def main():
for f in file_sound:
print f
df = mf.parse_sounding2(f)
fname = os.path.basename(f)
''' removes file extension and split date '''
raw_date = fname[:-4].split('_')
''' some files have preffix, so I take only datetime'''
raw_date = raw_date[-2:]
if len(raw_date[1]) == 6:
raw_date = raw_date[0] + raw_date[1]
date = dt.datetime.strptime(raw_date, "%Y%m%d%H%M%S")
else:
raw_date = raw_date[0] + raw_date[1]
date = dt.datetime.strptime(raw_date, "%Y%m%d%H%M")
# plot_skew(df,date)
plot_thermo(df, date, top=5.)
# compare_potential_temp(df,date)
# break
# print df
plt.show()
def main():
for f in file_sound:
print f
df = mf.parse_sounding2(f)
fname = os.path.basename(f)
''' removes file extension and split date '''
raw_date = fname[:-4].split('_')
''' some files have preffix, so I take only datetime'''
raw_date = raw_date[-2:]
if len(raw_date[1]) == 6:
raw_date = raw_date[0]+raw_date[1]
date = dt.datetime.strptime(raw_date, "%Y%m%d%H%M%S")
else:
raw_date = raw_date[0]+raw_date[1]
date = dt.datetime.strptime(raw_date, "%Y%m%d%H%M")
# plot_skew(df,date)
plot_thermo(df, date, top=5.)
# compare_potential_temp(df,date)
# break
# print df
plt.show()
for n,par in enumerate(params):
" synth data "
out = vitas.main(par[0])
p3u = out[0]
p3v = out[1]
p3wU = out[2] # from 230deg
p3z = out[3]
p3ulist.extend(p3u[:16])
p3vlist.extend(p3v[:16])
" balloon data "
infiles3,_ = so.get_sounding_files(par[1], homedir='/localdata')
infiles3.sort()
df = mf.parse_sounding2(infiles3[par[2]])
bSu = df.u.values
bSv = df.v.values
bSz = df.index.values
""" reduce resolution of bS by averaging 100-m layer
centered at p3 altitude """
bot = 25
top = 26
for z in p3z[:16]:
center_idx = np.where(bSz==z*1000.)[0]
if not center_idx:
bSulist.append(np.nan)
bSvlist.append(np.nan)
else:
c = center_idx[0]
def get_raw_array(soundvar, file_sound):
file_sound.sort()
''' height grid with 10 m resolution'''
var = [] # 2D list
timestamps = []
top_limit = 10000 # [m]
hgtgrid = np.asarray(range(10, 4010, 10))
for f in file_sound:
''' sounding timestamps approached to
the nearest 20 minute grid point'''
fname = os.path.basename(f)
raw_date = fname[:-4].split('_')[-2:]
y = raw_date[0][:4]
m = raw_date[0][4:6]
d = raw_date[0][6:8]
hh = int(raw_date[1][0:2])
mm = int(raw_date[1][2:4])
nearest = round_to_nearest(mm, 20)
if nearest == 60:
hh += 1
mm = 0
else:
mm = nearest
if hh == 24:
d = str(int(d) + 1)
hh = 0
raw_date = y+'-'+m+'-'+d+'T' + \
str(hh).zfill(2)+':'+str(mm).zfill(2)+':'+str(0).zfill(2)
timestamps.append(np.datetime64(raw_date))
''' variable '''
df = mf.parse_sounding2(f)
# print df
sv = df[soundvar][df.index < top_limit].values
Z = df.index[df.index < top_limit].values
''' interpolate to a common vertical grid '''
f = interp1d(Z, sv)
svinterp = np.asarray([])
for h in hgtgrid:
try:
svinterp = np.append(svinterp, f(h))
except ValueError:
svinterp = np.append(svinterp, np.nan)
''' take care of vertical nan values in raw soundings;
more gaps can show up if the top of the vertical grid
is increased (i.e. higher balloon altitudes show more
missing data)
'''
if np.any(np.isnan(svinterp)):
nanidx = np.where(np.isnan(svinterp))[0]
