OREGON STATE UNIVERSITY

You are here

The Tsunami Computational Portal: Distributed Infrastructure for Executing and Comparing Multiple Computational Models

TitleThe Tsunami Computational Portal: Distributed Infrastructure for Executing and Comparing Multiple Computational Models
Publication TypeConference Paper
Year of Publication2010
AuthorsKeon, D., C. M. Pancake, H. Yeh, and T. Logan
Conference NameAssociation of American Geographers (AAG)
Date Published03/2009
Keywordscomputational models, cyberinfrastucture, distributed computing, GIS, high-performance computing, open source, spatial databases
Abstract

The relative isolation of useful computational models and the difficulty of comparing alternative approaches are common problems shared by most research disciplines. Models have unique input/output formats and parameterization schemes, and the codes may require specific hardware/software environments. This makes it difficult to run similar jobs against different models in order to compare results. The Tsunami Computational Portal (TCP) is a cyberinfrastructure service that brings together computational models in a common environment to facilitate comparison. Its web-based interface facilitates executing the same inputs and controls against multiple models, or modifying parameters incrementally for a particular model. Researchers retrieve and visualize the results using tools available in the same interface. The wizard-style interface is driven by a large PostgreSQL database (spatially-enabled by PostGIS) that contains both global and local bathymetry/topography datasets. A custom mapping tool allows users to select nested bathy/topo grids of interest, automatically snapping them to parent grids in the database to satisfy alignment requirements of the model codes. Additional custom utilities automatically handle the export of spatial objects in the binary format required for processing on supercomputers and manage communications among components at Oregon State University's Northwest Alliance for Computational Science & Engineering and the Arctic Region Supercomputing Center. Automatic post-processing code generates images and video of each model run. TCP has made it possible to compare tsunami models and observe the effects of alternative numerical methods. A similar approach would benefit any domain where alternative models need to be exercised broadly and compared to increase scientific understanding.