def PSD_palm(t_para, tSeq, zInd, xNum, yNum, varSeq, segNum):
t_start = t_para[0]
t_end = t_para[1]
t_delta = t_para[2]
fs = 1 / t_delta # sampling frequency
t_num = int((t_end - t_start) / t_delta + 1)
t_seq = np.linspace(t_start, t_end, t_num)
PSD_list = []
for yInd in range(yNum):
for xInd in range(xNum):
vSeq = varSeq[:, zInd, yInd, xInd]
# interpolate
f = interp1d(tSeq, vSeq, kind='linear', fill_value='extrapolate')
v_seq = f(t_seq)
# detrend
deg_ = 1
polyFunc = np.poly1d(np.polyfit(t_seq, v_seq, deg=deg_))
tmp = v_seq - polyFunc(t_seq)
tmp = tmp - tmp.mean()
# bell tapering
tmp = funcs.window_weight(tmp)
# FFT
# omega_seq, tmp = PSD_omega(t_seq,tmp)
f_seq, tmp = scipy.signal.csd(tmp,
tmp,
fs,
nperseg=segNum,
noverlap=None)
PSD_list.append(tmp)
PSD_seq = np.average(np.array(PSD_list), axis=0)
return f_seq, PSD_seq
def PSD_ky_sowfa(tSeq, ySeq, xSeq, zInd, d, varSeq, segNum):
tNum = tSeq.size
xNum = xSeq.size
yNum = ySeq.size
pInd_start = xNum * yNum * zInd
pInd_end = xNum * yNum * (zInd + 1)
psd_list = []
for t in range(tNum):
v = varSeq[pInd_start:pInd_end, t]
# detrend
deg_ = 1
polyFunc = np.poly1d(np.polyfit(ySeq, v, deg=deg_))
v_ = v - polyFunc(ySeq)
# bell tapering
v_ = funcs.window_weight(v_)
# FFT
# omega_seq, tmp = PSD_omega(t_seq,tmp)
ky_seq, psd = scipy.signal.csd(v_,
v_,
2 * np.pi / d,
nperseg=segNum,
noverlap=None)
psd_list.append(psd)
psd_seq = np.average(np.array(psd_list), axis=0)
return ky_seq, psd_seq
def PSD_sowfa(t_para, tSeq, zInd, xNum, yNum, varSeq, segNum):
"""
Calculate power spectral density
INPUT
t_para: time parameters, tuple, e.g. (144000.0, 146400, 0.1)
tSeq: 1D array of times
zInd: height
xNum: length of xSeq
yNum: length of ySeq
varSeq: 2D array of the target variable, 1st dim point ID, 2nd dim time
segNum: length of the segment
OUTPUT
f_seq: frequency sequence
PSD_seq: PSD sequence
"""
t_start = t_para[0]
t_end = t_para[1]
t_delta = t_para[2]
fs = 1 / t_delta # sampling frequency
t_num = int((t_end - t_start) / t_delta + 1)
t_seq = np.linspace(t_start, t_end, t_num)
data = varSeq
tNum = tSeq.size
pInd_start = xNum * yNum * zInd
pInd_end = xNum * yNum * (zInd + 1)
# coors = data_org['coors'][xNum*yNum*zInd:xNum*yNum*(zInd+1)]
# pNum = coors.shape[0]
PSD_list = []
for p in range(pInd_start, pInd_end):
vSeq = varSeq[p]
# interpolate
f = interp1d(tSeq, vSeq, kind='linear', fill_value='extrapolate')
v_seq = f(t_seq)
# detrend
deg_ = 1
polyFunc = np.poly1d(np.polyfit(t_seq, v_seq, deg=deg_))
tmp = v_seq - polyFunc(t_seq)
tmp = tmp - tmp.mean()
# bell tapering
tmp = funcs.window_weight(tmp)
# FFT
# omega_seq, tmp = PSD_omega(t_seq,tmp)
f_seq, tmp = scipy.signal.csd(tmp,
tmp,
fs,
nperseg=segNum,
noverlap=None)
PSD_list.append(tmp)
PSD_seq = np.average(np.array(PSD_list), axis=0)
return f_seq, PSD_seq
def PSD_ky_palm(tSeq, ySeq, zInd, d, varSeq, segNum):
psd_list = []
for tInd in range(tSeq.size):
v = varSeq[tInd, zInd, :, 0]
# detrend
deg_ = 1
polyFunc = np.poly1d(np.polyfit(ySeq, v, deg=deg_))
v_ = v - polyFunc(ySeq)
# bell tapering
v_ = funcs.window_weight(v_)
ky_seq, psd = scipy.signal.csd(v_,
v_,
2 * np.pi / d,
nperseg=segNum,
noverlap=None)
psd_list.append(psd)
psd_seq = np.average(np.array(psd_list), axis=0)
return ky_seq, psd_seq
def PSD(t_para, tSeq, zInd, xNum, yNum, varSeq, segNum):
t_start = t_para[0]
t_end = t_para[1]
t_delta = t_para[2]
fs = 1 / t_delta # sampling frequency
t_num = int((t_end - t_start) / t_delta + 1)
t_seq = np.linspace(t_start, t_end, t_num)
data = varSeq
tNum = tSeq.size
pInd_start = xNum * yNum * zInd
pInd_end = xNum * yNum * (zInd + 1)
# coors = data_org['coors'][xNum*yNum*zInd:xNum*yNum*(zInd+1)]
# pNum = coors.shape[0]
PSD_list = []
for p in range(pInd_start, pInd_end):
vSeq = varSeq[p]
# interpolate
f = interp1d(tSeq, vSeq, kind='linear', fill_value='extrapolate')
v_seq = f(t_seq)
# detrend
deg_ = 1
polyFunc = np.poly1d(np.polyfit(t_seq, v_seq, deg=deg_))
tmp = v_seq - polyFunc(t_seq)
tmp = tmp - tmp.mean()
# bell tapering
tmp = funcs.window_weight(tmp)
# FFT
# omega_seq, tmp = PSD_omega(t_seq,tmp)
f_seq, tmp = scipy.signal.csd(tmp,
tmp,
fs,
nperseg=segNum,
noverlap=None)
PSD_list.append(tmp)
PSD_seq = np.average(np.array(PSD_list), axis=0)
return f_seq, PSD_seq
# coors = data_org['coors'][xNum*yNum*zInd:xNum*yNum*(zInd+1)]
# pNum = coors.shape[0]
PSD_list = []
for p in range(pInd_start, pInd_end):
vSeq = data[p][:, varD]
# interpolate
f = interp1d(tSeq, vSeq, fill_value='extrapolate')
v_seq = f(t_seq)
# detrend
deg_ = 1
polyFunc = np.poly1d(np.polyfit(t_seq, v_seq, deg=deg_))
tmp = v_seq - polyFunc(t_seq)
tmp = tmp - tmp.mean()
# bell tapering
tmp = funcs.window_weight(tmp)
# FFT
# omega_seq, tmp = PSD_omega(t_seq,tmp)
f_seq, tmp = scipy.signal.csd(tmp,
tmp,
fs,
nperseg=segNum,
noverlap=None)
PSD_list.append(tmp)
PSD_seq = np.average(np.array(PSD_list), axis=0)
plotDataList.append((f_seq, PSD_seq))
# plot
fig, ax = plt.subplots(figsize=(8, 5))
colors = plt.cm.jet(np.linspace(0, 1, zNum))