Multiscale method for hazard map construction

E. Ramona Stefanescu, Abani Patra, E. Bruce Pitman, Marcus Bursik, Puneet Singla, Tarunraj Singh

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This work describes a multiscale approach for creating a fast surrogate of physics based simulators, to improve the speed of applications that requires large ensembles like hazard map creation. The novel framework is applied in determining the probability of the presence of airborne ash at a specific height when an explosive volcanic eruption occurs. The procedure involves representing both the parameter space (sample points at which the numerical model is evaluated) and physical space (ash concentration at a certain height covered by well delimited parcel) by a weighted graph. The combination of graph representation and low rank approximation gives a good approximation of the original graph (allows us to identify a well-conditioned basis of the adjacency matrix for its numerical range) that is less computationally intensive and more accurate when out-of-sample extension is performed at re-sample points as higher resolution parcels.

Original languageEnglish (US)
Title of host publicationDynamic Data-Driven Environmental Systems Science - 1st International Conference, DyDESS 2014, Revised Selected Papers
EditorsAdrian Sandu, Sai Ravela
PublisherSpringer Verlag
Pages41-53
Number of pages13
ISBN (Print)9783319251370
DOIs
StatePublished - 2015
Event1st International Conference on Dynamic Data-Driven Environmental Systems Science, DyDESS 2014 - Cambridge, United States
Duration: Nov 5 2014Nov 7 2014

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8964
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other1st International Conference on Dynamic Data-Driven Environmental Systems Science, DyDESS 2014
CountryUnited States
CityCambridge
Period11/5/1411/7/14

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint Dive into the research topics of 'Multiscale method for hazard map construction'. Together they form a unique fingerprint.

Cite this