged in more than one path.
Image courtesy of NASA.