def test(p, parameters):
"""
Runs the quality control check on profile p and returns a numpy array
of quality control decisions with False where the data value has
passed the check and True where it failed.
"""
cruise = p.cruise()
uid = p.uid()
# don't bother if cruise == 0 or None, or if timestamp is corrupt
if (cruise in [0, None]) or (None in [p.year(), p.month(), p.day(), p.time()]):
return np.zeros(1, dtype=bool)
# don't bother if this has already been analyzed
command = 'SELECT en_track_check FROM ' + parameters["table"] + ' WHERE uid = ' + str(uid) + ';'
en_track_result = main.dbinteract(command)
if en_track_result[0][0] is not None:
en_track_result = main.unpack_row(en_track_result[0])[0]
result = np.zeros(1, dtype=bool)
result[0] = np.any(en_track_result)
return result
# some detector types cannot be assessed by this test; do not raise flag.
if p.probe_type() in [None]:
return np.zeros(1, dtype=bool)
# fetch all profiles on track, sorted chronologically, earliest first (None sorted as highest)
command = 'SELECT uid, year, month, day, time, lat, long, probe FROM ' + parameters["table"] + ' WHERE cruise = ' + str(cruise) + ' and year is not null and month is not null and day is not null and time is not null ORDER BY year, month, day, time, uid ASC;'
track_rows = main.dbinteract(command)
# start all as passing by default:
EN_track_results = {}
for i in range(len(track_rows)):
EN_track_results[track_rows[i][0]] = np.zeros(1, dtype=bool)
# copy the list of headers;
# remove entries as they are flagged.
passed_rows = copy.deepcopy(track_rows)
rejects = findOutlier(passed_rows, EN_track_results)
while rejects != []:
passed_index = [x for x in range(len(passed_rows)) if x not in rejects ]
passed_rows = [passed_rows[index] for index in passed_index ]
rejects = findOutlier(passed_rows, EN_track_results)
# if more than half got rejected, reject everyone
if len(passed_rows) < len(track_rows) / 2:
for i in range(len(track_rows)):
EN_track_results[track_rows[i][0]][0] = True
# write all to db
result = []
for i in range(len(track_rows)):
result.append((main.pack_array(EN_track_results[track_rows[i][0]]), track_rows[i][0]))
query = "UPDATE " + sys.argv[1] + " SET en_track_check=? WHERE uid=?"
main.interact_many(query, result)
return EN_track_results[uid]
def test(p, parameters):
"""
Runs the quality control check on profile p and returns a numpy array
of quality control decisions with False where the data value has
passed the check and True where it failed.
"""
country = p.primary_header['Country code']
cruise = p.cruise()
originator_cruise = p.originator_cruise()
uid = p.uid()
# don't bother if this has already been analyzed
command = 'SELECT en_track_check FROM ' + parameters["table"] + ' WHERE uid = ' + str(uid) + ';'
en_track_result = main.dbinteract(command)
if en_track_result[0][0] is not None:
en_track_result = main.unpack_row(en_track_result[0])[0]
result = np.zeros(1, dtype=bool)
result[0] = np.any(en_track_result)
return result
# make sure this profile makes sense in the track check
if not assess_usability(p):
return np.zeros(1, dtype=bool)
# fetch all profiles on track, sorted chronologically, earliest first (None sorted as highest), then by uid
command = 'SELECT uid, year, month, day, time, lat, long, probe, raw FROM ' + parameters["table"] + ' WHERE cruise = ' + str(cruise) + ' and country = "' + str(country) + '" and ocruise = "' + str(originator_cruise) + '" and year is not null and month is not null and day is not null and time is not null ORDER BY year, month, day, time, uid ASC;'
track_rows = main.dbinteract(command)
# avoid inappropriate profiles
track_rows = [tr for tr in track_rows if assess_usability_raw(tr[8][1:-1])]
# start all as passing by default
EN_track_results = {}
for i in range(len(track_rows)):
EN_track_results[track_rows[i][0]] = np.zeros(1, dtype=bool)
# copy the list of headers;
# remove entries as they are flagged.
passed_rows = copy.deepcopy(track_rows)
rejects = findOutlier(passed_rows, EN_track_results)
while rejects != []:
passed_index = [x for x in range(len(passed_rows)) if x not in rejects ]
passed_rows = [passed_rows[index] for index in passed_index ]
rejects = findOutlier(passed_rows, EN_track_results)
# if more than half got rejected, reject everyone
if len(passed_rows) < len(track_rows) / 2:
for i in range(len(track_rows)):
EN_track_results[track_rows[i][0]][0] = True
# write all to db
result = []
for i in range(len(track_rows)):
result.append((main.pack_array(EN_track_results[track_rows[i][0]]), track_rows[i][0]))
query = "UPDATE " + sys.argv[1] + " SET en_track_check=? WHERE uid=?"
main.interact_many(query, result)
return EN_track_results[uid]
def stdLevelData(p, parameters):
"""
Combines data that have passed other QC checks to create a
set of observation minus background data on standard levels.
