def log(pro, logname, savepng=True):
"""
Log processed data (*.csv) and echograms (*.png) in rapidkrill/log/.
Args:
pro (dict): processed data output from "process" routine.
logname (str ): directory name under which log results will be saved.
savepng (bool): True to save echogram images, False to skip it.
"""
# Load processed data variables
rawfiles = pro['rawfiles']
transect = pro['transect']
t120 = pro['t120']
r120 = pro['r120']
Sv120 = pro['Sv120']
Sv120sw = pro['Sv120sw']
t120r = pro['t120r']
nm120r = pro['nm120r']
lon120r = pro['lon120r']
lat120r = pro['lat120r']
sbline120r = pro['sbliner'][0, :]
NASC120swr = pro['NASC120swr'][0, :]
pc120swr = pro['pc120swr'][0, :]
# Build summary results
results = {
'Time': np.array(t120r[:-1], dtype=str),
'Longitude': np.round(lon120r[:-1], 5),
'Latitude': np.round(lat120r[:-1], 5),
'Transect': np.ones(len(t120r[:-1]), dtype=int) * transect,
'Miles': nm120r[:-1],
'Seabed': np.round(sbline120r[:-1], 1),
'NASC': np.round(NASC120swr[:-1], 2),
'% samples': np.round(pc120swr[:-1], 1)
}
results = pd.DataFrame(results,
columns=[
'Time', 'Longitude', 'Latitude', 'Transect',
'Miles', 'Seabed', 'NASC', '% samples'
])
# Create new log subdirectory
path = os.path.join(os.path.dirname(__file__), '..', 'log', logname, '')
if not os.path.exists(path):
os.makedirs(path)
# Write results in CSV log file
with open(path + logname + '.csv', 'a') as f:
results.to_csv(path + logname + '.csv',
index=False,
mode='a',
header=f.tell() == 0)
# save png image
if savepng:
# set figure
plt.close()
plt.figure(figsize=(8, 8))
plt.subplots_adjust(left=0.066,
right=1.055,
bottom=0.065,
top=0.985,
wspace=0,
hspace=0.05)
plt.rcParams.update({'font.size': 9, 'lines.linewidth': 1})
# plot raw echogram
plt.subplot(211).invert_yaxis()
im = plt.pcolormesh(t120,
r120,
Sv120,
vmin=-80,
vmax=-50,
cmap=cmaps().ek500)
plt.colorbar(im).set_label('Sv raw (dB re 1m$^{-1}$)')
plt.gca().set_ylim(270, 0)
plt.gca().set_ylabel('Depth (m)')
plt.gca().set_xlim(t120r[0], t120r[-1])
plt.gca().set_xticks(t120r[[0, -1]])
plt.tick_params(labelright=False, labelbottom=False)
# plot processed echogram
ax = plt.subplot(212)
ax = [ax, ax.twinx()]
im = ax[0].pcolormesh(t120,
r120,
Sv120sw,
vmin=-80,
vmax=-50,
cmap=cmaps().ek500)
plt.colorbar(im).set_label('Sv pro (dB re 1m$^{-1}$)')
ax[0].invert_yaxis()
ax[0].set_ylim(270, 0)
ax[0].set_ylabel('Depth (m)')
# overlay distance/NASC info
for t, nm, NASC in zip(t120r[:-1], nm120r[:-1], NASC120swr[:-1]):
ax[1].plot([t, t], [0, 1], color=[0, .8, 0], linewidth=2)
ax[1].text(t,
.95,
' ' + str(transect) + ': ' + str(round(nm, 2)),
fontweight='bold',
color=[0, .8, 0])
ax[1].text(t,
.02,
' ' + str(round(NASC, 2)),
fontweight='bold',
color=[1, 0, 0])
ax[1].set_ylim(0, 1)
ax[1].set_xlim(t120r[0], t120r[-1])
ax[1].set_xticks(t120r[[0, -1]])
ax[1].tick_params(labelright=False)
ax[1].xaxis.set_major_formatter(mdates.DateFormatter('%d%b-%H:%M:%S'))
# save figure
pf = rawfiles[0].split('-')[0]
fn = pd.to_datetime(str(t120[0])).strftime(pf + '-D%Y%m%d-T%H%M%S')
plt.savefig(path + fn + '.png', figsize=(8, 8), dpi=100)
# =============================================================================
# Seabed masking
# =============================================================================
print('Masking seabed...')
