def test_spectrum():
freq = np.logspace(-2, 2, 101)
moment = spectrum.moment_from_magnitude(6.7)
mod = spectrum.model((moment, 150), freq, 10, kappa=0.035)
np.testing.assert_allclose(mod[0], 0.21764373, atol=1e-5)
np.testing.assert_allclose(mod[50], 113.5025146, atol=1e-5)
np.testing.assert_allclose(mod[-1], 0.0032295, atol=1e-5)
def summary_plots(st, directory, origin):
"""Stream summary plot.
Args:
st (gmprocess.stationtrace.StationStream):
Stream of data.
directory (str):
Directory for saving plots.
origin (ScalarEvent):
Flattened subclass of Obspy Event.
"""
mpl.rcParams['font.size'] = 8
# Check if directory exists, and if not, create it.
if not os.path.exists(directory):
os.makedirs(directory)
# Setup figure for stream
nrows = 4
ntrace = min(len(st), 3)
fig = plt.figure(figsize=(3.9 * ntrace, 10))
gs = fig.add_gridspec(nrows, ntrace, height_ratios=[1, 1, 2, 2])
ax = [plt.subplot(g) for g in gs]
stream_id = st.get_id()
logging.debug('stream_id: %s' % stream_id)
logging.debug('passed: %s' % st.passed)
if st.passed:
plt.suptitle("M%s %s | %s (passed)" %
(origin.magnitude, origin.id, stream_id),
x=0.5,
y=1.02)
else:
plt.suptitle("M%s %s | %s (failed)" %
(origin.magnitude, origin.id, stream_id),
color='red',
x=0.5,
y=1.02)
# Compute velocity
st_vel = st.copy()
st_vel = st_vel.integrate()
# process channels in preferred sort order (i.e., HN1, HN2, HNZ)
channels = [tr.stats.channel for tr in st]
if len(channels) < 3:
channelidx = np.argsort(channels).tolist()
else:
channelidx = range(3)
for j in channelidx:
tr = st[channelidx.index(j)]
# Break if j>3 becasue we can't on a page.
if j > 2:
logging.warning('Only plotting first 3 traces in stream.')
break
# ---------------------------------------------------------------------
# Get trace info
if tr.hasCached('snr'):
snr_dict = tr.getCached('snr')
else:
snr_dict = None
if tr.hasCached('signal_spectrum'):
signal_dict = tr.getCached('signal_spectrum')
else:
signal_dict = None
if tr.hasCached('noise_spectrum'):
noise_dict = tr.getCached('noise_spectrum')
else:
noise_dict = None
if tr.hasCached('smooth_signal_spectrum'):
smooth_signal_dict = tr.getCached('smooth_signal_spectrum')
else:
smooth_signal_dict = None
if tr.hasCached('smooth_noise_spectrum'):
smooth_noise_dict = tr.getCached('smooth_noise_spectrum')
else:
smooth_noise_dict = None
if tr.hasParameter('snr_conf'):
snr_conf = tr.getParameter('snr_conf')
else:
snr_conf = None
trace_failed = tr.hasParameter('failure')
if trace_failed:
failure_reason = tr.getParameter('failure')['reason']
else:
failure_reason = ''
# Note that the theoretical spectra will only be available for
# horizontal channels
if tr.hasParameter('fit_spectra'):
fit_spectra_dict = tr.getParameter('fit_spectra')
else:
fit_spectra_dict = None
# ---------------------------------------------------------------------
# Compute model spectra
if fit_spectra_dict is not None:
model_spec = spectrum.model(
(fit_spectra_dict['moment'], fit_spectra_dict['stress_drop']),
freq=np.array(smooth_signal_dict['freq']),
dist=fit_spectra_dict['epi_dist'],
kappa=fit_spectra_dict['kappa'])
# ---------------------------------------------------------------------
# Acceleration time series plot
if trace_failed:
trace_status = " (failed)"
trace_title = tr.get_id() + trace_status
ax[j].set_title(trace_title, color="red")
else:
trace_status = " (passed)"
trace_title = tr.get_id() + trace_status
ax[j].set_title(trace_title)
