Satellite images allow Earth System scientists to monitor changes in vegetation, map species distributions, and even measure surface temperatures. The amount of data available for such efforts is enormous; the U.S. government’s Moderate Resolution Imaging Spectro-radiometers (MODIS), whose polar orbits circle the Earth in opposite directions, view the entire Earth’s surface every 1 to 2 days and generating about 400GB of data per day. One of the most difficult challenges of using the data they gather is accounting for the presence of clouds, which cover up to 70% of the earth’s surface and tend to obscure the satellite’s view of Earth’s surface.
MODIS uses a combination of algorithms to identify and mask clouds, but there has been little discussion about their performance over different parts of the world. A recent YCEI study found that one of the cloud detection programs overestimates the frequency of clouds over some land cover types (such as grassland and savanna) by up to 20% in some areas around the world. These artifacts have resulted in significantly decreased and spatially biased data availability for MODIS data products such as land surface temperature and ecosystem productivity. This important issue calls into focus the complexity of working with satellite-derived information and the need to accurately account for uncertainties in the data. Our findings suggest that both re-interpretation of some previous MODIS based results that are sensitive to the demonstrated biases and a cautious future use of data derived from these products may be necessary.
Wilson, Adam M., Benoit Parmentier, and Walter Jetz. 2014. “Systematic Land Cover Bias in Collection 5 MODIS Cloud Mask and Derived Products — A Global Overview.” Remote Sensing of Environment 141 (February 5): 149–154. doi:10.1016/j.rse.2013.10.025. http://www.sciencedirect.com/science/article/pii/S0034425713003933