diff = np.diff(nanidx)
idxjump = np.where(diff > 1)[0]
ngaps = len(idxjump) + 1
if ngaps > 1 and nanidx[-1] != svinterp.size - 1:
gapidx = np.split(nanidx, idxjump + 1)
for g in gapidx:
first = g[0]
last = g[-1]
''' pick good values between nans and
make a linear interpolation (hopefully there
few nans) '''
if last + 1 == svinterp.size:
x = [hgtgrid[first - 1], hgtgrid[last]]
y = [svinterp[first - 1], svinterp[last]]
elif first == 0:
x = [hgtgrid[first], hgtgrid[last + 1]]
y = [svinterp[first], svinterp[last + 1]]
else:
x = [hgtgrid[first - 1], hgtgrid[last + 1]]
y = [svinterp[first - 1], svinterp[last + 1]]
f = interp1d(x, y)
svinterp[g] = [f(h) for h in hgtgrid[g]]
elif ngaps == 1 and nanidx[-1] != svinterp.size - 1:
first = nanidx[0]
last = nanidx[-1]
if last + 1 == svinterp.size:
x = [hgtgrid[first - 1], hgtgrid[last]]
y = [svinterp[first - 1], svinterp[last]]
elif first == 0:
x = [hgtgrid[first], hgtgrid[last + 1]]
y = [svinterp[first], svinterp[last + 1]]
else:
x = [hgtgrid[first - 1], hgtgrid[last + 1]]
y = [svinterp[first - 1], svinterp[last + 1]]
f = interp1d(x, y)
svinterp[nanidx] = [f(h) for h in hgtgrid[nanidx]]
var.append(svinterp)
else:
''' add column to list '''
var.append(svinterp)
''' time grid with 20 minute spacing'''
t1 = timestamps[0] - np.timedelta64(20, 'm')
t2 = timestamps[-1] + np.timedelta64(20, 'm')
dates = np.arange(t1, t2, dtype='datetime64[20m]')
''' get index of each sounding date '''
tidx = []
for t in timestamps:
tidx.append(np.where(dates == t)[0][0])
''' assign soundings to 2D array '''
soundarray = np.full((hgtgrid.size, dates.size), np.nan)
for n, i in enumerate(tidx):
soundarray[:, i] = var[n]
return soundarray, var, tidx, hgtgrid, dates, timestamps
def get_raw_array(soundvar, file_sound):
file_sound.sort()
''' height grid with 10 m resolution'''
var = [] # 2D list
timestamps = []
top_limit = 10000 # [m]
hgtgrid = np.asarray(range(10, 4010, 10))
for f in file_sound:
''' sounding timestamps approached to
the nearest 20 minute grid point'''
fname = os.path.basename(f)
raw_date = fname[:-4].split('_')[-2:]
y = raw_date[0][:4]
m = raw_date[0][4:6]
d = raw_date[0][6:8]
hh = int(raw_date[1][0:2])
mm = int(raw_date[1][2:4])
nearest = round_to_nearest(mm, 20)
if nearest == 60:
hh += 1
mm = 0
else:
mm = nearest
if hh == 24:
d = str(int(d)+1)
hh = 0
raw_date = y+'-'+m+'-'+d+'T' + \
str(hh).zfill(2)+':'+str(mm).zfill(2)+':'+str(0).zfill(2)
timestamps.append(np.datetime64(raw_date))
''' variable '''
df = mf.parse_sounding2(f)
# print df
sv = df[soundvar][df.index < top_limit].values
Z = df.index[df.index < top_limit].values
''' interpolate to a common vertical grid '''
f = interp1d(Z, sv)
svinterp = np.asarray([])
for h in hgtgrid:
try:
svinterp = np.append(svinterp, f(h))
except ValueError:
svinterp = np.append(svinterp, np.nan)
''' take care of vertical nan values in raw soundings;
more gaps can show up if the top of the vertical grid
is increased (i.e. higher balloon altitudes show more
missing data)
'''
if np.any(np.isnan(svinterp)):
nanidx = np.where(np.isnan(svinterp))[0]
diff = np.diff(nanidx)
idxjump = np.where(diff > 1)[0]
