SC Conference - Activity Details

Scaling the Science: Volcanic Hazard Analysis using HPC, Hazardous Mass Flow Modeling, Statistical Modeling and Paralllel Analytics

Abani Patra  (National Science Foundation)
Keith R. Dalbey  (University at Buffalo)
Matthew Jones  (University at Buffalo)
E. Bruce Pitman  (University at Buffalo)
Eliza S. Calder  (University at Buffalo)
Posters Session
Tuesday,  05:15PM - 07:00PM
Room Rotunda Lobby
We show that the efficient use of statistical models based on the Bayes Linear methodologies can be used to implement a large-scale workflow for the measure of hazard probabilities in pyroclastic flows. A triggering event requires the evaluation or reevaluation of hazard at all locations in a jurisdiction, which necessitates running either vast numbers of direct simulations or simulators in ensembles to evaluate all the possibilities in a timely (hours, depending on the need) fashion. We use a multi-level hierarchical construct for the statistical model, and a load balancing master/server utility for allocating the mission-critical workflow tasks on large scale computing platforms for the case of Montserrat island. This methodology is sufficiently general that we anticipate application to a wide array of hazard analyses in response to critical geophysical events.
   IEEE Computer Society  /  ACM     2 0   Y E A R S   -   U N L E A S H I N G   T H E   P O W E R   O F   H P C