marcel/pylot
3
0
Fork 0
Code Issues 7 Pull Requests 1 Releases Wiki Activity

WALL-E: Einmal aufräumen und zurück!

Browse Source
This commit is contained in:
Sebastian Wehling-Benatelli
2016-03-30 08:14:58 +02:00
parent a2640e3126
commit d7cfd0d176
29 changed files with 1285 additions and 1105 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
pylot/core/pick/autopick.py
View File

@@ -144,7 +144,7 @@ def autopickstation(wfstream, pickparam, verbose=False):
Sflag = 0
Pmarker = []
Ao = None # Wood-Anderson peak-to-peak amplitude
picker = 'autoPyLoT' # name of the picking programm
picker = 'autoPyLoT' # name of the picking programm
# split components
zdat = wfstream.select(component="Z")
@@ -867,19 +867,19 @@ def iteratepicker(wf, NLLocfile, picks, badpicks, pickparameter):
pickparameter.setParam(noisefactor=1.0)
pickparameter.setParam(zfac=1.0)
print(
"iteratepicker: The following picking parameters have been modified for iterative picking:")
"iteratepicker: The following picking parameters have been modified for iterative picking:")
print(
"pstart: %fs => %fs" % (pstart_old, pickparameter.getParam('pstart')))
"pstart: %fs => %fs" % (pstart_old, pickparameter.getParam('pstart')))
print(
"pstop: %fs => %fs" % (pstop_old, pickparameter.getParam('pstop')))
"pstop: %fs => %fs" % (pstop_old, pickparameter.getParam('pstop')))
print(
"sstop: %fs => %fs" % (sstop_old, pickparameter.getParam('sstop')))
"sstop: %fs => %fs" % (sstop_old, pickparameter.getParam('sstop')))
print("pickwinP: %fs => %fs" % (
pickwinP_old, pickparameter.getParam('pickwinP')))
pickwinP_old, pickparameter.getParam('pickwinP')))
print("Precalcwin: %fs => %fs" % (
Precalcwin_old, pickparameter.getParam('Precalcwin')))
Precalcwin_old, pickparameter.getParam('Precalcwin')))
print("noisefactor: %f => %f" % (
noisefactor_old, pickparameter.getParam('noisefactor')))
noisefactor_old, pickparameter.getParam('noisefactor')))
print("zfac: %f => %f" % (zfac_old, pickparameter.getParam('zfac')))
# repick station

255
pylot/core/pick/charfuns.py
View File

@@ -21,10 +21,12 @@ import matplotlib.pyplot as plt
import numpy as np
from obspy.core import Stream
class CharacteristicFunction(object):
'''
SuperClass for different types of characteristic functions.
'''
def __init__(self, data, cut, t2=None, order=None, t1=None, fnoise=None, stealthMode=False):
'''
Initialize data type object with information from the original
@@ -103,9 +105,9 @@ class CharacteristicFunction(object):
def setARdetStep(self, t1):
if t1:
self.ARdetStep = []
self.ARdetStep.append(t1 / 4)
self.ARdetStep.append(int(np.ceil(self.getTime2() / self.getIncrement()) / 4))
self.ARdetStep = []
self.ARdetStep.append(t1 / 4)
self.ARdetStep.append(int(np.ceil(self.getTime2() / self.getIncrement()) / 4))
def getOrder(self):
return self.order
@@ -150,14 +152,14 @@ class CharacteristicFunction(object):
if cut is not None:
if len(self.orig_data) == 1:
if self.cut[0] == 0 and self.cut[1] == 0:
start = 0
stop = len(self.orig_data[0])
start = 0
stop = len(self.orig_data[0])
elif self.cut[0] == 0 and self.cut[1] is not 0:
start = 0
stop = self.cut[1] / self.dt
start = 0
stop = self.cut[1] / self.dt
else:
start = self.cut[0] / self.dt
stop = self.cut[1] / self.dt
start = self.cut[0] / self.dt
stop = self.cut[1] / self.dt
zz = self.orig_data.copy()
z1 = zz[0].copy()
zz[0].data = z1.data[int(start):int(stop)]
@@ -165,16 +167,16 @@ class CharacteristicFunction(object):
return data
elif len(self.orig_data) == 2:
if self.cut[0] == 0 and self.cut[1] == 0:
start = 0
stop = min([len(self.orig_data[0]), len(self.orig_data[1])])
start = 0
stop = min([len(self.orig_data[0]), len(self.orig_data[1])])
elif self.cut[0] == 0 and self.cut[1] is not 0:
start = 0
stop = min([self.cut[1] / self.dt, len(self.orig_data[0]),
len(self.orig_data[1])])
start = 0
stop = min([self.cut[1] / self.dt, len(self.orig_data[0]),
len(self.orig_data[1])])
else:
start = max([0, self.cut[0] / self.dt])
stop = min([self.cut[1] / self.dt, len(self.orig_data[0]),
len(self.orig_data[1])])
start = max([0, self.cut[0] / self.dt])
stop = min([self.cut[1] / self.dt, len(self.orig_data[0]),
len(self.orig_data[1])])
hh = self.orig_data.copy()
h1 = hh[0].copy()
h2 = hh[1].copy()
@@ -184,16 +186,16 @@ class CharacteristicFunction(object):
return data
elif len(self.orig_data) == 3:
if self.cut[0] == 0 and self.cut[1] == 0:
start = 0
stop = min([self.cut[1] / self.dt, len(self.orig_data[0]),
len(self.orig_data[1]), len(self.orig_data[2])])
start = 0
stop = min([self.cut[1] / self.dt, len(self.orig_data[0]),
len(self.orig_data[1]), len(self.orig_data[2])])
elif self.cut[0] == 0 and self.cut[1] is not 0:
start = 0
stop = self.cut[1] / self.dt
start = 0
stop = self.cut[1] / self.dt
else:
start = max([0, self.cut[0] / self.dt])
stop = min([self.cut[1] / self.dt, len(self.orig_data[0]),
len(self.orig_data[1]), len(self.orig_data[2])])
start = max([0, self.cut[0] / self.dt])
stop = min([self.cut[1] / self.dt, len(self.orig_data[0]),
len(self.orig_data[1]), len(self.orig_data[2])])
hh = self.orig_data.copy()
h1 = hh[0].copy()
h2 = hh[1].copy()
@@ -223,13 +225,13 @@ class AICcf(CharacteristicFunction):
def calcCF(self, data):
#if self._getStealthMode() is False:
# if self._getStealthMode() is False:
# print 'Calculating AIC ...'
x = self.getDataArray()
xnp = x[0].data
nn = np.isnan(xnp)
if len(nn) > 1:
xnp[nn] = 0
xnp[nn] = 0
datlen = len(xnp)
k = np.arange(1, datlen)
cf = np.zeros(datlen)
@@ -247,6 +249,7 @@ class AICcf(CharacteristicFunction):
self.cf = cf - np.mean(cf)
self.xcf = x
class HOScf(CharacteristicFunction):
'''
Function to calculate skewness (statistics of order 3) or kurtosis
@@ -257,38 +260,38 @@ class HOScf(CharacteristicFunction):
def calcCF(self, data):
x = self.getDataArray(self.getCut())
xnp =x[0].data
xnp = x[0].data
nn = np.isnan(xnp)
if len(nn) > 1:
xnp[nn] = 0
xnp[nn] = 0
if self.getOrder() == 3: # this is skewness
#if self._getStealthMode() is False:
# if self._getStealthMode() is False:
# print 'Calculating skewness ...'
y = np.power(xnp, 3)
y1 = np.power(xnp, 2)
elif self.getOrder() == 4: # this is kurtosis
#if self._getStealthMode() is False:
# if self._getStealthMode() is False:
# print 'Calculating kurtosis ...'
y = np.power(xnp, 4)
y1 = np.power(xnp, 2)
#Initialisation
#t2: long term moving window
# Initialisation
# t2: long term moving window
ilta = int(round(self.getTime2() / self.getIncrement()))
lta = y[0]
lta1 = y1[0]
#moving windows
# moving windows
LTA = np.zeros(len(xnp))
for j in range(0, len(xnp)):
if j < 4:
LTA[j] = 0
elif j <= ilta:
lta = (y[j] + lta * (j-1)) / j
lta1 = (y1[j] + lta1 * (j-1)) / j
lta = (y[j] + lta * (j - 1)) / j
lta1 = (y1[j] + lta1 * (j - 1)) / j
else:
lta = (y[j] - y[j - ilta]) / ilta + lta
lta1 = (y1[j] - y1[j - ilta]) / ilta + lta1
#define LTA
# define LTA
if self.getOrder() == 3:
LTA[j] = lta / np.power(lta1, 1.5)
elif self.getOrder() == 4:
@@ -296,13 +299,12 @@ class HOScf(CharacteristicFunction):
nn = np.isnan(LTA)
if len(nn) > 1:
LTA[nn] = 0
LTA[nn] = 0
self.cf = LTA
self.xcf = x
class ARZcf(CharacteristicFunction):
def calcCF(self, data):
print 'Calculating AR-prediction error from single trace ...'