"""
# Combine other QC results.
preQC = (EN_background_check.test(p, parameters) |
EN_constant_value_check.test(p, parameters) |
EN_increasing_depth_check.test(p, parameters) |
EN_range_check.test(p, parameters) |
EN_spike_and_step_check.test(p, parameters) |
EN_stability_check.test(p, parameters))
# Get the data stored by the EN background check.
# As it was run above we know that the data is available in the db.
query = 'SELECT origlevels, ptlevels, bglevels FROM enbackground WHERE uid = ' + str(p.uid())
enbackground_pars = main.dbinteract(query)
enbackground_pars = main.unpack_row(enbackground_pars[0])
origlevels = enbackground_pars[0]
ptlevels = enbackground_pars[1]
bglevels = enbackground_pars[2]
origLevels = np.array(origlevels)
diffLevels = (np.array(ptlevels) - np.array(bglevels))
nLevels = len(origLevels)
if nLevels == 0: return None # Nothing more to do.
# Remove any levels that failed previous QC.
nLevels, origLevels, diffLevels = filterLevels(preQC, origLevels, diffLevels)
if nLevels == 0: return None
levels, assocLevs = meanDifferencesAtStandardLevels(origLevels, diffLevels, p.z(), parameters)
return levels, origLevels, assocLevs
def test(p, parameters):
"""
Runs the quality control check on profile p and returns a numpy array
of quality control decisions with False where the data value has
passed the check and True where it failed.
"""
country = p.primary_header['Country code']
cruise = p.cruise()
originator_cruise = p.originator_cruise()
uid = p.uid()
# don't bother if this has already been analyzed
command = 'SELECT en_track_check FROM ' + parameters["table"] + ' WHERE uid = ' + str(uid) + ';'
en_track_result = main.dbinteract(command)
if en_track_result[0][0] is not None:
en_track_result = main.unpack_row(en_track_result[0])[0]
result = np.zeros(1, dtype=bool)
result[0] = np.any(en_track_result)
return result
# make sure this profile makes sense in the track check
if not assess_usability(p):
return np.zeros(1, dtype=bool)
# fetch all profiles on track, sorted chronologically, earliest first (None sorted as highest), then by uid
command = 'SELECT uid, year, month, day, time, lat, long, probe, raw FROM ' + parameters["table"] + ' WHERE cruise = ' + str(cruise) + ' and country = "' + str(country) + '" and ocruise = "' + str(originator_cruise) + '" and year is not null and month is not null and day is not null and time is not null ORDER BY year, month, day, time, uid ASC;'
track_rows = main.dbinteract(command)
# avoid inappropriate profiles
track_rows = [tr for tr in track_rows if assess_usability_raw(tr[8][1:-1])]
# start all as passing by default
EN_track_results = {}
for i in range(len(track_rows)):
EN_track_results[track_rows[i][0]] = np.zeros(1, dtype=bool)
# copy the list of headers;
# remove entries as they are flagged.
passed_rows = copy.deepcopy(track_rows)
rejects = findOutlier(passed_rows, EN_track_results)
while rejects != []:
passed_index = [x for x in range(len(passed_rows)) if x not in rejects ]
passed_rows = [passed_rows[index] for index in passed_index ]
rejects = findOutlier(passed_rows, EN_track_results)
# if more than half got rejected, reject everyone
if len(passed_rows) < len(track_rows) / 2:
for i in range(len(track_rows)):
EN_track_results[track_rows[i][0]][0] = True
# write all to db
result = []
for i in range(len(track_rows)):
result.append((main.pack_array(EN_track_results[track_rows[i][0]]), track_rows[i][0]))
query = "UPDATE " + sys.argv[1] + " SET en_track_check=? WHERE uid=?"
main.interact_many(query, result)
return EN_track_results[uid]
def test(p, parameters):
"""
Runs the quality control check on profile p and returns a numpy array
of quality control decisions with False where the data value has
passed the check and True where it failed.