r0, r1, roff, thr = 10, 1000, 10, (-40, -60) # (m, m, m, dB)
mask38sb = maskSB.maxSv(Sv38smooth, r38, r0=r0, r1=r1, roff=roff, thr=thr)
Sv38sboff = Sv38.copy()
Sv38sboff[mask38sb] = np.nan
# =============================================================================
# plot results
# =============================================================================
print('Displaying results...')
plt.figure(figsize=(8, 6))
c = cmaps()
plt.subplot(211).invert_yaxis()
plt.pcolormesh(t38, r38, Sv38, vmin=-80, vmax=-50, cmap=c.ek500)
plt.colorbar().set_label('dB')
plt.ylabel('Depth (m)')
plt.tick_params(labelbottom=False)
plt.title('Sv 38 kHz')
plt.subplot(212).invert_yaxis()
plt.pcolormesh(t38, r38, Sv38sboff, vmin=-80, vmax=-50, cmap=c.ek500)
plt.colorbar().set_label('dB')
plt.ylabel('Depth (m)')
plt.xlabel('Time (dd HH:MM)')
plt.title('Sv 38 kHz - seabed masked')
m120sn = mSN.derobertis(Sv120, bn120, thr=12)
# In this case, Sv samples 12 dB above the background noise estimation, or
# lower, will be masked.
# -----------------------------------------------------------------------------
# use the mask to turn the Sv samples into "empty water" (-999)
Sv120clean[m120sn] = -999
#------------------------------------------------------------------------------
# Figures
plt.figure(figsize=(8, 7))
# Sv original
plt.subplot(311).invert_yaxis()
plt.pcolormesh(p120, r120, Sv120, vmin=-80, vmax=-50, cmap=cmaps().ek500)
plt.tick_params(labelbottom=False)
plt.colorbar()
plt.title('Sv original')
# Background noise estimation
plt.subplot(312).invert_yaxis()
plt.pcolormesh(p120, r120, bn120, cmap=cmaps().ek500)
plt.tick_params(labelbottom=False)
plt.ylabel('Depth (m)')
plt.colorbar()
plt.title('Background noise')
# Sv after correcting for background noise
plt.subplot(313).invert_yaxis()
plt.pcolormesh(p120, r120, Sv120clean, vmin=-80, vmax=-50, cmap=cmaps().ek500)
Sv120cvv = tf.log(convolve2d(tf.lin(Sv120clean), k, 'same', boundary='symm'))
#------------------------------------------------------------------------------
# Mask swarms using Weill's algorithm
m120sh = mSH.weill(Sv120cvv, thr=-70, maxvgap=15, maxhgap=0,
minhlen= 3, minvlen=15)[0]
#------------------------------------------------------------------------------
# Figures
plt.figure(figsize=(8,5))
plt.subplots_adjust(left=0.08, right=0.91, bottom=0.08, top=0.95, wspace=0.08)
gs = gridspec.GridSpec(1, 3, width_ratios=[1, 1, .05])
# Sv original
plt.subplot(gs[0]).invert_yaxis()
im= plt.pcolormesh(t120, r120, Sv120, vmin=-80, vmax=-50, cmap=cmaps().ek500)
plt.ylabel('Depth (m)')
plt.xlabel('Time (dd HH:MM)')
plt.title('Sv')
# Swarms mask
plt.subplot(gs[1]).invert_yaxis()
plt.pcolormesh(t120, r120, m120sh*1, cmap='Greys')
plt.tick_params(labelleft=False)
plt.xlabel('Time (dd HH:MM)')
plt.title('Swarms')
# colorbar
ax=plt.subplot(gs[2])
plt.colorbar(im, cax=ax).set_label('dB re m$^{-1}$')
def log(pro, outputdir, savepng=True):
"""
Log processed data (*.csv) and echograms (*.png) in rapidkrill/log/.