@@ -310,33 +312,33 @@ class ARZcf(CharacteristicFunction):
xnp = x[0].data
nn = np.isnan(xnp)
if len(nn) > 1:
xnp[nn] = 0
#some parameters needed
#add noise to time series
xnp[nn] = 0
# some parameters needed
# add noise to time series
xnoise = xnp + np.random.normal(0.0, 1.0, len(xnp)) * self.getFnoise() * max(abs(xnp))
tend = len(xnp)
#Time1: length of AR-determination window [sec]
#Time2: length of AR-prediction window [sec]
ldet = int(round(self.getTime1() / self.getIncrement())) #length of AR-determination window [samples]
lpred = int(np.ceil(self.getTime2() / self.getIncrement())) #length of AR-prediction window [samples]
# Time1: length of AR-determination window [sec]
# Time2: length of AR-prediction window [sec]
ldet = int(round(self.getTime1() / self.getIncrement())) # length of AR-determination window [samples]
lpred = int(np.ceil(self.getTime2() / self.getIncrement())) # length of AR-prediction window [samples]
cf = np.zeros(len(xnp))
loopstep = self.getARdetStep()
arcalci = ldet + self.getOrder() #AR-calculation index
arcalci = ldet + self.getOrder() # AR-calculation index
for i in range(ldet + self.getOrder(), tend - lpred - 1):
if i == arcalci:
#determination of AR coefficients
#to speed up calculation, AR-coefficients are calculated only every i+loopstep[1]!
self.arDetZ(xnoise, self.getOrder(), i-ldet, i)
# determination of AR coefficients
# to speed up calculation, AR-coefficients are calculated only every i+loopstep[1]!
self.arDetZ(xnoise, self.getOrder(), i - ldet, i)
arcalci = arcalci + loopstep[1]
#AR prediction of waveform using calculated AR coefficients
# AR prediction of waveform using calculated AR coefficients
self.arPredZ(xnp, self.arpara, i + 1, lpred)
#prediction error = CF
cf[i + lpred-1] = np.sqrt(np.sum(np.power(self.xpred[i:i + lpred-1] - xnp[i:i + lpred-1], 2)) / lpred)
# prediction error = CF
cf[i + lpred - 1] = np.sqrt(np.sum(np.power(self.xpred[i:i + lpred - 1] - xnp[i:i + lpred - 1], 2)) / lpred)
nn = np.isnan(cf)
if len(nn) > 1:
cf[nn] = 0
#remove zeros and artefacts
cf[nn] = 0
# remove zeros and artefacts
tap = np.hanning(len(cf))
cf = tap * cf
io = np.where(cf == 0)
@@ -366,25 +368,25 @@ class ARZcf(CharacteristicFunction):
Output: AR parameters arpara
'''
#recursive calculation of data vector (right part of eq. 6.5 in Kueperkoch et al. (2012)
# recursive calculation of data vector (right part of eq. 6.5 in Kueperkoch et al. (2012)
rhs = np.zeros(self.getOrder())
for k in range(0, self.getOrder()):
for i in range(rind, ldet+1):
for i in range(rind, ldet + 1):
ki = k + 1
rhs[k] = rhs[k] + data[i] * data[i - ki]
#recursive calculation of data array (second sum at left part of eq. 6.5 in Kueperkoch et al. 2012)
A = np.zeros((self.getOrder(),self.getOrder()))
# recursive calculation of data array (second sum at left part of eq. 6.5 in Kueperkoch et al. 2012)
A = np.zeros((self.getOrder(), self.getOrder()))
for k in range(1, self.getOrder() + 1):
for j in range(1, k + 1):
for i in range(rind, ldet+1):
for i in range(rind, ldet + 1):
ki = k - 1
ji = j - 1
A[ki,ji] = A[ki,ji] + data[i - j] * data[i - k]
A[ki, ji] = A[ki, ji] + data[i - j] * data[i - k]
A[ji,ki] = A[ki,ji]
A[ji, ki] = A[ki, ji]
#apply Moore-Penrose inverse for SVD yielding the AR-parameters
# apply Moore-Penrose inverse for SVD yielding the AR-parameters
self.arpara = np.dot(np.linalg.pinv(A), rhs)
def arPredZ(self, data, arpara, rind, lpred):
@@ -406,10 +408,10 @@ class ARZcf(CharacteristicFunction):
Output: predicted waveform z
'''
#be sure of the summation indeces
# be sure of the summation indeces
if rind < len(arpara):
rind = len(arpara)
if rind > len(data) - lpred :
if rind > len(data) - lpred:
rind = len(data) - lpred
if lpred < 1:
lpred = 1
@@ -426,7 +428,6 @@ class ARZcf(CharacteristicFunction):
class ARHcf(CharacteristicFunction):
def calcCF(self, data):
print 'Calculating AR-prediction error from both horizontal traces ...'
@@ -434,41 +435,42 @@ class ARHcf(CharacteristicFunction):
xnp = self.getDataArray(self.getCut())
n0 = np.isnan(xnp[0].data)
if len(n0) > 1:
xnp[0].data[n0] = 0
xnp[0].data[n0] = 0
n1 = np.isnan(xnp[1].data)
if len(n1) > 1:
xnp[1].data[n1] = 0
xnp[1].data[n1] = 0
#some parameters needed
#add noise to time series
# some parameters needed
# add noise to time series
xenoise = xnp[0].data + np.random.normal(0.0, 1.0, len(xnp[0].data)) * self.getFnoise() * max(abs(xnp[0].data))
xnnoise = xnp[1].data + np.random.normal(0.0, 1.0, len(xnp[1].data)) * self.getFnoise() * max(abs(xnp[1].data))
Xnoise = np.array( [xenoise.tolist(), xnnoise.tolist()] )
Xnoise = np.array([xenoise.tolist(), xnnoise.tolist()])
tend = len(xnp[0].data)
#Time1: length of AR-determination window [sec]
#Time2: length of AR-prediction window [sec]
ldet = int(round(self.getTime1() / self.getIncrement())) #length of AR-determination window [samples]
lpred = int(np.ceil(self.getTime2() / self.getIncrement())) #length of AR-prediction window [samples]
# Time1: length of AR-determination window [sec]
# Time2: length of AR-prediction window [sec]
ldet = int(round(self.getTime1() / self.getIncrement())) # length of AR-determination window [samples]
lpred = int(np.ceil(self.getTime2() / self.getIncrement())) # length of AR-prediction window [samples]
cf = np.zeros(len(xenoise))
loopstep = self.getARdetStep()
arcalci = lpred + self.getOrder() - 1 #AR-calculation index
#arcalci = ldet + self.getOrder() - 1 #AR-calculation index
arcalci = lpred + self.getOrder() - 1 # AR-calculation index
# arcalci = ldet + self.getOrder() - 1 #AR-calculation index
for i in range(lpred + self.getOrder() - 1, tend - 2 * lpred + 1):
if i == arcalci:
#determination of AR coefficients
#to speed up calculation, AR-coefficients are calculated only every i+loopstep[1]!
