def obtain_timeshifts(st):
'''function to compute time shift of overlapping files '''
'''st: obspy stream containing one station channel pair '''
# check if all traces are from same station/channel
for tr in st:
if tr.stats.station == st[0].stats.station:
pass
else:
raise NameError("Traces in st recorded by different stations")
if tr.stats.channel == st[0].stats.channel:
pass
else:
raise NameError("Traces in st recorded on different channels")
# empty stream for time corrected traces; add first trace
cor = Stream()
# "take_previous" determines if first trace is subject to syncronization to previous mseed file
take_previous = False
# check if previous mseed file is available
try:
stn = st[0].stats.station
chn = st[0].stats.channel
stime = st[0].stats.starttime
prev = read("/scratch/flindner/G7Jul01/MSEED/4D.%s..%s.%04d%02d%02d%02d*" \
% (stn, chn, stime.year, stime.month, stime.day, stime.hour), starttime=stime - 5)
cor += prev[0]
# available -> synchronize
take_previous = True
except:
pass
# in case of no sync to previous file, start with timeshift between first and second trace
if take_previous == False:
cor += st[0]
st = st[1:]
warnings.warn("Synchronization to previous MSEED file failed!")
# empty list for dts
dts = []
# add next trace
for i, tr in enumerate(st):
cor += tr
# helper stream - overlap
helper = cor.copy()
t1 = helper[1].stats.starttime
t2 = helper[0].stats.endtime
helper.trim(t1, t2)
delta = helper[0].stats.delta
npts = helper[0].stats.npts
# even number of samples
if (npts % 2) != 0:
helper.trim(t1, t2-delta)
delta = helper[0].stats.delta
# cross-correlation in order to obtain timeshift
xcorr = np.correlate(helper[0].data, helper[1].data, "same")
xcorr = Trace(data=xcorr)
xcorr.stats.delta = delta
xcorr.resample(1000)
ndelta = xcorr.stats.delta
nnpts = xcorr.stats.npts
dt = (np.argmax(xcorr.data) - nnpts/2.) * ndelta
dt = round(dt, 3)
# add estimated time shift to second trace
cor[1].stats.starttime += dt
# test if time series are perfectly aligned
check = True
count = 0
while check:
start = cor[1].stats.starttime
end = cor[0].stats.endtime
test = cor.slice(start, end)
eq_st = False
if round(test[0].stats.starttime.timestamp, 4) == round(test[1].stats.starttime.timestamp, 4):
eq_st = True
eq_ar = np.array_equal(test[0].data, test[1].data)
if eq_ar == True and eq_st == True:
# if perfectly aligned, quit testing loop
check = False
if count == 0:
dts.append(dt)
# if not perfectly aligned, undo time shift obtained by xcorr and shift manually
else:
if count == 0:
# undo time shift obtained from xcorr
cor[1].stats.starttime -= dt
mshift = int((end - start) / delta) + 1000
cor[1].stats.starttime -= mshift*delta
elif count > 0:
cor[1].stats.starttime += delta
cor[1].stats.starttime.microsecond = np.round(cor[1].stats.starttime.microsecond, 3)
count += 1
# in case of take_previous == True, remove previous file from stream (was just needed for time sync)
if i==0 and take_previous:
for tr in cor.select(network="4D"):
cor.remove(tr)
# time shift obtained from shifting
if count > 1:
dt = (-mshift + count - 2) * delta
dts.append(dt)
cor.merge(method=1)
# check number of written times shifts
dts = np.asarray(dts)
if len(dts) != len(st):
raise ValueError("Number of time shifts is not correct")
return dts, take_previous