Args:
pro (dict): processed data output from "process" routine.
outputdir (str ): directory name under which log results will be saved.
savepng (bool): True to save echogram images, False to skip it.
"""
results_data = get_results(pro)
results = pd.DataFrame(results_data,
columns=[
'Time', 'Longitude', 'Latitude', 'Transect',
'Miles', 'Seabed', 'NASC', '% samples'
])
# Create new log subdirectory
if not os.path.exists(outputdir):
os.makedirs(outputdir)
csvfile = os.path.join(outputdir, 'output.csv')
# Write results in CSV log file
with open(csvfile, 'a') as f:
results.to_csv(csvfile, index=False, mode='a', header=f.tell() == 0)
# save png image
if savepng:
rawfiles = pro['rawfiles']
transect = pro['transect']
t120 = pro['t120']
t120r = pro['t120r']
t120intrvls = pro['t120intervals']
nm120r = pro['nm120r']
r120 = pro['r120']
Sv120 = pro['Sv120']
Sv120sw = pro['Sv120sw']
NASC120swr = pro['NASC120swr'][0, :]
# set figure
plt.close()
plt.figure(figsize=(8, 8))
plt.subplots_adjust(left=0.066,
right=1.055,
bottom=0.065,
top=0.985,
wspace=0,
hspace=0.05)
plt.rcParams.update({'font.size': 9, 'lines.linewidth': 1})
# plot raw echogram
plt.subplot(211).invert_yaxis()
im = plt.pcolormesh(t120,
r120,
Sv120,
vmin=-80,
vmax=-50,
cmap=cmaps().ek500)
plt.colorbar(im).set_label('Sv raw (dB re 1m$^{-1}$)')
plt.gca().set_ylim(270, 0)
plt.gca().set_ylabel('Depth (m)')
plt.gca().set_xlim(t120intrvls[0], t120intrvls[-1])
plt.gca().set_xticks(t120intrvls[[0, -1]])
plt.tick_params(labelright=False, labelbottom=False)
# plot processed echogram
ax = plt.subplot(212)
ax = [ax, ax.twinx()]
im = ax[0].pcolormesh(t120,
r120,
Sv120sw,
vmin=-80,
vmax=-50,
cmap=cmaps().ek500)
plt.colorbar(im, ax=ax).set_label('Sv pro (dB re 1m$^{-1}$)')
ax[0].invert_yaxis()
ax[0].set_ylim(270, 0)
ax[0].set_ylabel('Depth (m)')
# overlay distance/NASC info
for t, nm, NASC in zip(t120r, nm120r, NASC120swr):
ax[1].plot([t, t], [0, 1], color=[0, .8, 0], linewidth=2)
ax[1].text(t,
.95,
' ' + str(transect) + ': ' + str(round(nm, 2)),
fontweight='bold',
color=[0, .8, 0])
ax[1].text(t,
.02,
' ' + str(round(NASC, 2)),
fontweight='bold',
color=[1, 0, 0])
ax[1].set_ylim(0, 1)
ax[1].set_xlim(t120intrvls[0], t120intrvls[-1])
ax[1].set_xticks(t120intrvls[[0, -1]])
ax[1].tick_params(labelright=False)
ax[1].xaxis.set_major_formatter(mdates.DateFormatter('%d%b-%H:%M:%S'))
# save figure
pf = rawfiles[0].split('-')[0]
fn = pd.to_datetime(str(t120[0])).strftime(pf + '-D%Y%m%d-T%H%M%S')
imagefile = os.path.join(outputdir, fn + '.png')
fig, ax = plt.subplots()
fig.set_size_inches(8, 8)
plt.savefig(imagefile, dpi=100)
plt.close()
#------------------------------------------------------------------------------
# integrate Nautical Area Scattering Coefficient (NASC), from 20 to 250 metres
NASC, NASCper = tf.Sv2NASC(Sv38inoff, r38, 20, 250, method='sum')