self.arDetH(Xnoise, self.getOrder(), i-ldet, i)
# determination of AR coefficients
# to speed up calculation, AR-coefficients are calculated only every i+loopstep[1]!
self.arDetH(Xnoise, self.getOrder(), i - ldet, i)
arcalci = arcalci + loopstep[1]
#AR prediction of waveform using calculated AR coefficients
# AR prediction of waveform using calculated AR coefficients
self.arPredH(xnp, self.arpara, i + 1, lpred)
#prediction error = CF
# prediction error = CF
cf[i + lpred] = np.sqrt(np.sum(np.power(self.xpred[0][i:i + lpred] - xnp[0][i:i + lpred], 2) \
+ np.power(self.xpred[1][i:i + lpred] - xnp[1][i:i + lpred], 2)) / (2 * lpred))
+ np.power(self.xpred[1][i:i + lpred] - xnp[1][i:i + lpred], 2)) / (
2 * lpred))
nn = np.isnan(cf)
if len(nn) > 1:
cf[nn] = 0
#remove zeros and artefacts
cf[nn] = 0
# remove zeros and artefacts
tap = np.hanning(len(cf))
cf = tap * cf
io = np.where(cf == 0)
@@ -500,24 +502,24 @@ class ARHcf(CharacteristicFunction):
Output: AR parameters arpara
'''
#recursive calculation of data vector (right part of eq. 6.5 in Kueperkoch et al. (2012)
# recursive calculation of data vector (right part of eq. 6.5 in Kueperkoch et al. (2012)
rhs = np.zeros(self.getOrder())
for k in range(0, self.getOrder()):
for i in range(rind, ldet):
rhs[k] = rhs[k] + data[0,i] * data[0,i - k] + data[1,i] * data[1,i - k]
rhs[k] = rhs[k] + data[0, i] * data[0, i - k] + data[1, i] * data[1, i - k]
#recursive calculation of data array (second sum at left part of eq. 6.5 in Kueperkoch et al. 2012)
A = np.zeros((4,4))
# recursive calculation of data array (second sum at left part of eq. 6.5 in Kueperkoch et al. 2012)
A = np.zeros((4, 4))
for k in range(1, self.getOrder() + 1):
for j in range(1, k + 1):
for i in range(rind, ldet):
ki = k - 1
ji = j - 1
A[ki,ji] = A[ki,ji] + data[0,i - ji] * data[0,i - ki] + data[1,i - ji] *data[1,i - ki]
A[ki, ji] = A[ki, ji] + data[0, i - ji] * data[0, i - ki] + data[1, i - ji] * data[1, i - ki]
A[ji,ki] = A[ki,ji]
A[ji, ki] = A[ki, ji]
#apply Moore-Penrose inverse for SVD yielding the AR-parameters
# apply Moore-Penrose inverse for SVD yielding the AR-parameters
self.arpara = np.dot(np.linalg.pinv(A), rhs)
def arPredH(self, data, arpara, rind, lpred):
@@ -540,7 +542,7 @@ class ARHcf(CharacteristicFunction):
Output: predicted waveform z
:type: structured array
'''
#be sure of the summation indeces
# be sure of the summation indeces
if rind < len(arpara) + 1:
rind = len(arpara) + 1
if rind > len(data[0]) - lpred + 1:
@@ -558,11 +560,11 @@ class ARHcf(CharacteristicFunction):
z1[i] = z1[i] + arpara[ji] * z1[i - ji]
z2[i] = z2[i] + arpara[ji] * z2[i - ji]
z = np.array( [z1.tolist(), z2.tolist()] )
z = np.array([z1.tolist(), z2.tolist()])
self.xpred = z
class AR3Ccf(CharacteristicFunction):
class AR3Ccf(CharacteristicFunction):
def calcCF(self, data):
print 'Calculating AR-prediction error from all 3 components ...'
@@ -570,46 +572,47 @@ class AR3Ccf(CharacteristicFunction):
xnp = self.getDataArray(self.getCut())
n0 = np.isnan(xnp[0].data)
if len(n0) > 1:
xnp[0].data[n0] = 0
xnp[0].data[n0] = 0
n1 = np.isnan(xnp[1].data)
if len(n1) > 1:
xnp[1].data[n1] = 0
xnp[1].data[n1] = 0
n2 = np.isnan(xnp[2].data)
if len(n2) > 1:
xnp[2].data[n2] = 0
xnp[2].data[n2] = 0
#some parameters needed
#add noise to time series
# some parameters needed
# add noise to time series
xenoise = xnp[0].data + np.random.normal(0.0, 1.0, len(xnp[0].data)) * self.getFnoise() * max(abs(xnp[0].data))
xnnoise = xnp[1].data + np.random.normal(0.0, 1.0, len(xnp[1].data)) * self.getFnoise() * max(abs(xnp[1].data))
xznoise = xnp[2].data + np.random.normal(0.0, 1.0, len(xnp[2].data)) * self.getFnoise() * max(abs(xnp[2].data))
Xnoise = np.array( [xenoise.tolist(), xnnoise.tolist(), xznoise.tolist()] )
Xnoise = np.array([xenoise.tolist(), xnnoise.tolist(), xznoise.tolist()])
tend = len(xnp[0].data)
#Time1: length of AR-determination window [sec]
#Time2: length of AR-prediction window [sec]
ldet = int(round(self.getTime1() / self.getIncrement())) #length of AR-determination window [samples]
lpred = int(np.ceil(self.getTime2() / self.getIncrement())) #length of AR-prediction window [samples]
# Time1: length of AR-determination window [sec]
# Time2: length of AR-prediction window [sec]
ldet = int(round(self.getTime1() / self.getIncrement())) # length of AR-determination window [samples]
lpred = int(np.ceil(self.getTime2() / self.getIncrement())) # length of AR-prediction window [samples]
cf = np.zeros(len(xenoise))
loopstep = self.getARdetStep()
arcalci = ldet + self.getOrder() - 1 #AR-calculation index
arcalci = ldet + self.getOrder() - 1 # AR-calculation index
for i in range(ldet + self.getOrder() - 1, tend - 2 * lpred + 1):
if i == arcalci:
#determination of AR coefficients
#to speed up calculation, AR-coefficients are calculated only every i+loopstep[1]!
self.arDet3C(Xnoise, self.getOrder(), i-ldet, i)
# determination of AR coefficients
# to speed up calculation, AR-coefficients are calculated only every i+loopstep[1]!
self.arDet3C(Xnoise, self.getOrder(), i - ldet, i)
arcalci = arcalci + loopstep[1]
#AR prediction of waveform using calculated AR coefficients
# AR prediction of waveform using calculated AR coefficients
self.arPred3C(xnp, self.arpara, i + 1, lpred)
#prediction error = CF
# prediction error = CF
cf[i + lpred] = np.sqrt(np.sum(np.power(self.xpred[0][i:i + lpred] - xnp[0][i:i + lpred], 2) \
+ np.power(self.xpred[1][i:i + lpred] - xnp[1][i:i + lpred], 2) \
+ np.power(self.xpred[2][i:i + lpred] - xnp[2][i:i + lpred], 2)) / (3 * lpred))
+ np.power(self.xpred[1][i:i + lpred] - xnp[1][i:i + lpred], 2) \
+ np.power(self.xpred[2][i:i + lpred] - xnp[2][i:i + lpred], 2)) / (
3 * lpred))
nn = np.isnan(cf)
if len(nn) > 1:
cf[nn] = 0
#remove zeros and artefacts
cf[nn] = 0
# remove zeros and artefacts
tap = np.hanning(len(cf))
cf = tap * cf
io = np.where(cf == 0)
@@ -641,26 +644,26 @@ class AR3Ccf(CharacteristicFunction):
Output: AR parameters arpara
'''
#recursive calculation of data vector (right part of eq. 6.5 in Kueperkoch et al. (2012)
# recursive calculation of data vector (right part of eq. 6.5 in Kueperkoch et al. (2012)
rhs = np.zeros(self.getOrder())
for k in range(0, self.getOrder()):
for i in range(rind, ldet):
rhs[k] = rhs[k] + data[0,i] * data[0,i - k] + data[1,i] * data[1,i - k] \
+ data[2,i] * data[2,i - k]
rhs[k] = rhs[k] + data[0, i] * data[0, i - k] + data[1, i] * data[1, i - k] \
+ data[2, i] * data[2, i - k]
#recursive calculation of data array (second sum at left part of eq. 6.5 in Kueperkoch et al. 2012)
A = np.zeros((4,4))
# recursive calculation of data array (second sum at left part of eq. 6.5 in Kueperkoch et al. 2012)
A = np.zeros((4, 4))
for k in range(1, self.getOrder() + 1):
for j in range(1, k + 1):
for i in range(rind, ldet):
ki = k - 1
ji = j - 1
A[ki,ji] = A[ki,ji] + data[0,i - ji] * data[0,i - ki] + data[1,i - ji] *data[1,i - ki] \
+ data[2,i - ji] *data[2,i - ki]
A[ki, ji] = A[ki, ji] + data[0, i - ji] * data[0, i - ki] + data[1, i - ji] * data[1, i - ki] \
