Source code for openquake.hazardlib.gsim.gmpe_table

# -*- coding: utf-8 -*-
# vim: tabstop=4 shiftwidth=4 softtabstop=4
#
# Copyright (C) 2015-2022 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 :mod:`openquake.hazardlib.gsim.gmpe_table` defines the
:class:`openquake.hazardlib.gsim.gmpe_table.GMPETable` for defining GMPEs
in the form of binary tables, and
:class:`openquake.hazardlib.gsim.gmpe_table.AmplificationTable` for defining
the corresponding amplification of the IMLs
"""
from copy import deepcopy

import h5py
from scipy.interpolate import interp1d
import numpy

from openquake.baselib.general import CallableDict
from openquake.baselib.python3compat import decode
from openquake.hazardlib.const import TRT, StdDev
from openquake.hazardlib import site
from openquake.hazardlib import imt as imt_module
from openquake.hazardlib.contexts import RuptureContext
from openquake.hazardlib.gsim.base import GMPE
from openquake.baselib.python3compat import round


_get_mean = CallableDict()


@_get_mean.add("base", "nga_east")
def _get_mean_base(kind, distance_type, data, dctx, dists):
    """
    Returns the mean intensity measure level from the tables
    :param data:
        The intensity measure level vector for the given magnitude and IMT
    :param key:
        The distance type
    :param distances:
        The distance vector for the given magnitude and IMT
    """
    dist = getattr(dctx, distance_type)
    # For values outside of the interpolation range use -999. to ensure
    # value is identifiable and outside of potential real values
    interpolator_mean = interp1d(
        dists, data, bounds_error=False, fill_value=-999.)
    mean = interpolator_mean(dist)
    # For those distances less than or equal to the shortest distance
    # extrapolate the shortest distance value
    mean[dist < (dists[0] + 1.0E-3)] = data[0]
    # For those distances significantly greater than the furthest distance
    # set to 1E-20.
    mean[dist > (dists[-1] + 1.0E-3)] = 1E-20
    # If any distance is between the final distance and a margin of 0.001
    # km then assign to smallest distance
    mean[mean < -1.] = data[-1]
    return mean


[docs]def hdf_arrays_to_dict(hdfgroup): """ Convert an hdf5 group contains only data sets to a dictionary of data sets :param hdfgroup: Instance of :class:`h5py.Group` :returns: Dictionary containing each of the datasets within the group arranged by name """ return {key: hdfgroup[key][:] for key in hdfgroup}
[docs]class AmplificationTable(object): """ Class to apply amplification from the GMPE tables. :attr shape: Shape of the amplification arrays as a tuple of (Number Distances, Number IMTs, Number Magnitudes, Number Amplification Levels) :attr periods: Spectral periods defined in table :attr mean: Amplification factors for the mean ground motion :attr sigma: List of modification factors for the standard deviation of ground motion :attr magnitudes: Magnitude values for the tables :attr distances: Distance values for the tables :attr parameter: Parameter to which the amplification applies. There is a check on the parameter name. :attr values: Array of values to which each amplification table corresponds :attr element: Indicates if the amplification corresponds to a rupture attribute or a site attribute """ def __init__(self, amplification_group, magnitudes, distances): """ Setup the amplification factors. :param amplification_group: Amplification model as instance of :class:`h5py.Group` :param magnitudes: Array of magnitudes :param distances: Array of distances """ self.shape = None self.periods = None self.mean = None self.sigma = None self.magnitudes = magnitudes self.distances = distances self.parameter = decode(amplification_group.attrs["apply_to"]) self.values = numpy.array([float(key) for key in amplification_group]) self.argrp_id = numpy.argsort(self.values) self.values = self.values[self.argrp_id] if self.parameter in RuptureContext._slots_: self.element = "Rupture" elif self.parameter in site.site_param_dt: self.element = "Sites" else: raise ValueError("Amplification parameter %s not recognised!" % self.parameter) self._build_data(amplification_group) def _build_data(self, amplification_group): """ Creates the numpy array tables from the hdf5 tables """ # Determine shape of the tables n_levels = len(amplification_group) # Checks the first group in the amplification group and returns the # shape of the SA array - implicitly assumes the SA array in all # amplification groups is the same shape level = next(iter(amplification_group)) n_d, n_p, n_m = amplification_group[level]["IMLs/SA"].shape assert n_d == len(self.distances), (n_d, len(self.distances)) assert n_m == len(self.magnitudes), (n_m, len(self.magnitudes)) # Instantiate the arrays with ones self.mean = {"SA": numpy.ones([n_d, n_p, n_m, n_levels]), "PGA": numpy.ones([n_d, 1, n_m, n_levels]), "PGV": numpy.ones([n_d, 1, n_m, n_levels])} self.sigma = {} for stddev_type in [StdDev.TOTAL, StdDev.INTER_EVENT, StdDev.INTRA_EVENT]: level = next(iter(amplification_group)) if stddev_type in amplification_group[level]: self.sigma[stddev_type] = deepcopy(self.mean) for iloc, (level, amp_model) in enumerate(amplification_group.items()): if "SA" in amp_model["IMLs"]: if iloc == 0: self.periods = amp_model["IMLs/T"][:] else: assert numpy.allclose(self.periods, amp_model["IMLs/T"][:]) for imt in ["SA", "PGA", "PGV"]: if imt in amp_model["IMLs"]: self.mean[imt][:, :, :, self.argrp_id[iloc]] = \ amp_model["IMLs/" + imt][:] for stddev_type in self.sigma: self.sigma[stddev_type][imt][ :, :, :, self.argrp_id[iloc]] = \ amp_model["/".join([stddev_type, imt])][:] self.shape = (n_d, n_p, n_m, n_levels)
[docs] def get_set(self): """ Return the parameter as an instance a Python set """ return {self.parameter}
[docs] def get_amplification_factors(self, imt, ctx, dists, stddev_types): """ Returns the amplification factors for the given rupture and site conditions. :param imt: Intensity measure type as an instance of the :class: `openquake.hazardlib.imt` :param ctx: RuptureContext :param dists: Source to site distances (km) :param stddev_types: List of required standard deviation types :returns: * mean_amp - Amplification factors applied to the median ground motion * sigma_amps - List of modification factors applied to the standard deviations of ground motion """ dist_level_table = self.get_mean_table(imt, ctx) sigma_tables = self.get_sigma_tables(imt, ctx, stddev_types) mean_interpolator = interp1d(self.values, numpy.log10(dist_level_table), axis=1) sigma_interpolators = [interp1d(self.values, sigma_table, axis=1) for sigma_table in sigma_tables] if self.element == "Rupture": mean_amp = 10.0 ** mean_interpolator( getattr(ctx, self.parameter))[0] * numpy.ones_like(dists) sigma_amps = [] for sig_interpolator in sigma_interpolators: sigma_amps.append(sig_interpolator( getattr(ctx, self.parameter))[0] * numpy.ones_like(dists)) else: mean_amp = 10.0 ** mean_interpolator( getattr(ctx, self.parameter))[0, :] sigma_amps = [] for sig_interpolator in sigma_interpolators: sigma_amps.append(sig_interpolator( getattr(ctx, self.parameter))[0, :] * numpy.ones_like(dists)) return mean_amp, sigma_amps
[docs] def get_mean_table(self, imt, rctx): """ Returns amplification factors for the mean, given the rupture and intensity measure type. :returns: amplification table as an array of [Number Distances, Number Levels] """ # Levels by Distances if imt.string in 'PGA PGV': interpolator = interp1d(self.magnitudes, numpy.log10(self.mean[imt.string]), axis=2) output_table = 10.0 ** ( interpolator(rctx.mag).reshape(self.shape[0], self.shape[3])) else: # For spectral accelerations - need two step process # Interpolate period - log-log space interpolator = interp1d(numpy.log10(self.periods), numpy.log10(self.mean["SA"]), axis=1) period_table = interpolator(numpy.log10(imt.period)) # Interpolate magnitude - linear-log space mag_interpolator = interp1d(self.magnitudes, period_table, axis=1) output_table = 10.0 ** mag_interpolator(rctx.mag) return output_table
