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 - Jan 1 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

Fingerprint

Ashes
Multiscale Methods
Sample point
Hazard
Hazards
Low-rank Approximation
Numerical Range
Graph Representation
Adjacency Matrix
Weighted Graph
Parameter Space
Numerical models
Simulator
Ensemble
High Resolution
Physics
Simulators
Approximation
Graph in graph theory

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Ramona Stefanescu, E., Patra, A., Bruce Pitman, E., Bursik, M., Singla, P., & Singh, T. (2015). Multiscale method for hazard map construction. In A. Sandu, & S. Ravela (Eds.), Dynamic Data-Driven Environmental Systems Science - 1st International Conference, DyDESS 2014, Revised Selected Papers (pp. 41-53). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8964). Springer Verlag. https://doi.org/10.1007/978-3-319-25138-7_5
Ramona Stefanescu, E. ; Patra, Abani ; Bruce Pitman, E. ; Bursik, Marcus ; Singla, Puneet ; Singh, Tarunraj. / Multiscale method for hazard map construction. Dynamic Data-Driven Environmental Systems Science - 1st International Conference, DyDESS 2014, Revised Selected Papers. editor / Adrian Sandu ; Sai Ravela. Springer Verlag, 2015. pp. 41-53 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{bae82f036b1e415a8d2fdceabb93a158,
title = "Multiscale method for hazard map construction",
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.",
author = "{Ramona Stefanescu}, E. and Abani Patra and {Bruce Pitman}, E. and Marcus Bursik and Puneet Singla and Tarunraj Singh",
year = "2015",
month = "1",
day = "1",
doi = "10.1007/978-3-319-25138-7_5",
language = "English (US)",
isbn = "9783319251370",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "41--53",
editor = "Adrian Sandu and Sai Ravela",
booktitle = "Dynamic Data-Driven Environmental Systems Science - 1st International Conference, DyDESS 2014, Revised Selected Papers",
address = "Germany",

}

Ramona Stefanescu, E, Patra, A, Bruce Pitman, E, Bursik, M, Singla, P & Singh, T 2015, Multiscale method for hazard map construction. in A Sandu & S Ravela (eds), Dynamic Data-Driven Environmental Systems Science - 1st International Conference, DyDESS 2014, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 8964, Springer Verlag, pp. 41-53, 1st International Conference on Dynamic Data-Driven Environmental Systems Science, DyDESS 2014, Cambridge, United States, 11/5/14. https://doi.org/10.1007/978-3-319-25138-7_5

Multiscale method for hazard map construction. / Ramona Stefanescu, E.; Patra, Abani; Bruce Pitman, E.; Bursik, Marcus; Singla, Puneet; Singh, Tarunraj.

Dynamic Data-Driven Environmental Systems Science - 1st International Conference, DyDESS 2014, Revised Selected Papers. ed. / Adrian Sandu; Sai Ravela. Springer Verlag, 2015. p. 41-53 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8964).

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

TY - GEN

T1 - Multiscale method for hazard map construction

AU - Ramona Stefanescu, E.

AU - Patra, Abani

AU - Bruce Pitman, E.

AU - Bursik, Marcus

AU - Singla, Puneet

AU - Singh, Tarunraj

PY - 2015/1/1

Y1 - 2015/1/1

N2 - 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.

AB - 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.

UR - http://www.scopus.com/inward/record.url?scp=84951797117&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84951797117&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-25138-7_5

DO - 10.1007/978-3-319-25138-7_5

M3 - Conference contribution

AN - SCOPUS:84951797117

SN - 9783319251370

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 41

EP - 53

BT - Dynamic Data-Driven Environmental Systems Science - 1st International Conference, DyDESS 2014, Revised Selected Papers

A2 - Sandu, Adrian

A2 - Ravela, Sai

PB - Springer Verlag

ER -

Ramona Stefanescu E, Patra A, Bruce Pitman E, Bursik M, Singla P, Singh T. Multiscale method for hazard map construction. In Sandu A, Ravela S, editors, Dynamic Data-Driven Environmental Systems Science - 1st International Conference, DyDESS 2014, Revised Selected Papers. Springer Verlag. 2015. p. 41-53. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-25138-7_5