dtimes = np.linspace(0, tr.stats.endtime - tr.stats.starttime,
tr.stats.npts)
ax[j].plot(dtimes, tr.data, 'k', linewidth=0.5)
# Show signal split as vertical dashed line
if tr.hasParameter('signal_split'):
split_dict = tr.getParameter('signal_split')
sptime = UTCDateTime(split_dict['split_time'])
dsec = sptime - tr.stats.starttime
ax[j].axvline(dsec, color='red', linestyle='dashed')
ax[j].set_xlabel('Time (s)')
ax[j].set_ylabel('Acceleration (cm/s/s)')
# ---------------------------------------------------------------------
# Velocity time series plot
tr_vel = st_vel[j]
dtimes = np.linspace(0, tr_vel.stats.endtime - tr_vel.stats.starttime,
tr_vel.stats.npts)
ax[j + ntrace].plot(dtimes, tr_vel.data, 'k', linewidth=0.5)
# Show signal split as vertical dashed line
if tr.hasParameter('signal_split'):
split_dict = tr.getParameter('signal_split')
sptime = UTCDateTime(split_dict['split_time'])
dsec = sptime - tr.stats.starttime
ax[j + ntrace].axvline(dsec, color='red', linestyle='dashed')
ax[j + ntrace].set_xlabel('Time (s)')
ax[j + ntrace].set_ylabel('Velocity (cm/s)')
# ---------------------------------------------------------------------
# Spectral plot
# Raw signal spec
if signal_dict is not None:
ax[j + 2 * ntrace].loglog(signal_dict['freq'],
signal_dict['spec'],
color='lightblue')
# Smoothed signal spec
if smooth_signal_dict is not None:
ax[j + 2 * ntrace].loglog(smooth_signal_dict['freq'],
smooth_signal_dict['spec'],
color='blue',
label='Signal')
# Raw noise spec
if noise_dict is not None:
ax[j + 2 * ntrace].loglog(noise_dict['freq'],
noise_dict['spec'],
color='salmon')
# Smoothed noise spec
if smooth_noise_dict is not None:
ax[j + 2 * ntrace].loglog(smooth_noise_dict['freq'],
smooth_noise_dict['spec'],
color='red',
label='Noise')
if fit_spectra_dict is not None:
# Model spec
ax[j + 2 * ntrace].loglog(smooth_signal_dict['freq'],
model_spec,
color='black',
linestyle='dashed')
# Corner frequency
ax[j + 2 * ntrace].axvline(fit_spectra_dict['f0'],
color='black',
linestyle='dashed')
ax[j + 2 * ntrace].set_xlabel('Frequency (Hz)')
ax[j + 2 * ntrace].set_ylabel('Amplitude (cm/s)')
# ---------------------------------------------------------------------
# Signal-to-noise ratio plot
if 'corner_frequencies' in tr.getParameterKeys():
hp = tr.getParameter('corner_frequencies')['highpass']
lp = tr.getParameter('corner_frequencies')['lowpass']
ax[j + 3 * ntrace].axvline(hp,
color='black',
linestyle='--',
label='Highpass')
ax[j + 3 * ntrace].axvline(lp,
color='black',
linestyle='--',
label='Lowpass')
if snr_conf is not None:
ax[j + 3 * ntrace].axhline(snr_conf['threshold'],
color='0.75',
linestyle='-',
linewidth=2)
ax[j + 3 * ntrace].axvline(snr_conf['max_freq'],
color='0.75',
linewidth=2,
linestyle='-')
ax[j + 3 * ntrace].axvline(snr_conf['min_freq'],
color='0.75',
linewidth=2,
linestyle='-')
if snr_dict is not None:
ax[j + 3 * ntrace].loglog(snr_dict['freq'],
snr_dict['snr'],
label='SNR')
ax[j + 3 * ntrace].set_ylabel('SNR')
ax[j + 3 * ntrace].set_xlabel('Frequency (Hz)')
stream_id = st.get_id()
# Do not save files if running tests
file_name = None
if 'CALLED_FROM_PYTEST' not in os.environ:
plt.subplots_adjust(left=0.05,
right=0.97,
hspace=0.25,
wspace=0.2,
top=0.97)
file_name = os.path.join(directory,
origin.id + '_' + stream_id + '.png')
plt.savefig(fname=file_name)
plt.close('all')
return file_name
def summary_plots(st, directory, origin):
"""Stream summary plot.