[docs] def get_sigma_tables(self, imt, rctx, stddev_types): """ Returns modification factors for the standard deviations, given the rupture and intensity measure type. :returns: List of standard deviation modification tables, each as an array of [Number Distances, Number Levels] """ output_tables = [] for stddev_type in stddev_types: # For PGA and PGV only needs to apply magnitude interpolation if imt.string in 'PGA PGV': interpolator = interp1d(self.magnitudes, self.sigma[stddev_type][imt.string], axis=2) output_tables.append( interpolator(rctx.mag).reshape(self.shape[0], self.shape[3])) else: # For spectral accelerations - need two step process # Interpolate period interpolator = interp1d(numpy.log10(self.periods), self.sigma[stddev_type]["SA"], axis=1) period_table = interpolator(numpy.log10(imt.period)) mag_interpolator = interp1d(self.magnitudes, period_table, axis=1) output_tables.append(mag_interpolator(rctx.mag)) return output_tables
def _setup_amplification(self, fle): """ If amplification data is specified then reads into memory and updates the required rupture and site parameters """ self.amplification = AmplificationTable( fle["Amplification"], self.m_w, self.distances) if self.amplification.element == "Sites": self.REQUIRES_SITES_PARAMETERS = {self.amplification.parameter} elif self.amplification.element == "Rupture": # set the site and rupture parameters on the instance self.REQUIRES_SITES_PARAMETERS = set() self.REQUIRES_RUPTURE_PARAMETERS = ( self.REQUIRES_RUPTURE_PARAMETERS | {self.amplification.parameter}) def _return_tables(self, mag, imt, std_idx): """ Returns the vector of ground motions or standard deviations corresponding to the specific magnitude and intensity measure type. :param std_idx: the string "IMLs" or an integer 0, 1, 2 for total, inter, intra """ if imt.string in 'PGA PGV': # Get scalar imt if std_idx == "IMLs": iml_table = self.imls[imt.string][:] else: stds = self.stddevs[std_idx] if stds is None: raise KeyError("Unsupported StdDev#%d" % std_idx) iml_table = stds[imt.string][:] n_d, n_s, n_m = iml_table.shape iml_table = iml_table.reshape([n_d, n_m]) else: if std_idx == "IMLs": periods = self.imls["T"][:] iml_table = self.imls["SA"][:] else: stds = self.stddevs[std_idx] if stds is None: raise KeyError("Unsupported StdDev#%d" % std_idx) periods = stds["T"][:] iml_table = stds["SA"][:] low_period = round(periods[0], 7) high_period = round(periods[-1], 7) if (round(imt.period, 7) < low_period) or ( round(imt.period, 7) > high_period): raise ValueError("Spectral period %.3f outside of valid range " "(%.3f to %.3f)" % (imt.period, periods[0], periods[-1])) # Apply log-log interpolation for spectral period interpolator = interp1d(numpy.log10(periods), numpy.log10(iml_table), axis=1) iml_table = 10. ** interpolator(numpy.log10(imt.period)) return apply_magnitude_interpolation(self, mag, iml_table)
[docs]def apply_magnitude_interpolation(self, mag, iml_table): """ Interpolates the tables to the required magnitude level :param float mag: Magnitude :param iml_table: Intensity measure level table """ # do not allow "mag" to exceed maximum table magnitude mag = numpy.clip(mag, None, self.m_w[-1]) # Get magnitude values if (mag < self.m_w[0]).any() or (mag > self.m_w[-1]).any(): raise ValueError("Magnitude %s outside of supported range " "(%.2f to %.2f)" % (mag, self.m_w[0], self.m_w[-1])) # It is assumed that log10 of the spectral acceleration scales # linearly (or approximately linearly) with magnitude m_interpolator = interp1d(self.m_w, numpy.log10(iml_table), axis=1) return 10.0 ** m_interpolator(mag)
def _get_stddevs(self, dists, ctx, imt, stdis): """ Returns the total standard deviation of the intensity measure level from the tables. """ stddevs = [] dst = getattr(ctx, self.distance_type) for stdi in stdis: sigma = _return_tables(self, ctx.mag, imt, stdi) stddev = numpy.interp(dst, dists, sigma) stddev[dst < dists[0]] = sigma[0] stddev[dst > dists[-1]] = sigma[-1] stddevs.append(stddev) return stddevs
