Transcript A Simulation Approach for Better Understanding of Natural

Yelena Kropivnitskaya, Kristy F. Tiampo,
Jinhui Qin and Michael A. Bauer
A Simulation Approach for
Better Understanding of
Natural Catastrophe Hazard
and Risk
Motivation
Deep uncertainty concerning the future evolution of disaster losses
Socio-economic damages caused by extreme natural events, have risen steadily for decades
(Munich Re,2015);
Damage drivers:
– economic growth;
– changes in the
localization of assets
and population (Pielke
and Landsea, 1998);
– after 2004 and 2005
hurricane seasons in the
North Atlantic - climate
change (Emanuel,
2005a, 2005b; Webster
et al., 2005).
– The design of long-lived infrastructure much more difficult (inadequate design and illadaptation to climate conditions);
– Can provoke significant losses (Hallegatte, 2006a).
Area of Interest –
Eastern Canada (Montreal)
Earthquake Catalogue (Fereidoni et al.,2011)
Figure credit – (Atkinson, 2011)
Population Density
Figure credit – (NRC, 2001)
Eastern Canada seismic
zone model (GSC)
Thirty-nine zones (Adams an Halchuk, 2003, 2004; Halchuk and Adams, 2008)
with a magnitude-recurrence parameters, maximum magnitudes and depths
Consists of two alternative source zone models:
the H model - historical seismicity,
the R model - regional tectonic structure.
Eastern Canada GMPE Set
(Atkinson and Adams, 2013)
New way to quantify epistemic uncertainty:
Using existing peer-reviewed GMPEs to construct a set of three weighted equations
(medium (M), high (H) and lower(L)) for each region and event type
Frequency = 10 Hz
H
GMPE
M
M7.5
GM
Log(PSA), cm/sec^2
L
M6.5
M5.5
M4.5
Epicentral distance, km
Montreal Soil Map (Rosset et al. 2014)
The Montreal soil map is used to find
amplifications for:
• PGV,
• PGA,
• response spectra at frequencies of
1, 5, and 10 Hz.
Amplification values are calculated
based on Boore and Atkinson (2008)
relations:
• a linear part (Flin) which is
dependent only for Vs30;
• a non-linear term (Fnl) which is a
factor of both Vs30 and a reference
ground motion.
In this case, the reference motion is
the predicted peak ground acceleration
in g for Vref = 760 m/s. The
combination of these two terms is the
total amplification:
Fs = Flin + Fnl
Originally:
EqHaz1 - The
number of records
in each synthetic
catalog is limited to
1,000,000
EqHaz2 - Could
have no more than
100,000 points
Parallel execution
• Each single operator executes on a single core
• The entire process is divided into 3 stages:
- The first stage is exactly the same as before
- Monte Carlo simulations for the synthetic catalogue
- Spatial grid decomposition
The number of
records in the
each synthetic
catalog is now
limited to
10,000,000
Output
Example - Mean hazard map for a 2475 year
return period for pseudo acceleration at T=0.1
sec
Resolution of mean hazard map improved to 2,500,000
points
The number of periods is the specified input parameter
(in our case 0.2 sec, 1.0 sec and 2.0 sec)
2475 Year Stochastic Subcatalogs for
Eastern Canada
Montreal, located in the seismically active region of the western Quebec (WQU) seismic
zone, is one the largest and most populated cities in eastern Canada.
Figure demonstrates two random runs of seismicity in the region based on seismic zones
model (Halchuk and Adams, 2008).
Contributions of different magnitude-distance bins to PGA hazard
in Montreal from a 10 million year synthetic catalog
The maximum contribution is
observed at a distance of
approximately
50
km
for
magnitudes between 6 and 7.5.
PGA
There also are two peaks at 0.1 Hz
frequency which are close to 50 km
and 100 km for the same
magnitudes.
At 1 Hz frequency, the maximum
contribution also occurs at around
50 km.
The distance range of maximum
impact seems to be wider however,
starting at about 30 km and going
up to approximately 70 km.
Scenarios from a 10 Million Year Synthetic Catalog
for the Events with the Highest Contribution to
Hazard
In this regions single seismic events dominate the hazard and lead to a high level of
uncertainty of the input parameters in earthquake scenarios. The results of this work
which are earthquake scenarios from a 10 million year synthetic catalog are an
important input for the future uncertainty and PSHA model parameters analysis for
Montreal city.
Scenario 1
M 7.5, depth 10km
epi ~40 km (WQU)
Scenario 2
M 7.5, depth 10km
epi ~340 km (CHV)
NE Atlantic Hurricane
• Extreme hurricanes occur with low
frequency.
• Low probability of North Atlantic
Ocean hurricane landfall makes
statistics are difficult to estimate.
• Statistical methods can be
employed to resample the historical
data, creating a large number of
tracks used to improve estimates of
the probability extremes.
• A spatial domain 39-48N and 6085W
Historical Data Sources and Functional
Form of Parameters Distribution
• The historical best-track North Atlantic hurricane
dataset (HURDAT2; Landsea et al. 2015) from
1851 to 2015:
–
–
–
–
–
genesis
translational speed and heading
central pressure over water
inland filling rate
maximum wind velocity
Poisson
Gaussian
Gaussian
Gaussian
Lognormal
• The extended best-track dataset (Demuth et al.
2006):
– the radius of maximum wind
Lognormal
(truncated)
Thank you