# -*- coding: utf-8 -*-
# vim: tabstop=4 shiftwidth=4 softtabstop=4
#
# Copyright (C) 2013-2018 GEM Foundation
#
# OpenQuake is free software: you can redistribute it and/or modify it
# under the terms of the GNU Affero General Public License as published
# by the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# OpenQuake is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU Affero General Public License for more details.
#
# You should have received a copy of the GNU Affero General Public License
# along with OpenQuake. If not, see <http://www.gnu.org/licenses/>.
"""
Module exports :class:`AfshariStewart2016`,
:class:`AfshariStewart2016Japan`
"""
from __future__ import division
import numpy as np
from openquake.hazardlib.gsim.base import CoeffsTable, GMPE
from openquake.hazardlib import const
from openquake.hazardlib.imt import RSD595, RSD575, RSD2080
[docs]class AfshariStewart2016(GMPE):
"""
Implements the GMPE of Afshari & Stewart (2016) for relative significant
duration for 5 - 75 %, 5 - 95 % and 20 - 80 % Arias Intensity.
Afshari, K. and Stewart, J. P. (2016) "Physically Parameterized Prediction
Equations for Signficant Duration in Active Crustal Regions", Earthquake
Spectra, 32(4), 2057 - 2081
"""
#: Supported tectonic region type is active shallow crust
DEFINED_FOR_TECTONIC_REGION_TYPE = const.TRT.ACTIVE_SHALLOW_CRUST
#: Supported intensity measure types are 5 - 95 % Arias and 5 - 75 % Arias
#: significant duration
DEFINED_FOR_INTENSITY_MEASURE_TYPES = set([
RSD595,
RSD575,
RSD2080
])
#: Supported intensity measure component is the geometric mean horizontal
#: component
DEFINED_FOR_INTENSITY_MEASURE_COMPONENT = const.IMC.AVERAGE_HORIZONTAL
#: Supported standard deviation type is only total, see table 7, page 35
DEFINED_FOR_STANDARD_DEVIATION_TYPES = set([
const.StdDev.TOTAL,
const.StdDev.INTER_EVENT,
const.StdDev.INTRA_EVENT
])
#: Requires vs30
REQUIRES_SITES_PARAMETERS = set(('vs30', 'z1pt0'))
#: Required rupture parameters are magnitude and top of rupture depth
REQUIRES_RUPTURE_PARAMETERS = set(('mag', 'rake'))
#: Required distance measure is closest distance to rupture
REQUIRES_DISTANCES = set(('rrup', ))
[docs] def get_mean_and_stddevs(self, sites, rup, dists, imt, stddev_types):
"""
See :meth:`superclass method
<.base.GroundShakingIntensityModel.get_mean_and_stddevs>`
for spec of input and result values.
"""
C = self.COEFFS[imt]
mean = (np.log(self.get_magnitude_term(C, rup) +
self.get_distance_term(C, dists.rrup)) +
self.get_site_amplification(C, sites))
stddevs = self.get_stddevs(C, sites.vs30.shape, rup.mag, stddev_types)
return mean, stddevs
[docs] def get_magnitude_term(self, C, rup):
"""
Returns the magnitude scaling term in equation 3
"""
b0, stress_drop = self._get_sof_terms(C, rup.rake)
if rup.mag <= C["m1"]:
return b0
else:
# Calculate moment (equation 5)
m_0 = 10.0 ** (1.5 * rup.mag + 16.05)
# Get stress-drop scaling (equation 6)
if rup.mag > C["m2"]:
stress_drop += (C["b2"] * (C["m2"] - self.CONSTANTS["mstar"]) +
(C["b3"] * (rup.mag - C["m2"])))
else:
stress_drop += (C["b2"] * (rup.mag - self.CONSTANTS["mstar"]))
stress_drop = np.exp(stress_drop)
# Get corner frequency (equation 4)
f0 = 4.9 * 1.0E6 * 3.2 * ((stress_drop / m_0) ** (1. / 3.))