Args:
st (gmprocess.stationtrace.StationStream):
Stream of data.
directory (str):
Directory for saving plots.
origin (ScalarEvent):
Flattened subclass of Obspy Event.
"""
mpl.rcParams['font.size'] = 8
# Check if directory exists, and if not, create it.
if not os.path.exists(directory):
os.makedirs(directory)
# Setup figure for stream
nrows = 4
ntrace = min(len(st), 3)
fig = plt.figure(figsize=(3.8 * ntrace, 10))
gs = fig.add_gridspec(nrows, ntrace, height_ratios=[1, 1, 2, 2])
ax = [plt.subplot(g) for g in gs]
stream_id = st.get_id()
logging.debug('stream_id: %s' % stream_id)
logging.debug('passed: %s' % st.passed)
if st.passed:
plt.suptitle("M%s %s | %s (passed)" %
(origin.magnitude, origin.id, stream_id),
x=0.5, y=1.02)
else:
plt.suptitle("M%s %s | %s (failed)"
% (origin.magnitude, origin.id, stream_id),
color='red', x=0.5, y=1.02)
# Compute velocity
st_vel = st.copy()
st_vel = st_vel.integrate()
# process channels in preferred sort order (i.e., HN1, HN2, HNZ)
channels = [tr.stats.channel for tr in st]
if len(channels) < 3:
channelidx = np.argsort(channels).tolist()
else:
channelidx = range(3)
for j in channelidx:
tr = st[channelidx.index(j)]
# Break if j>3 becasue we can't on a page.
if j > 2:
logging.warning('Only plotting first 3 traces in stream.')
break
# ---------------------------------------------------------------------
# Get trace info
if tr.hasParameter('snr'):
snr_dict = tr.getParameter('snr')
else:
snr_dict = None
if tr.hasParameter('signal_spectrum'):
signal_dict = tr.getParameter('signal_spectrum')
else:
signal_dict = None
if tr.hasParameter('noise_spectrum'):
noise_dict = tr.getParameter('noise_spectrum')
else:
noise_dict = None
if tr.hasParameter('smooth_signal_spectrum'):
smooth_signal_dict = tr.getParameter('smooth_signal_spectrum')
else:
smooth_signal_dict = None
if tr.hasParameter('smooth_noise_spectrum'):
smooth_noise_dict = tr.getParameter('smooth_noise_spectrum')
else:
smooth_noise_dict = None
if tr.hasParameter('snr_conf'):
snr_conf = tr.getParameter('snr_conf')
else:
snr_conf = None
trace_failed = tr.hasParameter('failure')
if trace_failed:
failure_reason = tr.getParameter('failure')['reason']
else:
failure_reason = ''
# Note that the theoretical spectra will only be available for
# horizontal channels
if tr.hasParameter('fit_spectra'):
fit_spectra_dict = tr.getParameter('fit_spectra')
else:
fit_spectra_dict = None
# ---------------------------------------------------------------------
# Compute model spectra
if fit_spectra_dict is not None:
model_spec = spectrum.model(
freq=np.array(smooth_signal_dict['freq']),
dist=fit_spectra_dict['epi_dist'],
kappa=fit_spectra_dict['kappa'],
magnitude=fit_spectra_dict['magnitude'],
stress_drop=fit_spectra_dict['stress_drop']
)
# ---------------------------------------------------------------------
# Acceleration time series plot
if trace_failed:
trace_status = " (failed)"
trace_title = tr.get_id() + trace_status
ax[j].set_title(trace_title, color="red")
else:
trace_status = " (passed)"
trace_title = tr.get_id() + trace_status
ax[j].set_title(trace_title)
dtimes = tr.times('utcdatetime') - tr.times('utcdatetime')[0]
ax[j].plot(dtimes, tr.data, 'k', linewidth=0.5)
# Show signal split as vertical dashed line
if tr.hasParameter('signal_split'):
split_dict = tr.getParameter('signal_split')