"""
# Check if the QC of this profile was already done and if not
# run the QC.
query = 'SELECT en_constant_value_check FROM ' + parameters["table"] + ' WHERE uid = ' + str(p.uid()) + ';'
qc_log = main.dbinteract(query)
qc_log = main.unpack_row(qc_log[0])
if qc_log[0] is not None:
return qc_log[0]
return run_qc(p, parameters)
def test(p, parameters):
"""
Runs the quality control check on profile p and returns a numpy array
of quality control decisions with False where the data value has
passed the check and True where it failed.
"""
# Check if the QC of this profile was already done and if not
# run the QC.
query = 'SELECT en_increasing_depth_check FROM ' + parameters["table"] + ' WHERE uid = ' + str(p.uid()) + ';'
qc_log = main.dbinteract(query)
qc_log = main.unpack_row(qc_log[0])
if qc_log[0] is not None:
return qc_log[0]
return run_qc(p, parameters)
def stdLevelData(p, parameters):
"""
Combines data that have passed other QC checks to create a
set of observation minus background data on standard levels.
"""
# Combine other QC results.
preQC = (EN_background_check.test(p, parameters)
| EN_constant_value_check.test(p, parameters)
| EN_increasing_depth_check.test(p, parameters)
| EN_range_check.test(p, parameters)
| EN_spike_and_step_check.test(p, parameters)
| EN_stability_check.test(p, parameters))
# Get the data stored by the EN background check.
# As it was run above we know that the data is available in the db.
query = 'SELECT origlevels, ptlevels, bglevels FROM enbackground WHERE uid = ' + str(
p.uid())
enbackground_pars = main.dbinteract(query)
enbackground_pars = main.unpack_row(enbackground_pars[0])
origlevels = enbackground_pars[0]
ptlevels = enbackground_pars[1]
bglevels = enbackground_pars[2]
origLevels = np.array(origlevels)
diffLevels = (np.array(ptlevels) - np.array(bglevels))
nLevels = len(origLevels)
if nLevels == 0: return None # Nothing more to do.
# Remove any levels that failed previous QC.
nLevels, origLevels, diffLevels = filterLevels(preQC, origLevels,
diffLevels)
if nLevels == 0: return None
levels, assocLevs = meanDifferencesAtStandardLevels(
origLevels, diffLevels, p.z(), parameters)
return levels, origLevels, assocLevs
def test(p, parameters, allow_level_reinstating=True):
"""
Runs the quality control check on profile p and returns a numpy array
of quality control decisions with False where the data value has
passed the check and True where it failed.
If allow_level_reinstating is set to True then rejected levels can be
reprieved by comparing with levels above and below. NB this is done by
default in EN processing.
"""
# Define an array to hold results.
qc = np.zeros(p.n_levels(), dtype=bool)
# Obtain the obs minus background differences on standard levels.
result = stdLevelData(p, parameters)
if result is None:
return qc
# Unpack the results.
levels, origLevels, assocLevels = result
# Retrieve the background and observation error variances and
# the background values.
query = 'SELECT bgstdlevels, bgevstdlevels FROM enbackground WHERE uid = ' + str(
p.uid())
enbackground_pars = main.dbinteract(query)
enbackground_pars = main.unpack_row(enbackground_pars[0])
bgsl = enbackground_pars[0]
slev = parameters['enbackground']['depth']
bgev = enbackground_pars[1]
obev = parameters['enbackground']['obev']
#find initial pge
pgeData = determine_pge(levels, bgev, obev, p)
# Find buddy.
profiles = get_profile_info(parameters)
minDist = 1000000000.0
iMinDist = None
for iProfile, profile in enumerate(profiles):
pDist = assessBuddyDistance(p, profile)
if pDist is not None and pDist < minDist:
