Satellite and ground-based remote sensing produces large quantities of heterogeneous and multisource spatial-temporal data. This offers promise for highly accurate retrievals of geophysical parameters on a global scale, but it also opens a number of computational challenges related to the construction of efficient and accurate retrieval algorithms. This project is exploring this opportunity through development of data mining methods to: (a) improve existing single-sensor retrieval algorithms; and (b) allow high-quality joint-sensor retrieval. Aerosol-related data from the TERRA and AQUA satellites and the AERONET network of ground-based sensors is being used for development and validation of the proposed algorithms. The intellectual merit of the proposed work is in: addressing this challenge through the use of complex forward-simulation models; exploring spatial-temporal properties of large, heterogeneous and multi-resolution data; conditional probability modeling and uncertainty estimation; sampling design; use of advanced data structures; and integration and handling of multi-TB data. The broader impacts stem from development of accurate aerosol retrieval methods that will allow improved characterization of the effects of aerosols on the Earth's energy and water cycles. Additionally, the project is providing guidelines for developing accurate and fast retrieval algorithms in other geoscience applications, and will lead to advancements in spatial-temporal data mining. The project is assuring a broad participation of students through incorporating the research results into several courses, exposing diverse groups of students to research, and widely disseminating the results through publications and the project web site (www.ist.temple.edu/IIS-0612149).