dsec = split_dict['split_time'] - tr.times('utcdatetime')[0]
ax[j].axvline(dsec,
color='red', linestyle='dashed')
ax[j].set_xlabel('Time (s)')
ax[j].set_ylabel('Acceleration (cm/s/s)')
# ---------------------------------------------------------------------
# Velocity time series plot
tr_vel = st_vel[j]
dtimes = tr_vel.times('utcdatetime') - tr_vel.times('utcdatetime')[0]
ax[j + ntrace].plot(dtimes, tr_vel.data, 'k', linewidth=0.5)
# Show signal split as vertical dashed line
if tr.hasParameter('signal_split'):
split_dict = tr.getParameter('signal_split')
dsec = split_dict['split_time'] - tr.times('utcdatetime')[0]
ax[j + ntrace].axvline(dsec, color='red', linestyle='dashed')
ax[j + ntrace].set_xlabel('Time (s)')
ax[j + ntrace].set_ylabel('Velocity (cm/s)')
# ---------------------------------------------------------------------
# Spectral plot
# Raw signal spec
if signal_dict is not None:
ax[j + 2 * ntrace].loglog(signal_dict['freq'],
signal_dict['spec'],
color='lightblue')
# Smoothed signal spec
if smooth_signal_dict is not None:
ax[j + 2 * ntrace].loglog(smooth_signal_dict['freq'],
smooth_signal_dict['spec'],
color='blue',
label='Signal')
# Raw noise spec
if noise_dict is not None:
ax[j + 2 * ntrace].loglog(noise_dict['freq'],
noise_dict['spec'],
color='salmon')
# Smoothed noise spec
if smooth_noise_dict is not None:
ax[j + 2 * ntrace].loglog(smooth_noise_dict['freq'],
smooth_noise_dict['spec'],
color='red',
label='Noise')
if fit_spectra_dict is not None:
# Model spec
ax[j + 2 * ntrace].loglog(smooth_signal_dict['freq'],
model_spec,
color='black',
linestyle='dashed')
# Corner frequency
ax[j + 2 * ntrace].axvline(fit_spectra_dict['f0'],
color='black',
linestyle='dashed')
ax[j + 2 * ntrace].set_xlabel('Frequency (Hz)')
ax[j + 2 * ntrace].set_ylabel('Amplitude (cm/s)')
# ---------------------------------------------------------------------
# Signal-to-noise ratio plot
if 'corner_frequencies' in tr.getParameterKeys():
hp = tr.getParameter('corner_frequencies')['highpass']
lp = tr.getParameter('corner_frequencies')['lowpass']
ax[j + 3 * ntrace].axvline(hp,
color='black',
linestyle='--',
label='Highpass')
ax[j + 3 * ntrace].axvline(lp,
color='black',
linestyle='--',
label='Lowpass')
if snr_conf is not None:
ax[j + 3 * ntrace].axhline(snr_conf['threshold'],
color='0.75',
linestyle='-',
linewidth=2)
ax[j + 3 * ntrace].axvline(snr_conf['max_freq'],
color='0.75',
linewidth=2,
linestyle='-')
ax[j + 3 * ntrace].axvline(snr_conf['min_freq'],
color='0.75',
linewidth=2,
linestyle='-')
if snr_dict is not None:
ax[j + 3 * ntrace].loglog(snr_dict['freq'],
snr_dict['snr'],
label='SNR')
ax[j + 3 * ntrace].set_ylabel('SNR')
ax[j + 3 * ntrace].set_xlabel('Frequency (Hz)')
stream_id = st.get_id()
# Do not save files if running tests
if 'CALLED_FROM_PYTEST' not in os.environ:
plt.subplots_adjust(hspace=0.35, wspace=0.35, top=0.97)
file_name = os.path.join(
directory,
origin.id + '_' + stream_id + '.png')
plt.savefig(fname=file_name, bbox_inches='tight')
plt.close('all')
return st
def test_spectrum():
freq = np.logspace(-2, 2, 101)
mod = spectrum.model(freq, 10, kappa=0.035, magnitude=6.7)
np.testing.assert_allclose(mod[0], 0.21764373, atol=1e-5)
np.testing.assert_allclose(mod[50], 113.5025146, atol=1e-5)
np.testing.assert_allclose(mod[-1], 0.0032295, atol=1e-5)