Probabilistic Tsunami Hazard Analysis: High Performance Computing for Massive Scale Monte Carlo type Inundation Simulations Steven J. Gibbons 1 , Manuel J. Castro Díaz 3 , Sylfest Glimsdal 1 , Carl Bonnevie Harbitz 1 , Maria Concetta Lorenzino 2 , Stefano Lorito 2 , Finn Løvholt 1 , Massimo Nazaria 2 , Fabrizio Romano 2 , Jorge Macías Sánchez 3 , Jacopo Selva 2 , Roberto Tonini 2 , José Manuel González Vida 3 , Manuela Volpe 2 , and Malte Vöge 1 NGI, Norway 1 INGV, Italy 2 University of Málaga, Spain 3 #shareEGU20
16
Embed
Probabilistic Tsunami Hazard Assessment · 2020. 5. 5. · Probabilistic Tsunami Hazard Analysis: High Performance Computing for Massive Scale Monte Carlo type Inundation Simulations
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Probabilistic Tsunami Hazard Analysis: High Performance Computing for Massive Scale Monte Carlo type Inundation Simulations
Steven J. Gibbons1, Manuel J. Castro Díaz3, Sylfest Glimsdal1, Carl Bonnevie Harbitz1, Maria Concetta Lorenzino2, Stefano Lorito2, Finn Løvholt1, Massimo Nazaria2, Fabrizio Romano2, Jorge Macías Sánchez3, Jacopo Selva2, Roberto Tonini2, José Manuel González Vida3, Manuela Volpe2, andMalte Vöge1
NGI, Norway1
INGV, Italy2
University of Málaga, Spain3
#shareEGU20
Probabilistic Tsunami Hazard Assessment
PTHA gives: Probability and uncertainty of a
tsunami exceeding a given intensity at any point
on the coastline within a given time interval.
0
Pro
bab
ility
Tsunami Intensity
Lower
Uncertainty
Higher
Uncertainty
Suite of hazard curves
expressing uncertainty in
hazard model …
Probabilistic Tsunami Hazard Assessment
0
Pro
bab
ility
Tsunami Intensity
Suite of hazard curves
expressing uncertainty in
hazard model …
Disaster Risk Reduction applications:
PTHA based Coastal Planning (evacuation maps)
Facilitate scientific results as input to political decisions for different design levels
• 2500 yr Annual Return Period
• 84th percentile
i.e.
~ 2% in 50 yr
MultiBeam EchoSounder (MBES)
Based on the bathymetric gradient, the study area can be divided
in two sectors:
Sector 1: -2 m up to -20 m Sector 2: -20 m up to -50m
Digital Terrain Model with a resolution of 0.50 m x 0.50 m
◉ Prioritization for accelerating science and standards:○ Model calibration and benchmarking○ Acquisition of new data/enhanced real-
time monitoring○ Higher Resolution assessments
Conclusion - HPC can raise the bar on high resolution local tsunami hazard capabilities
◉ Inundation model computationally more demanding than offshore simulations○ Typical simulation time 1h on single GPU○ Higher resolution runs more demanding
◉ Catania example: ○ For retrieving 99 % of the hazard
~250 000 scenarios
◉ On Tier-0 HPC (Piz Daint)○ 2-3 days computation time for
resolutions used in present problems○ Testing of higher resolution is