# note the sum method is used.
# the second output is the percentange vertical samples integrated behind every
# computation of NASC.
#------------------------------------------------------------------------------
# Figures
plt.figure(figsize=(8, 6))
# Sv with impulse noise removed
plt.subplot(311).invert_yaxis()
plt.pcolormesh(t38, r38, Sv38inoff, vmin=-80, vmax=-50, cmap=cmaps().ek500)
plt.tick_params(labelbottom=False)
plt.ylabel('Depth (m)')
plt.title('Sv with impulse noise removed')
# integrated sa
ax = plt.subplot(312)
ax = [ax, ax.twinx()]
ax[0].plot(t38, sa, '-r')
ax[0].set_xlim(t38[0], t38[-1])
ax[0].tick_params(axis='y', colors='r')
ax[0].yaxis.tick_left()
ax[0].yaxis.set_label_position("left")
ax[0].set_ylabel('s$_a$ (m$^2$ m$^{-2}$)', color='r')
ax[0].tick_params(labelbottom=False)
ax[1].plot(t38, saper, '-b')
# This time we set "log=False" becaue theta is already at linear scale.
# "log=False" is the default, so might skip this setting.
#------------------------------------------------------------------------------
# Resample Sv and theta back to full resolution.
Sv120rsf, m120rsf_ = rs.full(Sv120rs, r120rs, t120rs, r120, t120)
theta120rsf, m120rsf_ = rs.full(theta120rs, r120rs, t120, r120, t120)
#------------------------------------------------------------------------------
# Figures
plt.figure(figsize=(8, 6))
# Sv original
plt.subplot(221).invert_yaxis()
plt.pcolormesh(t120, r120, Sv120, vmin=-80, vmax=-50, cmap=cmaps().ek500)
plt.tick_params(labelbottom=False)
plt.ylabel('Depth (m)')
plt.title('Sv')
# Theta original
plt.subplot(222).invert_yaxis()
plt.pcolormesh(t120, r120, theta120, cmap=cmaps().coolwarm)
plt.tick_params(labelbottom=False)
plt.tick_params(labelleft=False)
plt.title('Theta')
# Sv 2D-resampled
plt.subplot(223).invert_yaxis()
plt.pcolormesh(t120, r120, Sv120rsf, vmin=-80, vmax=-50, cmap=cmaps().ek500)
plt.ylabel('Depth (m)')
#------------------------------------------------------------------------------
# Clean Sv from impulse noise with Wang's algorithm
Sv120wang, m120wang_ = mIN.wang(Sv120)
# Note that Wang's algorithm does not return a mask with impulse noise detected
# but Sv without impulse noise and other Sv modifications. It also removes
# Sv signal below and above the target of interest. In this case, krill swarms.
#------------------------------------------------------------------------------
# Figures
plt.figure(figsize=(8, 4))
# Sv original
plt.subplot(131).invert_yaxis()
plt.pcolormesh(p120, r120, Sv120, vmin=-80, vmax=-50, cmap=cmaps().ek500)
plt.ylabel('Depth (m)')
plt.title('IN on')
# IN removed with Ryan's algorithm
plt.subplot(132).invert_yaxis()
plt.pcolormesh(p120, r120, Sv120ryan, vmin=-80, vmax=-50, cmap=cmaps().ek500)
plt.tick_params(labelleft=False)
plt.xlabel('Number of pings')
plt.title('IN off (Ryan)')
# IN removed (and further signal) with Wang's algorithm
plt.subplot(133).invert_yaxis()
plt.pcolormesh(p120, r120, Sv120wang, vmin=-80, vmax=-50, cmap=cmaps().ek500)
plt.tick_params(labelleft=False)
plt.title('IN off (Wang)')
