def multiprocess_fft(inlist):
tr1=inlist[0]
tr2=inlist[1]
window_length=inlist[2]
overlap = inlist[3]
try:
return noise.noisecorr(tr1,tr2,window_length,overlap)
except:
return None
import noise
from obspy import read
from obspy.geodetics.base import gps2dist_azimuth
import matplotlib.pyplot as plt
ref_curve = np.loadtxt("Average_phase_velocity_rayleigh")
tr1 = read("preprocessed_data/SULZ.LHZ.CH.2013.219.processed.SAC")[0]
tr2 = read("preprocessed_data/VDL.LHZ.CH.2013.219.processed.SAC")[0]
# bad example with only one day of correlation
dist, az, baz = gps2dist_azimuth(tr1.stats.sac.stla, tr1.stats.sac.stlo,
tr2.stats.sac.stla, tr2.stats.sac.stlo)
freq, xcorr, n_corr_wins = noise.noisecorr(tr1,
tr2,
window_length=3600.,
overlap=0.5)
smoothed = noise.velocity_filter(freq,
xcorr,
dist / 1000.,
velband=(6.0, 5.0, 1.5, 0.5),
return_all=False)
crossings,phase_vel = noise.extract_phase_velocity(freq,smoothed,dist/1000.,ref_curve,\
freqmin=0.004,freqmax=0.25, min_vel=1.5, max_vel=5.0,min_amp=0.0,\
horizontal_polarization=False, smooth_spectrum=False,plotting=True)
plt.figure(figsize=(16, 10))
plt.subplot(2, 2, 1)
plt.plot(freq, np.real(xcorr), label='original')
from obspy import read, Trace
from obspy.geodetics.base import gps2dist_azimuth
import matplotlib.pyplot as plt
ref_curve = np.loadtxt("avg.txt")
tr1 = read("/Users/tzompantli/Desktop/tr1.mseed")[0]
tr2 = read("/Users/tzompantli/Desktop/tr2.mseed")[0]
# bad example with only one day of correlation
#dist,az,baz = gps2dist_azimuth(tr1.stats.sac.stla,tr1.stats.sac.stlo,
# tr2.stats.sac.stla,tr2.stats.sac.stlo)
dist = 150
freq, xcorr = noise.noisecorr(tr1, tr2, window_length=30., overlap=0.3)
smoothed = noise.velocity_filter(freq,
xcorr,
dist / 1000.,
cmin=.1,
cmax=1.0,
return_all=False)
cc = np.real(np.fft.fftshift(np.fft.ifft(smoothed)))
tr = Trace(data=cc)
tr.stats.sampling_rate = 50
tr.write('test.mseed', format='MSEED')
dkjd
def process_noise(stream,pair,comp_correlations,window_length,overlap,year,julday,flog):
global statdict
global save_monthly
global pairdict
global existing_corrdays
#print(datetime.datetime.now(),">>>>>> Processing pair",stat1,stat2,file=flog)
if len(stream)==0:
print("empty stream")
return
stat1 = pair[0]
stat2 = pair[1]
net1,sta1 = stat1.split(".")
net2,sta2 = stat2.split(".")
# sort the correlation components, so that the unrotated ones are processed
# first, and afterwards the rotated (unrotated traces are no longer acces-
# sible after rotation)
unrotated_components = []
rotated_components = []
for comp in comp_correlations:
if comp[0] in 'RT' or comp[1] in 'RT':
rotated_components.append(comp)
else:
unrotated_components.append(comp)
component_list = unrotated_components+rotated_components
# results will be saved to this dictionary
corr_list = {}
for components in component_list:
corr_list[components] = {}
corr_list[components]['spec'] = []
corr_list[components]['no_windows'] = []
rotated = False
for components in component_list:
if ((components[0] in 'NE' and components[1] in 'RT') or
(components[0] in 'RT' and components[1] in 'NE')):
print("mixed correlations of rotated (RT) and unrotated (NE) components is currently not supported.")
continue
if len(components) != 2:
print("correlation",components,"is not valid, skipping.")
continue
if (year,julday) in existing_corrdays[pair][components]:
continue
# correlations of unrotated components (ZZ,NN,EE,ZN,...)
if components[0] in 'ZNE' and components[1] in 'ZNE':
st1 = stream.select(network=net1,station=sta1,component=components[0])
st2 = stream.select(network=net2,station=sta2,component=components[1])
if len(st1) == 0 or len(st2) == 0:
continue
st1,st2 = noise.adapt_timespan(st1,st2,min_overlap=window_length,
interpolate=True,copystreams=True)
if len(st1) == 0 or len(st2) == 0:
continue
# if R or T is in the correlation components, rotate streams
elif not rotated:
st1z = stream.select(network=net1,station=sta1,component='Z')
st1n = stream.select(network=net1,station=sta1,component='N')
st1e = stream.select(network=net1,station=sta1,component='E')
st2z = stream.select(network=net2,station=sta2,component='Z')
st2n = stream.select(network=net2,station=sta2,component='N')
st2e = stream.select(network=net2,station=sta2,component='E')
if len(st1n)==0 or len(st2n)==0 or len(st1e)==0 or len(st2e)==0:
return # abort, will not be possible to rotate to RT coords
st1,st2 = noise.adapt_timespan((st1z+st1n+st1e),
(st2z+st2n+st2e),
min_overlap=window_length,
interpolate=True,copystreams=True)
if len(st1) < 2 or len(st2) < 2:
return # abort, will not be possible to rotate to RT coords
# check that the time span is really the same
if (st1[0].stats.starttime != st1[1].stats.starttime or
st1[0].stats.starttime != st2[0].stats.starttime or
st1[0].stats.starttime != st2[1].stats.starttime or
st1[0].stats.endtime != st1[1].stats.endtime or
st1[0].stats.endtime != st2[0].stats.endtime or
st1[0].stats.endtime != st2[1].stats.endtime or
st1[0].stats.endtime - st1[0].stats.starttime < window_length):
raise Exception("this should not be possible!")