return 1. / f0
[docs] def get_distance_term(self, C, rrup):
"""
Returns the distance scaling term in equation 7
"""
f_p = C["c1"] * rrup
idx = np.logical_and(rrup > self.CONSTANTS["r1"],
rrup <= self.CONSTANTS["r2"])
f_p[idx] = (C["c1"] * self.CONSTANTS["r1"]) +\
C["c2"] * (rrup[idx] - self.CONSTANTS["r1"])
idx = rrup > self.CONSTANTS["r2"]
f_p[idx] = C["c1"] * self.CONSTANTS["r1"] +\
C["c2"] * (self.CONSTANTS["r2"] - self.CONSTANTS["r1"]) +\
C["c3"] * (rrup[idx] - self.CONSTANTS["r2"])
return f_p
def _get_sof_terms(self, C, rake):
"""
Returns the style-of-faulting scaling parameters
"""
if rake >= 45.0 and rake <= 135.0:
# Reverse faulting
return C["b0R"], C["b1R"]
elif rake <= -45. and rake >= -135.0:
# Normal faulting
return C["b0N"], C["b1N"]
else:
# Strike slip
return C["b0SS"], C["b1SS"]
def _get_lnmu_z1(self, vs30):
"""
Returns the z1.0 normalisation term for California (equation 11)
"""
return (-7.15 / 4.) * np.log(
(vs30 ** 4. + 570.94 ** 4) / (1360.0 ** 4. + 570.94 ** 4.)) -\
np.log(1000.0)
[docs] def get_site_amplification(self, C, sites):
"""
Returns the site amplification term
"""
# Gets delta normalised z1
dz1 = sites.z1pt0 - np.exp(self._get_lnmu_z1(sites.vs30))
f_s = C["c5"] * dz1
# Calculates site amplification term
f_s[dz1 > self.CONSTANTS["dz1ref"]] = (C["c5"] *
self.CONSTANTS["dz1ref"])
idx = sites.vs30 > self.CONSTANTS["v1"]
f_s[idx] += (C["c4"] * np.log(self.CONSTANTS["v1"] / C["vref"]))
idx = np.logical_not(idx)
f_s[idx] += (C["c4"] * np.log(sites.vs30[idx] / C["vref"]))
return f_s
[docs] def get_stddevs(self, C, nsites, mag, stddev_types):
"""
Returns the standard deviations
"""
tau = self._get_tau(C, mag) + np.zeros(nsites)
phi = self._get_phi(C, mag) + np.zeros(nsites)
stddevs = []
for stddev in stddev_types:
assert stddev in self.DEFINED_FOR_STANDARD_DEVIATION_TYPES
if stddev == const.StdDev.TOTAL:
stddevs.append(np.sqrt(tau ** 2. + phi ** 2.))
elif stddev == const.StdDev.INTER_EVENT:
stddevs.append(tau)
elif stddev == const.StdDev.INTRA_EVENT:
stddevs.append(phi)
return stddevs
def _get_tau(self, C, mag):
"""
Returns magnitude dependent inter-event standard deviation (tau)
(equation 14)
"""
if mag < 6.5:
return C["tau1"]
elif mag < 7.:
return C["tau1"] + (C["tau2"] - C["tau1"]) * ((mag - 6.5) / 0.5)
else:
return C["tau2"]
def _get_phi(self, C, mag):
"""
Returns the magnitude dependent intra-event standard deviation (phi)
(equation 15)
"""
if mag < 5.5:
return C["phi1"]
elif mag < 5.75:
return C["phi1"] + (C["phi2"] - C["phi1"]) * ((mag - 5.5) / 0.25)
else:
return C["phi2"]
COEFFS = CoeffsTable(sa_damping=5, table="""\
imt m1 m2 b0N b0R b0SS b0U b1N b1R b1SS b1U b2 b3 c1 c2 c3 c4 c5 vref tau1 tau2 phi1 phi2
rsd575 5.35 7.15 1.555 0.7806 1.2790 1.280 4.992 7.061 5.578 5.576 0.9011 -1.684 0.1159 0.1065 0.0682 -0.2246 0.0006 368.2 0.28 0.25 0.54 0.41
rsd595 5.2 7.40 2.541 1.6120 2.3020 2.182 3.170 4.536 3.467 3.628 0.9443 -3.911 0.3165 0.2539 0.0932 -0.3183 0.0006 369.9 0.25 0.19 0.43 0.35
rsd2080 5.2 7.40 1.409 0.7729 0.8804 0.8822 4.778 6.579 6.188 6.182 0.7414 -3.164 0.0646 0.0865 0.0373 -0.4237 0.0005 369.6 0.30 0.19 0.56 0.45
""")
CONSTANTS = {"mstar": 6.0,
"r1": 10.0,
"r2": 50.0,
"v1": 600.0,
"dz1ref": 200.0}
[docs]class AfshariStewart2016Japan(AfshariStewart2016):
"""
Adaption of the Afshari & Stewart (2016) GMPE for relative significant
duration for the case when the Japan basin model is preferred
"""
def _get_lnmu_z1(self, vs30):
"""
Returns the z1.0 normalisation term for Japan (equation 12)
"""
return (-5.23 / 2.) * np.log(
(vs30 ** 2. + 412.39 ** 2) / (1360.0 ** 2. + 412.39 ** 2.)) -\
np.log(1000.0)