[docs]class GMPETable(GMPE): """ Implements ground motion prediction equations in the form of a table from which the expected ground motion intensity levels and standard deviations are interpolated. In a GMPE tables the expected ground motions for each of the IMTs over the range of magnitudes and distances are stored in an hdf5 file on the path specified by the user. In this version of the GMPE the expected values are interpolated to the required IMT, magnitude and distance in three stages. i) Initially the correct IMT values are identified, interpolating in log-T|log-IML space between neighbouring spectral periods. ii) The IML values are then interpolated to the correct magnitude using linear-M|log-IML space iii) The IML values are then interpolated to the correct distance via linear-D|linear-IML interpolation """ DEFINED_FOR_TECTONIC_REGION_TYPE = TRT.ACTIVE_SHALLOW_CRUST DEFINED_FOR_INTENSITY_MEASURE_TYPES = set() DEFINED_FOR_INTENSITY_MEASURE_COMPONENT = "" DEFINED_FOR_STANDARD_DEVIATION_TYPES = {StdDev.TOTAL} REQUIRES_SITES_PARAMETERS = set() REQUIRES_DISTANCES = set() REQUIRES_RUPTURE_PARAMETERS = {"mag"} gmpe_table = None # see subclasses like NBCC2015_AA13_activecrustFRjb_low amplification = None kind = "base" def __init__(self, **kwargs): """ Executes the preprocessing steps at the instantiation stage to read in the tables from hdf5 and hold them in memory. """ super().__init__(**kwargs) fname = self.kwargs.get('gmpe_table', self.gmpe_table) with h5py.File(fname, "r") as fle: self.distance_type = decode(fle["Distances"].attrs["metric"]) self.REQUIRES_DISTANCES = set([self.distance_type]) # Load in magnitude self.m_w = fle["Mw"][:] # Load in distances self.distances = fle["Distances"][:] # Load intensity measure types and levels self.imls = hdf_arrays_to_dict(fle["IMLs"]) # Update the list of supported IMTs from the tables imt_list = [] # this is a list of factories, like PGA, SA, etc for key in self.imls: if "SA" in key: imt_list.append(imt_module.SA) elif key == "T": continue else: try: factory = getattr(imt_module, key) except Exception: continue imt_list.append(factory) self.DEFINED_FOR_INTENSITY_MEASURE_TYPES = set(imt_list) if "SA" in self.imls and "T" not in self.imls: raise ValueError("Spectral Acceleration must be accompanied by" " periods") # Load in standard deviations self.stddevs = [None] * 3 if self.kind in "nga_east usgs": # there are no stddevs in the hdf5 file return self.stddevs[0] = hdf_arrays_to_dict(fle["Total"]) self.DEFINED_FOR_STANDARD_DEVIATION_TYPES = set( self.DEFINED_FOR_STANDARD_DEVIATION_TYPES) for stddev_type in [StdDev.INTER_EVENT, StdDev.INTRA_EVENT]: if stddev_type in fle: self.stddevs[StdDev.idx[stddev_type]] = hdf_arrays_to_dict( fle[stddev_type]) self.DEFINED_FOR_STANDARD_DEVIATION_TYPES.add(stddev_type) if "Amplification" in fle: _setup_amplification(self, fle)
[docs] def compute(self, ctx, imts, mean, sig, tau, phi): stds = [sig, tau, phi] stdis = [StdDev.idx[sdt] for sdt in self.DEFINED_FOR_STANDARD_DEVIATION_TYPES] for m, imt in enumerate(imts): # Return Distance Tables imls = _return_tables(self, ctx.mag, imt, "IMLs") # Get distance vector for the given magnitude idx = numpy.searchsorted(self.m_w, ctx.mag) dists = self.distances[:, 0, idx - 1] # Get mean and standard deviations mean_ = _get_mean(self.kind, self.distance_type, imls, ctx, dists) stddevs = _get_stddevs(self, dists, ctx, imt, stdis) if self.amplification: # Apply amplification mean_amp, sigma_amp = ( self.amplification.get_amplification_factors( imt, ctx, getattr(ctx, self.distance_type), self.DEFINED_FOR_STANDARD_DEVIATION_TYPES)) mean[m] = numpy.log(mean_) + numpy.log(mean_amp) for i, amp in zip(stdis, sigma_amp): stds[i][m] = stddevs[i] * amp else: mean[m] = numpy.log(mean_) for i in stdis: stds[i][m] = stddevs[i]