+ data[2, i - ji] * data[2, i - ki]
A[ji,ki] = A[ki,ji]
A[ji, ki] = A[ki, ji]
#apply Moore-Penrose inverse for SVD yielding the AR-parameters
# apply Moore-Penrose inverse for SVD yielding the AR-parameters
self.arpara = np.dot(np.linalg.pinv(A), rhs)
def arPred3C(self, data, arpara, rind, lpred):
@@ -683,7 +686,7 @@ class AR3Ccf(CharacteristicFunction):
Output: predicted waveform z
:type: structured array
'''
#be sure of the summation indeces
# be sure of the summation indeces
if rind < len(arpara) + 1:
rind = len(arpara) + 1
if rind > len(data[0]) - lpred + 1:
@@ -703,5 +706,5 @@ class AR3Ccf(CharacteristicFunction):
z2[i] = z2[i] + arpara[ji] * z2[i - ji]
z3[i] = z3[i] + arpara[ji] * z3[i - ji]
z = np.array( [z1.tolist(), z2.tolist(), z3.tolist()] )
z = np.array([z1.tolist(), z2.tolist(), z3.tolist()])
self.xpred = z

59
pylot/core/pick/picker.py
View File

@@ -25,6 +25,7 @@ from pylot.core.pick.utils import getnoisewin, getsignalwin
from pylot.core.pick.charfuns import CharacteristicFunction
import warnings
class AutoPicker(object):
'''
Superclass of different, automated picking algorithms applied on a CF determined
@@ -87,7 +88,6 @@ class AutoPicker(object):
Tsmooth=self.getTsmooth(),
Pick1=self.getpick1())
def getTSNR(self):
return self.TSNR
@@ -152,14 +152,14 @@ class AICPicker(AutoPicker):
self.Pick = None
self.slope = None
self.SNR = None
#find NaN's
# find NaN's
nn = np.isnan(self.cf)
if len(nn) > 1:
self.cf[nn] = 0
#taper AIC-CF to get rid off side maxima
# taper AIC-CF to get rid off side maxima
tap = np.hanning(len(self.cf))
aic = tap * self.cf + max(abs(self.cf))
#smooth AIC-CF
# smooth AIC-CF
ismooth = int(round(self.Tsmooth / self.dt))
aicsmooth = np.zeros(len(aic))
if len(aic) < ismooth:
@@ -171,32 +171,32 @@ class AICPicker(AutoPicker):
ii1 = i - ismooth
aicsmooth[i] = aicsmooth[i - 1] + (aic[i] - aic[ii1]) / ismooth
else:
aicsmooth[i] = np.mean(aic[1 : i])
#remove offset
aicsmooth[i] = np.mean(aic[1: i])
# remove offset
offset = abs(min(aic) - min(aicsmooth))
aicsmooth = aicsmooth - offset
#get maximum of 1st derivative of AIC-CF (more stable!) as starting point
# get maximum of 1st derivative of AIC-CF (more stable!) as starting point
diffcf = np.diff(aicsmooth)
#find NaN's
# find NaN's
nn = np.isnan(diffcf)
if len(nn) > 1:
diffcf[nn] = 0
#taper CF to get rid off side maxima
# taper CF to get rid off side maxima
tap = np.hanning(len(diffcf))
diffcf = tap * diffcf * max(abs(aicsmooth))
icfmax = np.argmax(diffcf)
#find minimum in AIC-CF front of maximum
# find minimum in AIC-CF front of maximum
lpickwindow = int(round(self.PickWindow / self.dt))
for i in range(icfmax - 1, max([icfmax - lpickwindow, 2]), -1):
if aicsmooth[i - 1] >= aicsmooth[i]:
self.Pick = self.Tcf[i]
break
#if no minimum could be found:
#search in 1st derivative of AIC-CF
# if no minimum could be found:
# search in 1st derivative of AIC-CF
if self.Pick is None:
for i in range(icfmax -1, max([icfmax -lpickwindow, 2]), -1):
if diffcf[i -1] >= diffcf[i]:
for i in range(icfmax - 1, max([icfmax - lpickwindow, 2]), -1):
if diffcf[i - 1] >= diffcf[i]:
self.Pick = self.Tcf[i]
break
@@ -215,7 +215,7 @@ class AICPicker(AutoPicker):
max(abs(aic[inoise] - np.mean(aic[inoise])))
# calculate slope from CF after initial pick
# get slope window
tslope = self.TSNR[3] #slope determination window
tslope = self.TSNR[3] # slope determination window
islope = np.where((self.Tcf <= min([self.Pick + tslope, len(self.Data[0].data)])) \
& (self.Tcf >= self.Pick))
# find maximum within slope determination window
@@ -237,7 +237,7 @@ class AICPicker(AutoPicker):
raw_input()
plt.close(p)
return
islope = islope[0][0 :imax]
islope = islope[0][0:imax]
dataslope = self.Data[0].data[islope]
# calculate slope as polynomal fit of order 1
xslope = np.arange(0, len(dataslope), 1)
@@ -258,7 +258,7 @@ class AICPicker(AutoPicker):
p1, = plt.plot(self.Tcf, x / max(x), 'k')
p2, = plt.plot(self.Tcf, aicsmooth / max(aicsmooth), 'r')
if self.Pick is not None:
p3, = plt.plot([self.Pick, self.Pick], [-0.1 , 0.5], 'b', linewidth=2)
p3, = plt.plot([self.Pick, self.Pick], [-0.1, 0.5], 'b', linewidth=2)
plt.legend([p1, p2, p3], ['(HOS-/AR-) Data', 'Smoothed AIC-CF', 'AIC-Pick'])
else:
plt.legend([p1, p2], ['(HOS-/AR-) Data', 'Smoothed AIC-CF'])
@@ -273,7 +273,8 @@ class AICPicker(AutoPicker):
p13, = plt.plot(self.Tcf[isignal], self.Data[0].data[isignal], 'r')
p14, = plt.plot(self.Tcf[islope], dataslope, 'g--')
p15, = plt.plot(self.Tcf[islope], datafit, 'g', linewidth=2)
plt.legend([p11, p12, p13, p14, p15], ['Data', 'Noise Window', 'Signal Window', 'Slope Window', 'Slope'],
plt.legend([p11, p12, p13, p14, p15],
['Data', 'Noise Window', 'Signal Window', 'Slope Window', 'Slope'],
loc='best')
plt.title('Station %s, SNR=%7.2f, Slope= %12.2f counts/s' % (self.Data[0].stats.station,
self.SNR, self.slope))
@@ -303,7 +304,7 @@ class PragPicker(AutoPicker):
self.SNR = None
self.slope = None
pickflag = 0
#smooth CF
# smooth CF
ismooth = int(round(self.Tsmooth / self.dt))
cfsmooth = np.zeros(len(self.cf))
if len(self.cf) < ismooth:
@@ -315,28 +316,28 @@ class PragPicker(AutoPicker):
ii1 = i - ismooth
cfsmooth[i] = cfsmooth[i - 1] + (self.cf[i] - self.cf[ii1]) / ismooth
else:
cfsmooth[i] = np.mean(self.cf[1 : i])
cfsmooth[i] = np.mean(self.cf[1: i])
#select picking window
#which is centered around tpick1
# select picking window
# which is centered around tpick1
ipick = np.where((self.Tcf >= self.getpick1() - self.PickWindow / 2) \
& (self.Tcf <= self.getpick1() + self.PickWindow / 2))
cfipick = self.cf[ipick] - np.mean(self.cf[ipick])
Tcfpick = self.Tcf[ipick]
cfsmoothipick = cfsmooth[ipick]- np.mean(self.cf[ipick])
cfsmoothipick = cfsmooth[ipick] - np.mean(self.cf[ipick])
ipick1 = np.argmin(abs(self.Tcf - self.getpick1()))
cfpick1 = 2 * self.cf[ipick1]
#check trend of CF, i.e. differences of CF and adjust aus regarding this trend
#prominent trend: decrease aus
#flat: use given aus
# check trend of CF, i.e. differences of CF and adjust aus regarding this trend
# prominent trend: decrease aus
# flat: use given aus
cfdiff = np.diff(cfipick)
i0diff = np.where(cfdiff > 0)
cfdiff = cfdiff[i0diff]
minaus = min(cfdiff * (1 + self.aus))
aus1 = max([minaus, self.aus])
#at first we look to the right until the end of the pick window is reached
# at first we look to the right until the end of the pick window is reached
flagpick_r = 0
flagpick_l = 0
cfpick_r = 0
@@ -380,8 +381,8 @@ class PragPicker(AutoPicker):
if self.getiplot() > 1:
p = plt.figure(self.getiplot())
p1, = plt.plot(Tcfpick,cfipick, 'k')
p2, = plt.plot(Tcfpick,cfsmoothipick, 'r')
p1, = plt.plot(Tcfpick, cfipick, 'k')
p2, = plt.plot(Tcfpick, cfsmoothipick, 'r')
if pickflag > 0:
p3, = plt.plot([self.Pick, self.Pick], [min(cfipick), max(cfipick)], 'b', linewidth=2)
plt.legend([p1, p2, p3], ['CF', 'Smoothed CF', 'Pick'])

307
pylot/core/pick/run_makeCF.py
View File

@@ -1,307 +0,0 @@
#!/usr/bin/python
# -*- coding: utf-8 -*-
"""
Script to run autoPyLoT-script "run_makeCF.py".
