2004 Denver Annual Meeting (November 7–10, 2004)

Paper No. 6
Presentation Time: 3:05 PM

AUTONOMIC FUSION OF INFORMATION FOR MONITORING, CHARACTERIZING, AND FORECASTING SUBSURFACE PROCESSES


YEH, Tian-Chyi Jim, Hydrology and Water Resources, Univ of Arizona, John Harshbarger Building, 1133 E. North Campus Drive, Tucson, AZ 85721, yeh@hwr.arizona.edu

Currently, we lack the capability to generate 3-D pictures of the Earth’s subsurface distributions of water and related properties. 3-D pictures are necessary to improve our understanding and management of groundwater resources. Existing monitoring and characterization technologies cover only a small fraction of the subsurface, and their outputs cannot be used to reliably evaluate current and future water-availability issues. We are taking on the challenge of developing a system for subsurface simulation and imaging at the basin scale, the appropriate unit for water resources management issues. This will require detailed knowledge of the variability and characteristics of geologic formations at scales from meters to kilometers throughout the basin. Inverse modeling combined with hydrologic or geophysical tomographic surveys has become a viable, high-resolution characterization tool for meter-scale field problems. “Seeing” into a basin, however, requires significant scaling up and integration of different high-resolution tools. This will demand unprecedented levels of computation and information processing. Moreover, current methods relying on locally induced artificial stimuli (e.g., pumping at wells and geophysics) are much too costly to provide dense coverage over a basin-size region. To overcome these impediments, we are exploring the possibility of exploiting natural stimuli as large-scale hydraulic or geophysical tomographic surveys, supported with modeling and near real-time, networked information fusion technologies. Naturally recurring stimuli (e.g., lightning, storms, floods, earthquakes, etc.) can provide a sufficiently varied distribution of excitations in time and space to obtain the requisite basin responses. We expect to develop: 1) extensive, spatially-distributed, non-invasive, smart sensor networks to gather massive geologic, hydrologic, and geophysical data; 2) stochastic information fusion methods based on dynamic tomographic surveys; 3) spatially-explicit dynamic models of the subsurface to support such iterative information fusion; and 4) asynchronous, parallel/distributed, adaptive algorithms for rapidly simulating the states of a basin at high resolution. Our approach will provide much needed tools to effectively manage our water resources.