minDist = pDist
iMinDist = iProfile
# Check if we have found a buddy and process if so.
if minDist <= 400000:
pBuddy = main.get_profile_from_db(profiles[iMinDist][0])
# buddy vetos
Fail = False
if pBuddy.var_index() is None:
Fail = True
if Fail == False:
main.catchFlags(pBuddy)
if np.sum(pBuddy.t().mask == False) == 0:
Fail = True
if Fail == False:
result = stdLevelData(pBuddy, parameters)
query = 'SELECT bgevstdlevels FROM enbackground WHERE uid = ' + str(
pBuddy.uid())
buddy_pars = main.dbinteract(query)
buddy_pars = main.unpack_row(buddy_pars[0])
if result is not None:
levelsBuddy, origLevelsBuddy, assocLevelsBuddy = result
bgevBuddy = buddy_pars[0]
pgeBuddy = determine_pge(levels, bgevBuddy, obev, pBuddy)
pgeData = update_pgeData(pgeData, pgeBuddy, levels,
levelsBuddy, minDist, p, pBuddy, obev,
bgev, bgevBuddy)
# Check if levels should be reinstated.
if allow_level_reinstating:
if np.abs(p.latitude()) < 20.0:
depthTol = 300.0
else:
depthTol = 200.0
stdLevelFlags = pgeData >= 0.5
for i, slflag in enumerate(stdLevelFlags):
if slflag:
# Check for non rejected surrounding levels.
okbelow = False
if i > 0:
if stdLevelFlags[i - 1] == False and levels.mask[
i - 1] == False and bgsl.mask[i - 1] == False:
okbelow = True
okabove = False
nsl = len(stdLevelFlags)
if i < nsl - 1:
if stdLevelFlags[i + 1] == False and levels.mask[
i + 1] == False and bgsl.mask[i + 1] == False:
okabove = True
# Work out tolerances.
if slev[i] > depthTol + 100:
tolFactor = 0.5
elif slev[i] > depthTol:
tolFactor = 1.0 - 0.005 * (slev[i] - depthTol)
else:
tolFactor = 1.0
ttol = 0.5 * tolFactor
if okbelow == True and okabove == True:
xmax = levels[i - 1] + bgsl[i - 1] + ttol
xmin = levels[i + 1] + bgsl[i + 1] - ttol
elif okbelow == True:
xmax = levels[i - 1] + bgsl[i - 1] + ttol
xmin = levels[i - 1] + bgsl[i - 1] - ttol
elif okabove == True:
xmax = levels[i + 1] + bgsl[i + 1] + ttol
xmin = levels[i + 1] + bgsl[i + 1] - ttol
else:
continue
# Reassign PGE if level is within the tolerances.
if levels[i] + bgsl[i] >= xmin and levels[i] + bgsl[i] <= xmax:
pgeData[i] = 0.49
# Assign the QC flags to original levels.
for i, pge in enumerate(pgeData):
if pgeData.mask[i]: continue
if pge < 0.5: continue
for j, assocLevel in enumerate(assocLevels):
if assocLevel == i:
origLevel = origLevels[j]
qc[origLevel] = True
return qc
def run_qc(p, suspect):
# check for pre-registered suspect tabulation, if that's what we want:
if suspect:
query = 'SELECT suspect FROM enspikeandstep WHERE uid = ' + str(p.uid()) + ';'
susp = main.dbinteract(query)
if len(susp) > 0:
return main.unpack_row(susp[0])[0]
# Define tolerances used.
tolD = np.array([0, 200, 300, 500, 600])
tolDTrop = np.array([0, 300, 400, 500, 600])
tolT = np.array([5.0, 5.0, 2.5, 2.0, 1.5])
# Define an array to hold results.
qc = np.zeros(p.n_levels(), dtype=bool)
# Get depth and temperature values from the profile.
z = p.z()
t = p.t()
# Find which levels have data.
isTemperature = (t.mask==False)
isDepth = (z.mask==False)
isData = isTemperature & isDepth
# Array to hold temperature differences between levels and gradients.
dt, gt = composeDT(t, z, p.n_levels())
# Spikes and steps detection.
for i in range(1, p.n_levels()):
if i >= 2:
if (isData[i-2] and isData[i-1] and isData[i]) == False:
continue
if z[i] - z[i-2] >= 5.0:
wt1 = (z[i-1] - z[i-2]) / (z[i] - z[i-2])
else:
wt1 = 0.5
else:
if (isData[i-1] and isData[i]) == False:
continue
wt1 = 0.5
dTTol = determineDepthTolerance(z[i-1], np.abs(p.latitude()))