Only for test purposes!
"""
from obspy.core import read
import matplotlib.pyplot as plt
import numpy as np
from pylot.core.pick.charfuns import CharacteristicFunction
from pylot.core.pick.picker import AutoPicker
from pylot.core.pick.utils import *
import glob
import argparse
def run_makeCF(project, database, event, iplot, station=None):
#parameters for CF calculation
t2 = 7 #length of moving window for HOS calculation [sec]
p = 4 #order of HOS
cuttimes = [10, 50] #start and end time for CF calculation
bpz = [2, 30] #corner frequencies of bandpass filter, vertical component
bph = [2, 15] #corner frequencies of bandpass filter, horizontal components
tdetz= 1.2 #length of AR-determination window [sec], vertical component
tdeth= 0.8 #length of AR-determination window [sec], horizontal components
tpredz = 0.4 #length of AR-prediction window [sec], vertical component
tpredh = 0.4 #length of AR-prediction window [sec], horizontal components
addnoise = 0.001 #add noise to seismogram for stable AR prediction
arzorder = 2 #chosen order of AR process, vertical component
arhorder = 4 #chosen order of AR process, horizontal components
TSNRhos = [5, 0.5, 1, .6] #window lengths [s] for calculating SNR for earliest/latest pick and quality assessment
#from HOS-CF [noise window, safety gap, signal window, slope determination window]
TSNRarz = [5, 0.5, 1, 1.0] #window lengths [s] for calculating SNR for earliest/lates pick and quality assessment
#from ARZ-CF
#get waveform data
if station:
dpz = '/DATA/%s/EVENT_DATA/LOCAL/%s/%s/%s*HZ.msd' % (project, database, event, station)
dpe = '/DATA/%s/EVENT_DATA/LOCAL/%s/%s/%s*HE.msd' % (project, database, event, station)
dpn = '/DATA/%s/EVENT_DATA/LOCAL/%s/%s/%s*HN.msd' % (project, database, event, station)
#dpz = '/DATA/%s/EVENT_DATA/LOCAL/%s/%s/%s*_z.gse' % (project, database, event, station)
#dpe = '/DATA/%s/EVENT_DATA/LOCAL/%s/%s/%s*_e.gse' % (project, database, event, station)
#dpn = '/DATA/%s/EVENT_DATA/LOCAL/%s/%s/%s*_n.gse' % (project, database, event, station)
else:
dpz = '/DATA/%s/EVENT_DATA/LOCAL/%s/%s/*HZ.msd' % (project, database, event)
dpe = '/DATA/%s/EVENT_DATA/LOCAL/%s/%s/*HE.msd' % (project, database, event)
dpn = '/DATA/%s/EVENT_DATA/LOCAL/%s/%s/*HN.msd' % (project, database, event)
wfzfiles = glob.glob(dpz)
wfefiles = glob.glob(dpe)
wfnfiles = glob.glob(dpn)
if wfzfiles:
for i in range(len(wfzfiles)):
print 'Vertical component data found ...'
print wfzfiles[i]
st = read('%s' % wfzfiles[i])
st_copy = st.copy()
#filter and taper data
tr_filt = st[0].copy()
tr_filt.filter('bandpass', freqmin=bpz[0], freqmax=bpz[1], zerophase=False)
tr_filt.taper(max_percentage=0.05, type='hann')
st_copy[0].data = tr_filt.data
##############################################################
#calculate HOS-CF using subclass HOScf of class CharacteristicFunction
hoscf = HOScf(st_copy, cuttimes, t2, p) #instance of HOScf
##############################################################
#calculate AIC-HOS-CF using subclass AICcf of class CharacteristicFunction
#class needs stream object => build it
tr_aic = tr_filt.copy()
tr_aic.data = hoscf.getCF()
st_copy[0].data = tr_aic.data
aiccf = AICcf(st_copy, cuttimes) #instance of AICcf
##############################################################
#get prelimenary onset time from AIC-HOS-CF using subclass AICPicker of class AutoPicking
aicpick = AICPicker(aiccf, TSNRhos, 3, 10, None, 0.1)
##############################################################
#get refined onset time from HOS-CF using class Picker
hospick = PragPicker(hoscf, TSNRhos, 2, 10, 0.001, 0.2, aicpick.getpick())
#############################################################
#get earliest and latest possible picks
st_copy[0].data = tr_filt.data
[lpickhos, epickhos, pickerrhos] = earllatepicker(st_copy, 1.5, TSNRhos, hospick.getpick(), 10)
#############################################################
#get SNR
[SNR, SNRdB] = getSNR(st_copy, TSNRhos, hospick.getpick())
print 'SNR:', SNR, 'SNR[dB]:', SNRdB
##########################################################
#get first motion of onset
hosfm = fmpicker(st, st_copy, 0.2, hospick.getpick(), 11)
##############################################################
#calculate ARZ-CF using subclass ARZcf of class CharcteristicFunction
arzcf = ARZcf(st, cuttimes, tpredz, arzorder, tdetz, addnoise) #instance of ARZcf
##############################################################
#calculate AIC-ARZ-CF using subclass AICcf of class CharacteristicFunction
#class needs stream object => build it
tr_arzaic = tr_filt.copy()
tr_arzaic.data = arzcf.getCF()
st_copy[0].data = tr_arzaic.data
araiccf = AICcf(st_copy, cuttimes, tpredz, 0, tdetz) #instance of AICcf
##############################################################
#get onset time from AIC-ARZ-CF using subclass AICPicker of class AutoPicking
aicarzpick = AICPicker(araiccf, TSNRarz, 2, 10, None, 0.1)
##############################################################
#get refined onset time from ARZ-CF using class Picker
arzpick = PragPicker(arzcf, TSNRarz, 2.0, 10, 0.1, 0.05, aicarzpick.getpick())
#get earliest and latest possible picks
st_copy[0].data = tr_filt.data
[lpickarz, epickarz, pickerrarz] = earllatepicker(st_copy, 1.5, TSNRarz, arzpick.getpick(), 10)
elif not wfzfiles:
print 'No vertical component data found!'
if wfefiles and wfnfiles:
for i in range(len(wfefiles)):
print 'Horizontal component data found ...'