# az = azimuth from station1 -> station2
# baz = azimuth from station2 -> station1
# for stream2 the back azimuth points in direction of station1
# for stream1 the azimuth points in direction of station2
# BUT 180. degree shift is needed so that the radial components point in the same direction!
# otherwise they point towards each other => transverse comp would be also opposed
try:
st1.rotate('NE->RT',back_azimuth=(pairdict[pair]['az']+180.)%360.)
except:
print("Error rotating stream",file=flog)
print(st1,file=flog)
raise Exception("Error rotating stream")
continue
try:
st2.rotate('NE->RT',back_azimuth=pairdict[pair]['baz'])
except:
print("Error rotating stream",file=flog)
print(st2,file=flog)
raise Exception("Error rotating stream")
continue
rotated = True
# timewindows that have overlapping data windows longer than window_length
windows = []
for trace in st1:
window = (trace.stats.starttime,trace.stats.endtime)
if window in windows:
continue
windows.append(window)
for timewin in windows:
corrstream1 = st1.select(component=components[0]).slice(
starttime=timewin[0],endtime=timewin[1])
corrstream2 = st2.select(component=components[1]).slice(
starttime=timewin[0],endtime=timewin[1])
# check for nan/inf in data
data_errors = False
for tr in (corrstream1+corrstream2):
if tr.stats.endtime-tr.stats.starttime < window_length:
# trace too short
data_errors = True
if np.std(tr.data) == 0.:
print("data all zero",file=flog)
print(tr,file=flog)
data_errors = True
if np.isnan(tr.data).any() or np.isinf(tr.data).any():
print("nan/inf in data",file=flog)
print(tr,file=flog)
data_errors = True
if data_errors:
continue
# check that the time span is really the same
if (corrstream1[0].stats.starttime != corrstream2[0].stats.starttime or
corrstream1[0].stats.endtime != corrstream2[0].stats.endtime):
raise Exception("this should not be possible!")
# finally, do the correlation
try:
freq,spec,wincount = noise.noisecorr(
corrstream1[0],corrstream2[0],
window_length,overlap,whiten=whiten,onebit=onebit)
except:
print("could not correlate",corrstream1[0].stats.id,
corrstream2[0].stats.id,components)
continue
corr_list[components]['spec'].append(spec)
corr_list[components]['no_windows'].append(wincount)
# finished correlating. saving the results to a file
for components in component_list:
if len(corr_list[components]['spec']) == 0:
continue
corr_spectrum = np.average(
np.array(corr_list[components]['spec']),axis=0,
weights=corr_list[components]['no_windows'])
filepath = getfilepath(stat1,stat2,components,
pairdict[pair]['dist'],overlap)
if os.path.isfile(filepath):
with open(filepath,"rb") as f:
corr_dict = pickle.load(f)
if (year,julday) in corr_dict['corrdays']:
print("correlation day already in database!",filepath,year,julday)
continue
else:
corr_dict['corrdays'].append((year,julday))
corr_dict['spectrum'] = np.average(
[corr_spectrum,corr_dict['spectrum']],axis=0,
weights=[np.sum(corr_list[components]['no_windows']),
corr_dict['no_wins']])
corr_dict['no_wins'] += np.sum(corr_list[components]['no_windows'])
else:
corr_dict = {}
corr_dict['corrdays'] = [(year,julday)]
corr_dict['spectrum'] = corr_spectrum
corr_dict['freq'] = freq
corr_dict['no_wins'] = np.sum(corr_list[components]['no_windows'])
corr_dict['dist'] = pairdict[pair]['dist']
corr_dict['az'] = pairdict[pair]['az']
corr_dict['baz'] = pairdict[pair]['baz']
corr_dict['component'] = components
corr_dict['station1'] = statdict[stat1]
corr_dict['station2'] = statdict[stat2]
corr_dict['station1']['id'] = stat1
corr_dict['station2']['id'] = stat2
if save_monthly:
month = str(year)+"."+str(UTCDateTime(year=year,julday=julday).month)
if not 'spectrum.'+month in corr_dict.keys():
corr_dict['spectrum.'+month] = corr_spectrum
corr_dict['no_wins.'+month] = np.sum(
corr_list[components]['no_windows'])
else:
corr_dict['spectrum.'+month] = np.average(
[corr_spectrum,corr_dict['spectrum.'+month]],axis=0,
weights=[np.sum(corr_list[components]['no_windows']),
corr_dict['no_wins.'+month]])
corr_dict['no_wins.'+month] += np.sum(
corr_list[components]['no_windows'])
with open(filepath,"wb") as f:
pickle.dump(corr_dict,f)
#print("successfully correlated",stat1,stat2,"comp:",corrcomps,"day:",year,julday)
return