HR: 12:05h
AN: V32A-08    [Abstracts]
TI: Integrating TITAN2D Geophysical Mass Flow Model with GIS
AU: * Namikawa, L M
EM: namikawa@buffalo.edu
AF: Department of Geography, University at Buffalo - The State University of New York, 105 Wilkeson Quad, Buffalo, NY 14261-0023 United States
AU: Renschler, C
EM: rensch@buffalo.edu
AF: Department of Geography, University at Buffalo - The State University of New York, 105 Wilkeson Quad, Buffalo, NY 14261-0023 United States
AU: Renschler, C
EM: rensch@buffalo.edu
AF: National Center for Geographic Information and Analysis (NCGIA), University at Buffalo - The State University of New York, 301 Wilkeson Quad, Buffalo, NY 14261-0023 United States
AB: TITAN2D simulates geophysical mass flows over natural terrain using depth-averaged granular flow models and requires spatially distributed parameter values to solve differential equations. Since a Geographical Information System (GIS) main task is integration and manipulation of data covering a geographic region, the use of a GIS for implementation of simulation of complex, physically-based models such as TITAN2D seems a natural choice. However, simulation of geophysical flows requires computationally intensive operations that need unique optimizations, such as adaptative grids and parallel processing. Thus GIS developed for general use cannot provide an effective environment for complex simulations and the solution is to develop a linkage between GIS and simulation model. The present work presents the solution used for TITAN2D where data structure of a GIS is accessed by simulation code through an Application Program Interface (API). GRASS is an open source GIS with published data formats thus GRASS data structure was selected. TITAN2D requires elevation, slope, curvature, and base material information at every cell to be computed. Results from simulation are visualized by a system developed to handle the large amount of output data and to support a realistic dynamic 3-D display of flow dynamics, which requires elevation and texture, usually from a remote sensor image. Data required by simulation is in raster format, using regular rectangular grids. GRASS format for regular grids is based on data file (binary file storing data either uncompressed or compressed by grid row), header file (text file, with information about georeferencing, data extents, and grid cell resolution), and support files (text files, with information about color table and categories names). The implemented API provides access to original data (elevation, base material, and texture from imagery) and slope and curvature derived from elevation data. From several existing methods to estimate slope and curvature from elevation, the selected one is based on estimation by a third-order finite difference method, which has shown to perform better or with minimal difference when compared to more computationally expensive methods. Derivatives are estimated using weighted sum of 8 grid neighbor values. The method was implemented and simulation results compared to derivatives estimated by a simplified version of the method (uses only 4 neighbor cells) and proven to perform better. TITAN2D uses an adaptative mesh grid, where resolution (grid cell size) is not constant, and visualization tools also uses texture with varying resolutions for efficient display. The API supports different resolutions applying bilinear interpolation when elevation, slope and curvature are required at a resolution higher (smaller cell size) than the original and using a nearest cell approach for elevations with lower resolution (larger) than the original. For material information nearest neighbor method is used since interpolation on categorical data has no meaning. Low fidelity characteristic of visualization allows use of nearest neighbor method for texture. Bilinear interpolation estimates the value at a point as the distance-weighted average of values at the closest four cell centers, and interpolation performance is just slightly inferior compared to more computationally expensive methods such as bicubic interpolation and kriging.
DE: 0525 Data management
DE: 0545 Modeling (4255)
DE: 0550 Model verification and validation
DE: 8488 Volcanic hazards and risks
DE: 8499 General or miscellaneous
SC: Volcanology, Geochemistry, Petrology [V]
MN: Fall Meeting 2005