#
# 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/>.
"""
Implements the GMPE adjustment used for the Cauzzi et al. (2014) and
Derras et al. (2014) model, as used in the 2018 German National Seismic Hazard
Model of Gruenthal et al. (2018)
"Gruenthal, G., Strometer, D., Bosse, C., Cotton, F. and Bindi, D. (2018)
The probabilistic seismic hazard assessment of Germany - version 2016,
considering the range or epistemic uncertainties and aleatory variability."
Bulletin of Earthquake Engineering, 16(10), 4339-4395
Module Exports :class: `CauzziEtAl2014RhypoGermany`,
                       `DerrasEtAl2014RhypoGermany`
"""
import numpy as np
from scipy.constants import g
from openquake.hazardlib.imt import PGA, SA
from openquake.hazardlib.gsim.cauzzi_2014 import CauzziEtAl2014
from openquake.hazardlib.gsim.derras_2014 import DerrasEtAl2014
[docs]def rhypo_to_rrup(rhypo, mag):
    """
    Converts hypocentral distance to an equivalent rupture distance
    dependent on the magnitude
    """
    rrup = rhypo - (0.7108 + 2.496E-6 * (mag ** 7.982))
    rrup[rrup < 3.0] = 3.0
    return rrup 
[docs]def rhypo_to_rjb(rhypo, mag):
    """
    Converts hypocentral distance to an equivalent Joyner-Boore distance
    dependent on the magnitude
    """
    epsilon = rhypo - (4.853 + 1.347E-6 * (mag ** 8.163))
    rjb = np.zeros_like(rhypo)
    idx = epsilon >= 3.
    rjb[idx] = np.sqrt((epsilon[idx] ** 2.) - 9.0)
    rjb[rjb < 0.0] = 0.0
    return rjb 
# Cauzzi et al. 2014 - Converted from Rhypo
[docs]class CauzziEtAl2014RhypoGermany(CauzziEtAl2014):
    """
    Implements the Cauzzi et al. (2015) GMPE applying the rhypo to rrup
    adjustment factor adopted for Germany
    """
    REQUIRES_DISTANCES = {"rhypo", "rrup"}
    REQUIRES_RUPTURE_PARAMETERS = {"rake", "mag", "width"}
    def __init__(self, adjustment_factor=1.0, **kwargs):
        super().__init__(adjustment_factor=adjustment_factor, **kwargs)
        self.adjustment_factor = np.log(adjustment_factor)
    def _compute_mean(self, C, rup, dists, sites, imt):
        """
        Returns the mean ground motion acceleration and velocity
        """
        if rup.width > 1.0E-3:
            # Finite rupture source used
            rrup = np.copy(dists.rrup)
        else:
            # Point source MSR used - convert rhypo to rrup
            rrup = rhypo_to_rrup(dists.rhypo, rup.mag)
        mean = (self._get_magnitude_scaling_term(C, rup.mag) +
                self._get_distance_scaling_term(C, rup.mag, rrup) +
                self._get_style_of_faulting_term(C, rup.rake) +
                self._get_site_amplification_term(C, sites.vs30))
        # convert from cm/s**2 to g for SA and from cm/s**2 to g for PGA (PGV
        # is already in cm/s) and also convert from base 10 to base e.
        if isinstance(imt, PGA):
            mean = np.log((10 ** mean) * ((2 * np.pi / 0.01) ** 2) *
                          1e-2 / g)
        elif isinstance(imt, SA):
            mean = np.log((10 ** mean) * ((2 * np.pi / imt.period) ** 2) *
                          1e-2 / g)
        else:
            mean = np.log(10 ** mean)
        return mean + self.adjustment_factor 
# Derras et al 2014
[docs]class DerrasEtAl2014RhypoGermany(DerrasEtAl2014):
    """
    Re-calibration of the Derras et al. (2014) GMPE taking hypocentral
    distance as an input and converting to Rjb
    """
    #: The required distance parameter is hypocentral distance
    REQUIRES_DISTANCES = {'rjb', 'rhypo'}
    REQUIRES_RUPTURE_PARAMETERS = {"rake", "mag", "hypo_depth", "width"}
    def __init__(self, adjustment_factor=1.0, **kwargs):
        super().__init__(adjustment_factor=adjustment_factor, **kwargs)
        self.adjustment_factor = np.log(adjustment_factor)
[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]
        # Get the mean
        mean = self.get_mean(C, rup, sites, dists)
        if imt.name == "PGV":
            # Convert from log10 m/s to ln cm/s
            mean = np.log((10.0 ** mean) * 100.)
        else:
            # convert from log10 m/s/s to ln g
            mean = np.log((10.0 ** mean) / g)
        # Get the standard deviations
        stddevs = self.get_stddevs(C, mean.shape, stddev_types)
        return mean + self.adjustment_factor, stddevs 
[docs]    def get_mean(self, C, rup, sites, dists):
        """
        Returns the mean ground motion in terms of log10 m/s/s, implementing
        equation 2 (page 502)
        """
        # W2 needs to be a 1 by 5 matrix (not a vector
        w_2 = np.array([
            [C["W_21"], C["W_22"], C["W_23"], C["W_24"], C["W_25"]]
            ])
        # Gets the style of faulting dummy variable
        sof = self._get_sof_dummy_variable(rup.rake)
        # Get the normalised coefficients
        p_n = self.get_pn(rup, sites, dists, sof)
        mean = np.zeros_like(dists.rhypo)
        # Need to loop over sites - maybe this can be improved in future?
        # ndenumerate is used to allow for application to 2-D arrays
        for idx, rval in np.ndenumerate(p_n[0]):
            # Place normalised coefficients into a single array
            p_n_i = np.array([rval, p_n[1], p_n[2][idx], p_n[3], p_n[4]])
            # Executes the main ANN model
            mean_i = np.dot(w_2, np.tanh(np.dot(self.W_1, p_n_i) + self.B_1))
            mean[idx] = (0.5 * (mean_i + C["B_2"] + 1.0) *
                         (C["tmax"] - C["tmin"])) + C["tmin"]
        return mean 
[docs]    def get_pn(self, rup, sites, dists, sof):
        """
        Normalise the input parameters within their upper and lower
        defined range
        """
        # List must be in following order
        p_n = []
        # Rjb
        # Note that Rjb must be clipped at 0.1 km
        if rup.width > 1.0E-3:
            rjb = np.copy(dists.rjb)
        else:
            rjb = rhypo_to_rjb(dists.rhypo, rup.mag)
        rjb[rjb < 0.1] = 0.1
        p_n.append(self._get_normalised_term(np.log10(rjb),
                                             self.CONSTANTS["logMaxR"],
                                             self.CONSTANTS["logMinR"]))
        # Magnitude
        p_n.append(self._get_normalised_term(rup.mag,
                                             self.CONSTANTS["maxMw"],
                                             self.CONSTANTS["minMw"]))
        # Vs30
        p_n.append(self._get_normalised_term(np.log10(sites.vs30),
                                             self.CONSTANTS["logMaxVs30"],
                                             self.CONSTANTS["logMinVs30"]))
        # Depth
        p_n.append(self._get_normalised_term(rup.hypo_depth,
                                             self.CONSTANTS["maxD"],
                                             self.CONSTANTS["minD"]))
        # Style of Faulting
        p_n.append(self._get_normalised_term(sof,
                                             self.CONSTANTS["maxFM"],
                                             self.CONSTANTS["minFM"]))
        return p_n