print wfefiles[i]
print wfnfiles[i]
#merge streams
H = read('%s' % wfefiles[i])
H += read('%s' % wfnfiles[i])
H_copy = H.copy()
#filter and taper data
trH1_filt = H[0].copy()
trH2_filt = H[1].copy()
trH1_filt.filter('bandpass', freqmin=bph[0], freqmax=bph[1], zerophase=False)
trH2_filt.filter('bandpass', freqmin=bph[0], freqmax=bph[1], zerophase=False)
trH1_filt.taper(max_percentage=0.05, type='hann')
trH2_filt.taper(max_percentage=0.05, type='hann')
H_copy[0].data = trH1_filt.data
H_copy[1].data = trH2_filt.data
##############################################################
#calculate ARH-CF using subclass ARHcf of class CharcteristicFunction
arhcf = ARHcf(H_copy, cuttimes, tpredh, arhorder, tdeth, addnoise) #instance of ARHcf
##############################################################
#calculate AIC-ARH-CF using subclass AICcf of class CharacteristicFunction
#class needs stream object => build it
tr_arhaic = trH1_filt.copy()
tr_arhaic.data = arhcf.getCF()
H_copy[0].data = tr_arhaic.data
#calculate ARH-AIC-CF
arhaiccf = AICcf(H_copy, cuttimes, tpredh, 0, tdeth) #instance of AICcf
##############################################################
#get onset time from AIC-ARH-CF using subclass AICPicker of class AutoPicking
aicarhpick = AICPicker(arhaiccf, TSNRarz, 4, 10, None, 0.1)
###############################################################
#get refined onset time from ARH-CF using class Picker
arhpick = PragPicker(arhcf, TSNRarz, 2.5, 10, 0.1, 0.05, aicarhpick.getpick())
#get earliest and latest possible picks
H_copy[0].data = trH1_filt.data
[lpickarh1, epickarh1, pickerrarh1] = earllatepicker(H_copy, 1.5, TSNRarz, arhpick.getpick(), 10)
H_copy[0].data = trH2_filt.data
[lpickarh2, epickarh2, pickerrarh2] = earllatepicker(H_copy, 1.5, TSNRarz, arhpick.getpick(), 10)
#get earliest pick of both earliest possible picks
epick = [epickarh1, epickarh2]
lpick = [lpickarh1, lpickarh2]
pickerr = [pickerrarh1, pickerrarh2]
ipick =np.argmin([epickarh1, epickarh2])
epickarh = epick[ipick]
lpickarh = lpick[ipick]
pickerrarh = pickerr[ipick]
#create stream with 3 traces
#merge streams
AllC = read('%s' % wfefiles[i])
AllC += read('%s' % wfnfiles[i])
AllC += read('%s' % wfzfiles[i])
#filter and taper data
All1_filt = AllC[0].copy()
All2_filt = AllC[1].copy()
All3_filt = AllC[2].copy()
All1_filt.filter('bandpass', freqmin=bph[0], freqmax=bph[1], zerophase=False)
All2_filt.filter('bandpass', freqmin=bph[0], freqmax=bph[1], zerophase=False)
All3_filt.filter('bandpass', freqmin=bpz[0], freqmax=bpz[1], zerophase=False)
All1_filt.taper(max_percentage=0.05, type='hann')
All2_filt.taper(max_percentage=0.05, type='hann')
All3_filt.taper(max_percentage=0.05, type='hann')
AllC[0].data = All1_filt.data
AllC[1].data = All2_filt.data
AllC[2].data = All3_filt.data
#calculate AR3C-CF using subclass AR3Ccf of class CharacteristicFunction
ar3ccf = AR3Ccf(AllC, cuttimes, tpredz, arhorder, tdetz, addnoise) #instance of AR3Ccf
##############################################################
if iplot:
#plot vertical trace
plt.figure()
tr = st[0]
tdata = np.arange(0, tr.stats.npts / tr.stats.sampling_rate, tr.stats.delta)
p1, = plt.plot(tdata, tr_filt.data/max(tr_filt.data), 'k')
p2, = plt.plot(hoscf.getTimeArray(), hoscf.getCF() / max(hoscf.getCF()), 'r')
p3, = plt.plot(aiccf.getTimeArray(), aiccf.getCF()/max(aiccf.getCF()), 'b')
p4, = plt.plot(arzcf.getTimeArray(), arzcf.getCF()/max(arzcf.getCF()), 'g')
p5, = plt.plot(araiccf.getTimeArray(), araiccf.getCF()/max(araiccf.getCF()), 'y')
plt.plot([aicpick.getpick(), aicpick.getpick()], [-1, 1], 'b--')
plt.plot([aicpick.getpick()-0.5, aicpick.getpick()+0.5], [1, 1], 'b')
plt.plot([aicpick.getpick()-0.5, aicpick.getpick()+0.5], [-1, -1], 'b')
plt.plot([hospick.getpick(), hospick.getpick()], [-1.3, 1.3], 'r', linewidth=2)
plt.plot([hospick.getpick()-0.5, hospick.getpick()+0.5], [1.3, 1.3], 'r')
plt.plot([hospick.getpick()-0.5, hospick.getpick()+0.5], [-1.3, -1.3], 'r')
plt.plot([lpickhos, lpickhos], [-1.1, 1.1], 'r--')
plt.plot([epickhos, epickhos], [-1.1, 1.1], 'r--')
plt.plot([aicarzpick.getpick(), aicarzpick.getpick()], [-1.2, 1.2], 'y', linewidth=2)
plt.plot([aicarzpick.getpick()-0.5, aicarzpick.getpick()+0.5], [1.2, 1.2], 'y')
plt.plot([aicarzpick.getpick()-0.5, aicarzpick.getpick()+0.5], [-1.2, -1.2], 'y')
plt.plot([arzpick.getpick(), arzpick.getpick()], [-1.4, 1.4], 'g', linewidth=2)
plt.plot([arzpick.getpick()-0.5, arzpick.getpick()+0.5], [1.4, 1.4], 'g')
plt.plot([arzpick.getpick()-0.5, arzpick.getpick()+0.5], [-1.4, -1.4], 'g')
plt.plot([lpickarz, lpickarz], [-1.2, 1.2], 'g--')
plt.plot([epickarz, epickarz], [-1.2, 1.2], 'g--')
plt.yticks([])
plt.ylim([-1.5, 1.5])
plt.xlabel('Time [s]')
plt.ylabel('Normalized Counts')
plt.title('%s, %s, CF-SNR=%7.2f, CF-Slope=%12.2f' % (tr.stats.station,
tr.stats.channel, aicpick.getSNR(), aicpick.getSlope()))
plt.suptitle(tr.stats.starttime)
plt.legend([p1, p2, p3, p4, p5], ['Data', 'HOS-CF', 'HOSAIC-CF', 'ARZ-CF', 'ARZAIC-CF'])
#plot horizontal traces
plt.figure(2)
plt.subplot(2,1,1)
tsteph = tpredh / 4
th1data = np.arange(0, trH1_filt.stats.npts / trH1_filt.stats.sampling_rate, trH1_filt.stats.delta)
th2data = np.arange(0, trH2_filt.stats.npts / trH2_filt.stats.sampling_rate, trH2_filt.stats.delta)
tarhcf = np.arange(0, len(arhcf.getCF()) * tsteph, tsteph) + cuttimes[0] + tdeth +tpredh
p21, = plt.plot(th1data, trH1_filt.data/max(trH1_filt.data), 'k')
p22, = plt.plot(arhcf.getTimeArray(), arhcf.getCF()/max(arhcf.getCF()), 'r')
p23, = plt.plot(arhaiccf.getTimeArray(), arhaiccf.getCF()/max(arhaiccf.getCF()))
plt.plot([aicarhpick.getpick(), aicarhpick.getpick()], [-1, 1], 'b')
plt.plot([aicarhpick.getpick()-0.5, aicarhpick.getpick()+0.5], [1, 1], 'b')
plt.plot([aicarhpick.getpick()-0.5, aicarhpick.getpick()+0.5], [-1, -1], 'b')
plt.plot([arhpick.getpick(), arhpick.getpick()], [-1, 1], 'r')
plt.plot([arhpick.getpick()-0.5, arhpick.getpick()+0.5], [1, 1], 'r')
plt.plot([arhpick.getpick()-0.5, arhpick.getpick()+0.5], [-1, -1], 'r')
plt.plot([lpickarh, lpickarh], [-0.8, 0.8], 'r--')
plt.plot([epickarh, epickarh], [-0.8, 0.8], 'r--')
plt.plot([arhpick.getpick() + pickerrarh, arhpick.getpick() + pickerrarh], [-0.2, 0.2], 'r--')
plt.plot([arhpick.getpick() - pickerrarh, arhpick.getpick() - pickerrarh], [-0.2, 0.2], 'r--')
plt.yticks([])
plt.ylim([-1.5, 1.5])
plt.ylabel('Normalized Counts')
plt.title([trH1_filt.stats.station, trH1_filt.stats.channel])
plt.suptitle(trH1_filt.stats.starttime)
plt.legend([p21, p22, p23], ['Data', 'ARH-CF', 'ARHAIC-CF'])
plt.subplot(2,1,2)
plt.plot(th2data, trH2_filt.data/max(trH2_filt.data), 'k')
plt.plot(arhcf.getTimeArray(), arhcf.getCF()/max(arhcf.getCF()), 'r')
plt.plot(arhaiccf.getTimeArray(), arhaiccf.getCF()/max(arhaiccf.getCF()))
plt.plot([aicarhpick.getpick(), aicarhpick.getpick()], [-1, 1], 'b')
plt.plot([aicarhpick.getpick()-0.5, aicarhpick.getpick()+0.5], [1, 1], 'b')
plt.plot([aicarhpick.getpick()-0.5, aicarhpick.getpick()+0.5], [-1, -1], 'b')
plt.plot([arhpick.getpick(), arhpick.getpick()], [-1, 1], 'r')
plt.plot([arhpick.getpick()-0.5, arhpick.getpick()+0.5], [1, 1], 'r')
plt.plot([arhpick.getpick()-0.5, arhpick.getpick()+0.5], [-1, -1], 'r')
plt.plot([lpickarh, lpickarh], [-0.8, 0.8], 'r--')
plt.plot([epickarh, epickarh], [-0.8, 0.8], 'r--')
plt.plot([arhpick.getpick() + pickerrarh, arhpick.getpick() + pickerrarh], [-0.2, 0.2], 'r--')
plt.plot([arhpick.getpick() - pickerrarh, arhpick.getpick() - pickerrarh], [-0.2, 0.2], 'r--')
plt.title([trH2_filt.stats.station, trH2_filt.stats.channel])
plt.yticks([])
plt.ylim([-1.5, 1.5])
plt.xlabel('Time [s]')
plt.ylabel('Normalized Counts')
