The current suite of spaceborne and in-situ assets (e.g., as deployed and operated by NASA, NOAA, and other groups) provides distributed sensing of the Earth's atmosphere, oceans, and land masses. As part of a project supported through NASA's Earth Science Technology Office (ESTO), we have developed techniques that enable these assets to be dynamically combined to form sensor webs, which can respond quickly to short-lived events and provide rich multi-modal observations of objects that are evolving in space and time.

A key focus was relating observations made by one instrument with observations made by other instruments. Techniques adapted from the fields of data mining, computer vision, and machine learning were used to automatically establish correspondence between different sets of observations. Two science scenarios were chosen to steer this development: (1) matchups between the morning and afternoon constellations of the NASA Earth Observing System (EOS) satellites and (2) correspondences between satellite and ground-based observations of clouds. The EOS matchup scenario provided increased satellite-derived information about cloud formation and development, while providing important inter-calibration information with respect to the satellite sensors and retrieval algorithms. The second scenario yielded new perspectives related to the three-dimensional structure and development of clouds.

This work was performed at the Jet Propulsion Laboratory, California Institute of Technology under contract with the National Aeronautics and Space Administration.