def mfd_empirical_likelihood_test(cfg: dict, bin_gdf: GeoDataFrame) -> None:
"""
Calculates the (log)likelihood of observing the earthquakes in the seismic
catalog in each :class:`~openquake.hme.utils.bins.SpacemagBin` as the
geometric mean of the Poisson likelihoods of observing the earthquakes
within each :class:`~openquake.hme.utils.bins.MagBin` of the
:class:`~openquake.hme.utils.bins.SpacemagBin`.
The likelihoods are calculated using the empirical likelihood of observing
the number of events that occurred in each
:class:`~openquake.hme.utils.bins.MagBin` given the occurrence rate for
that :class:`~openquake.hme.utils.bins.MagBin`. This is done through a Monte
Carlo simulation, which returns the fraction of the total Monte Carlo
samples had the same number of events as observed.
The likelihoods for each :class:`~openquake.hme.utils.bins.SpacemagBin` are
then log-transformed and appended as a new column to the `bin_gdf`
:class:`GeoDataFrame` hosting the bins.
"""
test_config = cfg["config"]["model_framework"]["gem"]["likelihood"]
source_bin_gdf = get_source_bins(bin_gdf)
logging.info("calculating empirical MFDs for source bins")
if cfg["config"]["parallel"] is False:
source_bin_mfds = source_bin_gdf["SpacemagBin"].apply(
get_stochastic_mfd,
n_iters=test_config["n_iters"],
interval_length=test_config["investigation_time"],
)
else:
source_bin_mfds = get_stochastic_mfds_parallel(
source_bin_gdf["SpacemagBin"],
n_iters=test_config["n_iters"],
interval_length=test_config["investigation_time"],
)
def calc_row_log_like(row, mfd_df=source_bin_mfds):
obs_eqs = row.SpacemagBin.observed_earthquakes
mfd_dict = mfd_df.loc[row._name]
return calc_mfd_log_likelihood_independent(
obs_eqs,
mfd_dict,
not_modeled_val=test_config["not_modeled_val"],
likelihood_method="empirical",
)
logging.info("calculating log likelihoods for sources")
source_bin_log_likes = source_bin_gdf.apply(calc_row_log_like, axis=1)
bin_gdf["log_like"] = test_config["default_likelihood"]
bin_gdf["log_like"].update(source_bin_log_likes)
def mfd_poisson_likelihood_test(cfg: dict, bin_gdf: GeoDataFrame) -> None:
"""
Calculates the (log)likelihood of observing the earthquakes in the seismic
catalog in each :class:`~openquake.hme.utils.bins.SpacemagBin` as the
geometric mean of the Poisson likelihoods of observing the earthquakes
within each :class:`~openquake.hme.utils.bins.MagBin` of the
:class:`~openquake.hme.utils.bins.SpacemagBin`.
The likelihoods are calculated using the Poisson likelihood of observing
the number of events that occurred in each
:class:`~openquake.hme.utils.bins.MagBin` given the occurrence rate for
that :class:`~openquake.hme.utils.bins.MagBin`. See
:func:`~openquake.hme.utils.stats.poisson_likelihood` for more information.
The likelihoods for each :class:`~openquake.hme.utils.bins.SpacemagBin` are
then log-transformed and appended as a new column to the `bin_gdf`
:class:`GeoDataFrame` hosting the bins.
"""
test_config = cfg["config"]["model_framework"]["gem"]["likelihood"]
source_bin_gdf = get_source_bins(bin_gdf)
logging.info("calculating empirical MFDs for source bins")
source_bin_mfds = source_bin_gdf["SpacemagBin"].apply(
lambda x: x.get_rupture_mfd(cumulative=False)
)
def calc_row_log_like(row, mfd_df=source_bin_mfds):
obs_eqs = row.SpacemagBin.observed_earthquakes
mfd_dict = mfd_df.loc[row._name]
return calc_mfd_log_likelihood_independent(
obs_eqs,
mfd_dict,
time_interval=test_config["investigation_time"],
not_modeled_val=test_config["not_modeled_val"],
likelihood_method="poisson",
)
logging.info("calculating log likelihoods for sources")
source_bin_log_likes = source_bin_gdf.apply(calc_row_log_like, axis=1)
bin_gdf["log_like"] = test_config["default_likelihood"]
bin_gdf["log_like"].update(source_bin_log_likes)