#plot 3-component window
plt.figure(3)
plt.subplot(3,1,1)
p31, = plt.plot(tdata, tr_filt.data/max(tr_filt.data), 'k')
p32, = plt.plot(ar3ccf.getTimeArray(), ar3ccf.getCF()/max(ar3ccf.getCF()), 'r')
plt.plot([arhpick.getpick(), arhpick.getpick()], [-1, 1], 'b')
plt.plot([arhpick.getpick()-0.5, arhpick.getpick()+0.5], [-1, -1], 'b')
plt.plot([arhpick.getpick()-0.5, arhpick.getpick()+0.5], [1, 1], 'b')
plt.yticks([])
plt.xticks([])
plt.ylabel('Normalized Counts')
plt.title([tr.stats.station, tr.stats.channel])
plt.suptitle(trH1_filt.stats.starttime)
plt.legend([p31, p32], ['Data', 'AR3C-CF'])
plt.subplot(3,1,2)
plt.plot(th1data, trH1_filt.data/max(trH1_filt.data), 'k')
plt.plot(ar3ccf.getTimeArray(), ar3ccf.getCF()/max(ar3ccf.getCF()), 'r')
plt.plot([arhpick.getpick(), arhpick.getpick()], [-1, 1], 'b')
plt.plot([arhpick.getpick()-0.5, arhpick.getpick()+0.5], [-1, -1], 'b')
plt.plot([arhpick.getpick()-0.5, arhpick.getpick()+0.5], [1, 1], 'b')
plt.yticks([])
plt.xticks([])
plt.ylabel('Normalized Counts')
plt.title([trH1_filt.stats.station, trH1_filt.stats.channel])
plt.subplot(3,1,3)
plt.plot(th2data, trH2_filt.data/max(trH2_filt.data), 'k')
plt.plot(ar3ccf.getTimeArray(), ar3ccf.getCF()/max(ar3ccf.getCF()), 'r')
plt.plot([arhpick.getpick(), arhpick.getpick()], [-1, 1], 'b')
plt.plot([arhpick.getpick()-0.5, arhpick.getpick()+0.5], [-1, -1], 'b')
plt.plot([arhpick.getpick()-0.5, arhpick.getpick()+0.5], [1, 1], 'b')
plt.yticks([])
plt.ylabel('Normalized Counts')
plt.title([trH2_filt.stats.station, trH2_filt.stats.channel])
plt.xlabel('Time [s]')
plt.show()
raw_input()
plt.close()
parser = argparse.ArgumentParser()
parser.add_argument('--project', type=str, help='project name (e.g. Insheim)')
parser.add_argument('--database', type=str, help='event data base (e.g. 2014.09_Insheim)')
parser.add_argument('--event', type=str, help='event ID (e.g. e0010.015.14)')
parser.add_argument('--iplot', help='anything, if set, figure occurs')
parser.add_argument('--station', type=str, help='Station ID (e.g. INS3) (optional)')
args = parser.parse_args()
run_makeCF(args.project, args.database, args.event, args.iplot, args.station)

228
pylot/core/pick/utils.py
View File

@@ -15,7 +15,7 @@ from obspy.core import Stream, UTCDateTime
import warnings
def earllatepicker(X, nfac, TSNR, Pick1, iplot=None, stealthMode = False):
def earllatepicker(X, nfac, TSNR, Pick1, iplot=None, stealthMode=False):
'''
Function to derive earliest and latest possible pick after Diehl & Kissling (2009)
as reasonable uncertainties. Latest possible pick is based on noise level,
@@ -70,7 +70,8 @@ def earllatepicker(X, nfac, TSNR, Pick1, iplot=None, stealthMode = False):
# get earliest possible pick
EPick = np.nan; count = 0
EPick = np.nan;
count = 0
pis = isignal
# if EPick stays NaN the signal window size will be doubled
@@ -78,10 +79,10 @@ def earllatepicker(X, nfac, TSNR, Pick1, iplot=None, stealthMode = False):
if count > 0:
if stealthMode is False:
print("\nearllatepicker: Doubled signal window size %s time(s) "
"because of NaN for earliest pick." %count)
"because of NaN for earliest pick." % count)
isigDoubleWinStart = pis[-1] + 1
isignalDoubleWin = np.arange(isigDoubleWinStart,
isigDoubleWinStart + len(pis))
isigDoubleWinStart + len(pis))
if (isigDoubleWinStart + len(pis)) < X[0].data.size:
pis = np.concatenate((pis, isignalDoubleWin))
else:
@@ -92,8 +93,7 @@ def earllatepicker(X, nfac, TSNR, Pick1, iplot=None, stealthMode = False):
zc = crossings_nonzero_all(x[pis] - x[pis].mean())
# calculate mean half period T0 of signal as the average of the
T0 = np.mean(np.diff(zc)) * X[0].stats.delta # this is half wave length!
EPick = Pick1 - T0 # half wavelength as suggested by Diehl et al.
EPick = Pick1 - T0 # half wavelength as suggested by Diehl et al.
# get symmetric pick error as mean from earliest and latest possible pick
# by weighting latest possible pick two times earliest possible pick
@@ -395,7 +395,7 @@ def getnoisewin(t, t1, tnoise, tgap):
# get noise window
inoise, = np.where((t <= max([t1 - tgap, 0])) \
& (t >= max([t1 - tnoise - tgap, 0])))
& (t >= max([t1 - tnoise - tgap, 0])))
if np.size(inoise) < 1:
print ("getnoisewin: Empty array inoise, check noise window!")
@@ -419,7 +419,7 @@ def getsignalwin(t, t1, tsignal):
# get signal window
isignal, = np.where((t <= min([t1 + tsignal, len(t)])) \
& (t >= t1))
& (t >= t1))
if np.size(isignal) < 1:
print ("getsignalwin: Empty array isignal, check signal window!")
@@ -460,7 +460,7 @@ def getResolutionWindow(snr):
else:
time_resolution = res_wins['HRW']
return time_resolution/2
return time_resolution / 2
def wadaticheck(pickdic, dttolerance, iplot):
@@ -488,17 +488,16 @@ def wadaticheck(pickdic, dttolerance, iplot):
SPtimes = []
for key in pickdic:
if pickdic[key]['P']['weight'] < 4 and pickdic[key]['S']['weight'] < 4:
# calculate S-P time
spt = pickdic[key]['S']['mpp'] - pickdic[key]['P']['mpp']
# add S-P time to dictionary
pickdic[key]['SPt'] = spt
# add P onsets and corresponding S-P times to list
UTCPpick = UTCDateTime(pickdic[key]['P']['mpp'])
UTCSpick = UTCDateTime(pickdic[key]['S']['mpp'])
Ppicks.append(UTCPpick.timestamp)
Spicks.append(UTCSpick.timestamp)
SPtimes.append(spt)
# calculate S-P time
spt = pickdic[key]['S']['mpp'] - pickdic[key]['P']['mpp']
# add S-P time to dictionary
pickdic[key]['SPt'] = spt
# add P onsets and corresponding S-P times to list
UTCPpick = UTCDateTime(pickdic[key]['P']['mpp'])
UTCSpick = UTCDateTime(pickdic[key]['S']['mpp'])
Ppicks.append(UTCPpick.timestamp)
Spicks.append(UTCSpick.timestamp)
SPtimes.append(spt)
if len(SPtimes) >= 3:
# calculate slope
@@ -530,7 +529,7 @@ def wadaticheck(pickdic, dttolerance, iplot):
ibad += 1
else:
marker = 'goodWadatiCheck'
checkedPpick = UTCDateTime(pickdic[key]['P']['mpp'])
checkedPpick = UTCDateTime(pickdic[key]['P']['mpp'])
checkedPpicks.append(checkedPpick.timestamp)
checkedSpick = UTCDateTime(pickdic[key]['S']['mpp'])
checkedSpicks.append(checkedSpick.timestamp)
@@ -642,7 +641,7 @@ def checksignallength(X, pick, TSNR, minsiglength, nfac, minpercent, iplot):
# calculate minimum adjusted signal level
minsiglevel = max(rms[inoise]) * nfac
# minimum adjusted number of samples over minimum signal level
minnum = len(isignal) * minpercent/100
minnum = len(isignal) * minpercent / 100
# get number of samples above minimum adjusted signal level
numoverthr = len(np.where(rms[isignal] >= minsiglevel)[0])
@@ -657,10 +656,10 @@ def checksignallength(X, pick, TSNR, minsiglength, nfac, minpercent, iplot):
if iplot == 2:
plt.figure(iplot)
p1, = plt.plot(t,rms, 'k')
p1, = plt.plot(t, rms, 'k')
p2, = plt.plot(t[inoise], rms[inoise], 'c')
p3, = plt.plot(t[isignal],rms[isignal], 'r')
p4, = plt.plot([t[isignal[0]], t[isignal[len(isignal)-1]]],
p3, = plt.plot(t[isignal], rms[isignal], 'r')
p4, = plt.plot([t[isignal[0]], t[isignal[len(isignal) - 1]]],
[minsiglevel, minsiglevel], 'g', linewidth=2)
p5, = plt.plot([pick, pick], [min(rms), max(rms)], 'b', linewidth=2)
plt.legend([p1, p2, p3, p4, p5], ['RMS Data', 'RMS Noise Window',
@@ -701,15 +700,15 @@ def checkPonsets(pickdic, dttolerance, iplot):
stations = []
for key in pickdic:
if pickdic[key]['P']['weight'] < 4:
# add P onsets to list
UTCPpick = UTCDateTime(pickdic[key]['P']['mpp'])
Ppicks.append(UTCPpick.timestamp)
stations.append(key)
# add P onsets to list
UTCPpick = UTCDateTime(pickdic[key]['P']['mpp'])
Ppicks.append(UTCPpick.timestamp)
stations.append(key)
# apply jackknife bootstrapping on variance of P onsets
print ("###############################################")
print ("checkPonsets: Apply jackknife bootstrapping on P-onset times ...")
