Adaptive Fusion of Stochastic Information for Subsurface Imaging of Fractured Vadose Zones

Acknowledgment: NSF Award Number  0430884

NSF Organization: IIS (Division of Information & Intelligent Systems)

NSF Program: ITR-INFORMATION INTEGRATION

ObjectivesResultsReportsStudentsPublications

Objectives

The objective of this project is to develop technologies that combine a newly developed stochastic information fusion algorithm with dynamic, adaptive, parallel computing technologies to efficiently fuse different types of hydrological and geophysical information. The stochastic algorithm will be implemented on an innovative computing platform to provide a cost-effective monitoring, characterization, and predictive technology for the Vadose zone (VZ, fractured media between land surface and water table).

The technologies are in place that will enable:1) the development of physical scale models for hydro-factures at the University of Iowa, 2) a successful, highly efficient finite difference time-domain electromagnetic modeling algorithm from The Ohio State University, 3) improved algorithm for stochastic inversion of hydrological data from the University of Arizona, and 4) computational and computer science developments from Rutgers University.

P.I.:  Tian-Chyi Jim Yeh, University of Arizona            Co-P.I.:   Salim Hariri, University of Arizona  

Co-P.I.:  Jeffrey Daniels, Ohio State University            Co-P.I.:  Walter Illman, University of Iowa     

Co-P.I.:  Anton Kruger, University of Iowa                  Co-P.I.:  Manish Parashar, Rutgers University

For our purposes, the task components and information flow can be described for each primary task (indicated by the green pentagons) as follows:

I.        Experiment: The experiment is stand-alone. The fundamental input to the experimental tank is fluid (water) that is pumped into the model. Measurements (geophysical ERT and GPR, and hydrologic tomography) are made continuously after initiation of the experiment. These measurements can stream into the network upon demand from the Interpretation Fusion Controller (IFC).

1.      The hydraulic controller injects fluid into the model.

2.      Data from geophysical and hydraulic measurements, including time stamp, are stored in addressable sequential data buffers at the experimental site.

3.      Data are accessed upon demand from the interpretation node (3). 

II.     Control: The Network (internet, or satellite) is the hub for control and data flow. The heart of the hub is the IFC and a real time display that monitors the model as it is developed and acts as a controller to update and transmit the most recent model realization(s) back to the individual modeling nodes, depicted on the right hand side of Figure 1.

1.      Construct basic physical model (size, shape, other fixed parameters from static data base.

2.      Monitor and interrupt geophysical and/or hydraulic tomography models. Update and display evolving model output from geophysics and hydraulic tomography measurements.

3.      Validate model through hydraulic monitoring display.

4.      Provide updated model as exported boundary control for geophysics and hydraulic tomography.

 III.   Modeling: Two of the three modeling hubs (geophysical, and hydraulic tomography) are stand-alone-computer forward/inverse model systems that individually require intensive distributed computations (computations on a grid). The hydraulic monitoring program acts as an overall control on the process: if no changes in flow rates are noted over a pre-determined period of time, then the inversion process is stopped.

 IV.  Distributed Grid Computing: Grid and parallel computing is controlled solely by the individual computational controllers in each of the modeling functions in System Component III.