gTTol = 0.05
# Check for low temperatures in the Tropics.
# This might be more appropriate to appear in a separate EN regional
# range check but is included here for now for consistency with the
# original code.
if (np.abs(p.latitude()) < 20.0 and z[i-1] < 1000.0 and
t[i-1] < 1.0):
dt[i] = np.ma.masked
if suspect == True: qc[i-1] = True
continue
qc, dt = conditionA(dt, dTTol, qc, wt1, i, suspect)
qc, dt = conditionB(dt, dTTol, gTTol, qc, gt, i, suspect)
qc = conditionC(dt, dTTol, z, qc, t, i, suspect)
# End of loop over levels.
# Step or 0.0 at the bottom of a profile.
if isData[-1] and dt.mask[-1] == False:
dTTol = determineDepthTolerance(z[-1], np.abs(p.latitude()))
if np.abs(dt[-1]) > dTTol:
if suspect == True: qc[-1] = True
if isTemperature[-1]:
if t[-1] == 0.0:
if suspect == True: qc[-1] = True
# If 4 levels or more than half the profile is rejected then reject all.
if suspect == False:
nRejects = np.count_nonzero(qc)
if nRejects >= 4 or nRejects > p.n_levels()/2:
qc[:] = True
# register suspects, if computed, to db
if suspect:
query = "REPLACE INTO enspikeandstep VALUES(?,?);"
main.dbinteract(query, [p.uid(), main.pack_array(qc)] )
return qc
def run_qc(p, suspect, parameters):
# check for pre-registered suspect tabulation, if that's what we want:
if suspect:
query = 'SELECT suspect FROM enspikeandstep WHERE uid = ' + str(
p.uid()) + ';'
susp = main.dbinteract(query, targetdb=parameters["db"])
if len(susp) > 0:
return main.unpack_row(susp[0])[0]
# Define tolerances used.
tolD = np.array([0, 200, 300, 500, 600])
tolDTrop = np.array([0, 300, 400, 500, 600])
tolT = np.array([5.0, 5.0, 2.5, 2.0, 1.5])
# Define an array to hold results.
qc = np.zeros(p.n_levels(), dtype=bool)
# Get depth and temperature values from the profile.
z = p.z()
t = p.t()
# Find which levels have data.
isTemperature = (t.mask == False)
isDepth = (z.mask == False)
isData = isTemperature & isDepth
# Array to hold temperature differences between levels and gradients.
dt, gt = composeDT(t, z, p.n_levels())
# Spikes and steps detection.
for i in range(1, p.n_levels()):
if i >= 2:
if (isData[i - 2] and isData[i - 1] and isData[i]) == False:
continue
if z[i] - z[i - 2] >= 5.0:
wt1 = (z[i - 1] - z[i - 2]) / (z[i] - z[i - 2])
else:
wt1 = 0.5
else:
if (isData[i - 1] and isData[i]) == False:
continue
wt1 = 0.5
dTTol = determineDepthTolerance(z[i - 1], np.abs(p.latitude()))
gTTol = 0.05
# Check for low temperatures in the Tropics.
# This might be more appropriate to appear in a separate EN regional
# range check but is included here for now for consistency with the
# original code.
if (np.abs(p.latitude()) < 20.0 and z[i - 1] < 1000.0
and t[i - 1] < 1.0):
dt[i] = np.ma.masked
if suspect == True: qc[i - 1] = True
continue
qc, dt = conditionA(dt, dTTol, qc, wt1, i, suspect)
qc, dt = conditionB(dt, dTTol, gTTol, qc, gt, i, suspect)
qc = conditionC(dt, dTTol, z, qc, t, i, suspect)
# End of loop over levels.
# Step or 0.0 at the bottom of a profile.
if isData[-1] and dt.mask[-1] == False:
dTTol = determineDepthTolerance(z[-1], np.abs(p.latitude()))
if np.abs(dt[-1]) > dTTol:
if suspect == True: qc[-1] = True
if isTemperature[-1]:
if t[-1] == 0.0:
if suspect == True: qc[-1] = True
# If 4 levels or more than half the profile is rejected then reject all.
if suspect == False:
nRejects = np.count_nonzero(qc)
if nRejects >= 4 or nRejects > p.n_levels() / 2:
qc[:] = True
# register suspects, if computed, to db
if suspect:
query = "REPLACE INTO enspikeandstep VALUES(?,?);"
main.dbinteract(query, [p.uid(), main.pack_array(qc)],
targetdb=parameters["db"])
return qc
def test(p, parameters, allow_level_reinstating=True):
"""
Runs the quality control check on profile p and returns a numpy array
of quality control decisions with False where the data value has
passed the check and True where it failed.