[xjack,PHI_pseudo,PHI_sub] = jackknife(Ppicks, 'VAR', 1)
[xjack, PHI_pseudo, PHI_sub] = jackknife(Ppicks, 'VAR', 1)
# get pseudo variances smaller than average variances
# (times safety factor), these picks passed jackknife test
ij = np.where(PHI_pseudo <= 2 * xjack)
@@ -730,7 +729,7 @@ def checkPonsets(pickdic, dttolerance, iplot):
print ("checkPonsets: %d pick(s) deviate too much from median!" % len(ibad))
print ("checkPonsets: Skipped %d P pick(s) out of %d" % (len(badstations) \
+ len(badjkstations), len(stations)))
+ len(badjkstations), len(stations)))
goodmarker = 'goodPonsetcheck'
badmarker = 'badPonsetcheck'
@@ -881,10 +880,9 @@ def checkZ4S(X, pick, zfac, checkwin, iplot):
if len(ndat) == 0: # check for other components
ndat = X.select(component="1")
z = zdat[0].data
tz = np.arange(0, zdat[0].stats.npts / zdat[0].stats.sampling_rate,
zdat[0].stats.delta)
zdat[0].stats.delta)
# calculate RMS trace from vertical component
absz = np.sqrt(np.power(z, 2))
@@ -916,9 +914,9 @@ def checkZ4S(X, pick, zfac, checkwin, iplot):
if iplot > 1:
te = np.arange(0, edat[0].stats.npts / edat[0].stats.sampling_rate,
edat[0].stats.delta)
edat[0].stats.delta)
tn = np.arange(0, ndat[0].stats.npts / ndat[0].stats.sampling_rate,
ndat[0].stats.delta)
ndat[0].stats.delta)
plt.plot(tz, z / max(z), 'k')
plt.plot(tz[isignal], z[isignal] / max(z), 'r')
plt.plot(te, edat[0].data / max(edat[0].data) + 1, 'k')
@@ -960,65 +958,64 @@ def writephases(arrivals, fformat, filename):
:type: string
'''
if fformat == 'NLLoc':
print ("Writing phases to %s for NLLoc" % filename)
fid = open("%s" % filename, 'w')
# write header
fid.write('# EQEVENT: Label: EQ001 Loc: X 0.00 Y 0.00 Z 10.00 OT 0.00 \n')
for key in arrivals:
# P onsets
if arrivals[key]['P']:
fm = arrivals[key]['P']['fm']
if fm == None:
fm = '?'
onset = arrivals[key]['P']['mpp']
year = onset.year
month = onset.month
day = onset.day
hh = onset.hour
mm = onset.minute
ss = onset.second
ms = onset.microsecond
ss_ms = ss + ms / 1000000.0
if arrivals[key]['P']['weight'] < 4:
pweight = 1 # use pick
else:
pweight = 0 # do not use pick
fid.write('%s ? ? ? P %s %d%02d%02d %02d%02d %7.4f GAU 0 0 0 0 %d \n' % (key,
fm,
year,
month,
day,
hh,
mm,
ss_ms,
pweight))
# S onsets
if arrivals[key]['S']:
fm = '?'
onset = arrivals[key]['S']['mpp']
year = onset.year
month = onset.month
day = onset.day
hh = onset.hour
mm = onset.minute
ss = onset.second
ms = onset.microsecond
ss_ms = ss + ms / 1000000.0
if arrivals[key]['S']['weight'] < 4:
sweight = 1 # use pick
else:
sweight = 0 # do not use pick
fid.write('%s ? ? ? S %s %d%02d%02d %02d%02d %7.4f GAU 0 0 0 0 %d \n' % (key,
fm,
year,
month,
day,
hh,
mm,
ss_ms,
sweight))
# P onsets
if arrivals[key]['P']:
fm = arrivals[key]['P']['fm']
if fm == None:
fm = '?'
onset = arrivals[key]['P']['mpp']
year = onset.year
month = onset.month
day = onset.day
hh = onset.hour
mm = onset.minute
ss = onset.second
ms = onset.microsecond
ss_ms = ss + ms / 1000000.0
if arrivals[key]['P']['weight'] < 4:
pweight = 1 # use pick
else:
pweight = 0 # do not use pick
fid.write('%s ? ? ? P %s %d%02d%02d %02d%02d %7.4f GAU 0 0 0 0 %d \n' % (key,
fm,
year,
month,
day,
hh,
mm,
ss_ms,
pweight))
# S onsets
if arrivals[key]['S']:
fm = '?'
onset = arrivals[key]['S']['mpp']
year = onset.year
month = onset.month
day = onset.day
hh = onset.hour
mm = onset.minute
ss = onset.second
ms = onset.microsecond
ss_ms = ss + ms / 1000000.0
if arrivals[key]['S']['weight'] < 4:
sweight = 1 # use pick
else:
sweight = 0 # do not use pick
fid.write('%s ? ? ? S %s %d%02d%02d %02d%02d %7.4f GAU 0 0 0 0 %d \n' % (key,
fm,
year,
month,
day,
hh,
mm,
ss_ms,
sweight))
fid.close()
@@ -1043,9 +1040,9 @@ def writephases(arrivals, fformat, filename):
Ao = str('%7.2f' % Ao)
year = Ponset.year
if year >= 2000:
year = year -2000
year = year - 2000
else:
year = year - 1900
year = year - 1900
month = Ponset.month
day = Ponset.day
hh = Ponset.hour
@@ -1054,9 +1051,9 @@ def writephases(arrivals, fformat, filename):
ms = Ponset.microsecond
ss_ms = ss + ms / 1000000.0
if pweight < 2:
pstr = 'I'
pstr = 'I'
elif pweight >= 2:
pstr = 'E'
pstr = 'E'
if arrivals[key]['S']['weight'] < 4:
Sss = Sonset.second
Sms = Sonset.microsecond
@@ -1067,35 +1064,36 @@ def writephases(arrivals, fformat, filename):
elif sweight >= 2:
sstr = 'E'
fid.write('%s%sP%s%d %02d%02d%02d%02d%02d%5.2f %s%sS %d %s\n' % (key,
pstr,
fm,
pweight,
year,
month,
day,
hh,
mm,
ss_ms,
Sss_ms,
sstr,
sweight,
Ao))
pstr,
fm,
pweight,
year,
month,
day,
hh,
mm,
ss_ms,
Sss_ms,
sstr,
sweight,
Ao))
else:
fid.write('%s%sP%s%d %02d%02d%02d%02d%02d%5.2f %s\n' % (key,
pstr,
fm,
pweight,
year,
month,
day,
hh,
mm,
ss_ms,
Ao))
pstr,
fm,
pweight,
year,
month,
day,
hh,
mm,
ss_ms,
Ao))
fid.close()
if __name__ == '__main__':
import doctest
doctest.testmod()