ngaps = len(idxjump) + 1
if ngaps > 1 and nanidx[-1] != svinterp.size-1:
gapidx = np.split(nanidx, idxjump+1)
for g in gapidx:
first = g[0]
last = g[-1]
''' pick good values between nans and
make a linear interpolation (hopefully there
few nans) '''
if last+1 == svinterp.size:
x = [hgtgrid[first-1], hgtgrid[last]]
y = [svinterp[first-1], svinterp[last]]
elif first == 0:
x = [hgtgrid[first], hgtgrid[last+1]]
y = [svinterp[first], svinterp[last+1]]
else:
x = [hgtgrid[first-1], hgtgrid[last+1]]
y = [svinterp[first-1], svinterp[last+1]]
f = interp1d(x, y)
svinterp[g] = [f(h) for h in hgtgrid[g]]
elif ngaps == 1 and nanidx[-1] != svinterp.size-1:
first = nanidx[0]
last = nanidx[-1]
if last+1 == svinterp.size:
x = [hgtgrid[first-1], hgtgrid[last]]
y = [svinterp[first-1], svinterp[last]]
elif first == 0:
x = [hgtgrid[first], hgtgrid[last+1]]
y = [svinterp[first], svinterp[last+1]]
else:
x = [hgtgrid[first-1], hgtgrid[last+1]]
y = [svinterp[first-1], svinterp[last+1]]
f = interp1d(x, y)
svinterp[nanidx] = [f(h) for h in hgtgrid[nanidx]]
var.append(svinterp)
else:
''' add column to list '''
var.append(svinterp)
''' time grid with 20 minute spacing'''
t1 = timestamps[0]-np.timedelta64(20, 'm')
t2 = timestamps[-1]+np.timedelta64(20, 'm')
dates = np.arange(t1, t2, dtype='datetime64[20m]')
''' get index of each sounding date '''
tidx = []
for t in timestamps:
tidx.append(np.where(dates == t)[0][0])
''' assign soundings to 2D array '''
soundarray = np.full((hgtgrid.size, dates.size), np.nan)
for n, i in enumerate(tidx):
soundarray[:, i] = var[n]
return soundarray, var, tidx, hgtgrid, dates, timestamps
nobs = ('n=7', 'n=11')
infiles3, _ = so.get_sounding_files('3', homedir='/localdata')
infiles7, _ = so.get_sounding_files('7', homedir='/localdata')
cmap = discrete_cmap(7, base_cmap='Set1')
color = (cmap(0), cmap(1))
infiles = (infiles3, infiles7)
for n, ax in enumerate(axes):
first = True
for f in infiles[n]:
df = mf.parse_sounding2(f)
x = np.expand_dims(df.bvf_moist.values, axis=1) * 10000
y = np.expand_dims(df.index.values, axis=1)
ax.plot(x, y, color=color[n], lw=0.5)
top = 2000 # [m]
top_idx = np.where(y == top)[0]
if first is True:
prof = x[:top_idx]
first = False
else:
prof = np.hstack((prof, x[:top_idx]))
meanx = np.expand_dims(np.nanmean(prof, axis=1), axis=1)
y2 = y[:top_idx]
ax.plot(meanx, y2, color=color[n], lw=3)
xpos = 0.08
ax.text(
nobs=('n=7','n=11')
infiles3,_ = so.get_sounding_files('3', homedir='/localdata')
infiles7,_ = so.get_sounding_files('7', homedir='/localdata')
cmap = discrete_cmap(7, base_cmap='Set1')
color=(cmap(0),cmap(1))
infiles=(infiles3,infiles7)
for n,ax in enumerate(axes):
first = True
for f in infiles[n]:
df = mf.parse_sounding2(f)
x = np.expand_dims(df.bvf_moist.values,axis=1)*10000
y = np.expand_dims(df.index.values,axis=1)
ax.plot(x,y,color=color[n],lw=0.5)
top = 2000 # [m]
top_idx = np.where(y == top)[0]
if first is True:
prof = x[:top_idx]
first = False
else:
prof = np.hstack((prof,x[:top_idx]))
meanx = np.expand_dims(np.nanmean(prof,axis=1),axis=1)
y2 = y[:top_idx]
ax.plot(meanx,y2,color=color[n],lw=3)
xpos = 0.08
ax.text(xpos,0.9,ax.get_gid(),