If allow_level_reinstating is set to True then rejected levels can be
reprieved by comparing with levels above and below. NB this is done by
default in EN processing.
"""
# Define an array to hold results.
qc = np.zeros(p.n_levels(), dtype=bool)
# Obtain the obs minus background differences on standard levels.
result = stdLevelData(p, parameters)
if result is None:
return qc
# Unpack the results.
levels, origLevels, assocLevels = result
# Retrieve the background and observation error variances and
# the background values.
query = 'SELECT bgstdlevels, bgevstdlevels FROM enbackground WHERE uid = ' + str(p.uid())
enbackground_pars = main.dbinteract(query)
enbackground_pars = main.unpack_row(enbackground_pars[0])
bgsl = enbackground_pars[0]
slev = parameters['enbackground']['depth']
bgev = enbackground_pars[1]
obev = parameters['enbackground']['obev']
#find initial pge
pgeData = determine_pge(levels, bgev, obev, p)
# Find buddy.
profiles = get_profile_info(parameters)
minDist = 1000000000.0
iMinDist = None
for iProfile, profile in enumerate(profiles):
pDist = assessBuddyDistance(p, profile)
if pDist is not None and pDist < minDist:
minDist = pDist
iMinDist = iProfile
# Check if we have found a buddy and process if so.
if minDist <= 400000:
pBuddy = main.get_profile_from_db(profiles[iMinDist][0])
# buddy vetos
Fail = False
if pBuddy.var_index() is None:
Fail = True
if Fail == False:
main.catchFlags(pBuddy)
if np.sum(pBuddy.t().mask == False) == 0:
Fail = True
if Fail == False:
result = stdLevelData(pBuddy, parameters)
query = 'SELECT bgevstdlevels FROM enbackground WHERE uid = ' + str(pBuddy.uid())
buddy_pars = main.dbinteract(query)
buddy_pars = main.unpack_row(buddy_pars[0])
if result is not None:
levelsBuddy, origLevelsBuddy, assocLevelsBuddy = result
bgevBuddy = buddy_pars[0]
pgeBuddy = determine_pge(levels, bgevBuddy, obev, pBuddy)
pgeData = update_pgeData(pgeData, pgeBuddy, levels, levelsBuddy, minDist, p, pBuddy, obev, bgev, bgevBuddy)
# Check if levels should be reinstated.
if allow_level_reinstating:
if np.abs(p.latitude()) < 20.0:
depthTol = 300.0
else:
depthTol = 200.0
stdLevelFlags = pgeData >= 0.5
for i, slflag in enumerate(stdLevelFlags):
if slflag:
# Check for non rejected surrounding levels.
okbelow = False
if i > 0:
if stdLevelFlags[i - 1] == False and levels.mask[i - 1] == False and bgsl.mask[i - 1] == False:
okbelow = True
okabove = False
nsl = len(stdLevelFlags)
if i < nsl - 1:
if stdLevelFlags[i + 1] == False and levels.mask[i + 1] == False and bgsl.mask[i + 1] == False:
okabove = True
# Work out tolerances.
if slev[i] > depthTol + 100:
tolFactor = 0.5
elif slev[i] > depthTol:
tolFactor = 1.0 - 0.005 * (slev[i] - depthTol)
else:
tolFactor = 1.0
ttol = 0.5 * tolFactor
if okbelow == True and okabove == True:
xmax = levels[i - 1] + bgsl[i - 1] + ttol
xmin = levels[i + 1] + bgsl[i + 1] - ttol
elif okbelow == True:
xmax = levels[i - 1] + bgsl[i - 1] + ttol
xmin = levels[i - 1] + bgsl[i - 1] - ttol
elif okabove == True:
xmax = levels[i + 1] + bgsl[i + 1] + ttol
xmin = levels[i + 1] + bgsl[i + 1] - ttol
else:
continue
# Reassign PGE if level is within the tolerances.
if levels[i] + bgsl[i] >= xmin and levels[i] + bgsl[i] <= xmax:
pgeData[i] = 0.49
# Assign the QC flags to original levels.
for i, pge in enumerate(pgeData):
if pgeData.mask[i]: continue
if pge < 0.5: continue
for j, assocLevel in enumerate(assocLevels):
if assocLevel == i:
origLevel = origLevels[j]
qc[origLevel] = True
return qc