def timeseries_add_breaks(data, var, p=None, ax=None, color='k'):
""" Add Breakpoint lines to plot
Parameters
----------
data
var
p
ax
color
Returns
-------
"""
import matplotlib.pyplot as plt
funcid='TB'
if not isinstance(var, str):
raise ValueError("Variable is a string")
pressure_levels, dates, plevs, test, label = timeseries_preprocessor(data,
[var, '%s_breaks' % var, '%s_snht' % var],
p=p, funcid=funcid)
#
# Plot summed SNHT
#
if pressure_levels:
breaks = dates[(test["%s_breaks" % var] != 0).all(1)]
else:
breaks = dates[(test["%s_breaks" % var] != 0)]
#
if ax is None:
ax = plt.gca()
[ax.axvline(x=ib, color=color) for ib in breaks]
def timeseries_occurence(data, var, reference=None, levels=None, quantilen=None, limit=500, p=None, ax=None, legend=True, name=None,
verbose=0, **kwargs):
"""
Parameters
----------
data
var
reference
quantilen
p
ax
legend
name
verbose
kwargs
"""
import matplotlib.pyplot as plt
from raso.breakpoint import quantile_count
funcid = "[plot TOCC] "
if not isinstance(var, str):
raise ValueError("Variable is a string")
pressure_levels, dates, plevs, test, label = timeseries_preprocessor(data, var, p=p, funcid=funcid)
title = kwargs.pop('title', 'Radiosonde Occurence Frequency')
yticklabels = kwargs.pop('yticklabels', None)
xlabel = kwargs.pop('xlabel', 'Time')
if ax is None:
f, ax = plt.subplots(figsize=kwargs.pop('figsize', (10, 6)))
if name is None:
name = 'Occurence Freq. %s' % var
#
# Count per month
#
test = quantile_count(test, var, reference=reference, levels=levels, quantilen=quantilen, limit=limit)
# gives a panel if pressure_levels are present
#
# Plot (contour or contour per level ?)
#
if not pressure_levels:
test.plot(ax=ax, kind='area')
# ax.plot(dates, test, **kwargs)
ax.set_ylabel(name)
ax.grid(True)
ax.set_xlabel(xlabel)
ax.set_title(title)
else:
print "not sure how to plot all this information"
return test
def breakpoint_timeseries(data, var, unit='K', p=None, bins=None, departures=False, post_snht=False, raw=False,
window=365, min_periods=30, figsize=(16,9), ylim=None, verbose=0, **kwargs):
"""
Parameters
----------
data
var
unit
p
bins
departures
post_snht
window
min_periods
figsize
ylim
verbose
Returns
-------
"""
import matplotlib.pyplot as plt
from timeseries import timeseries_snht
from raso.breakpoint.detections import _1d_detection
funcid = '[BTS] '
pressure_levels, dates, plevs, test, label = timeseries_preprocessor(data,
[var, "%s_dep_breaks" % var, "%s_dep_snht" % var],
p=p,
add_vars=["%s_mcor" % var, "%s_qcor" % var,
"%s_qecor" % var, "%s_era" % var,
"%s_era_adj" % var, "%s_dep" % var],
funcid=funcid)
if plevs is None:
raise RuntimeError(funcid + " Require an availbale pressure level at least!")
else:
if isinstance(plevs, (int,float)):
plevs = [plevs] # as list
if bins is None:
bins = np.arange(0, 50)
varselection = []
varlabel = []
correction_present = False
for jvar in label:
if jvar == var:
varlabel = ['U']
varselection.append(jvar)
elif 'mcor' in jvar:
varlabel.append('M')
varselection.append(jvar)
correction_present = True
elif 'qcor' in jvar:
varlabel.append('Q')
varselection.append(jvar)
correction_present = True
elif 'qecor' in jvar:
varlabel.append('QE')
varselection.append(jvar)
correction_present = True
elif 'adj' in jvar:
varlabel.append('EA')
varselection.append(jvar)
elif jvar == "%s_era" % var:
varlabel.append('E')
varselection.append(jvar)
else:
if verbose > 0:
print "Not used: ", jvar
if verbose > 0:
print varlabel
print varselection
#
title = kwargs.pop('title', 'Radiosonde Breakpoint Detection and Correction')
xlabel = kwargs.pop('xlabel', 'Time')
if pressure_levels:
breaks = filter_series((test["%s_dep_breaks" % var] > 0).any(1))
else:
breaks = filter_series((test["%s_dep_breaks" % var] > 0))
#
nbreaks = len(breaks)
#
if nbreaks == 0:
nbreaks = 1
nplotlevels = 4 # snht, timeseries, hist, hist
if not correction_present:
nplotlevels = 3
for z, ilev in enumerate(plevs):
print "%d / %d : %d" % (z+1, len(plevs), ilev)
if pressure_levels:
plotdata = test.minor_xs(ilev)
else:
plotdata = test
if verbose > 0:
print plotdata.shape
print plotdata.columns
plotdata = plotdata.resample('12h').asfreq() # make sure it is the same freq
#
f = plt.figure(figsize=figsize)
#
# Figure 1 : SNHT
#
axs = plt.subplot2grid((nplotlevels, nbreaks), (0, 0), colspan=nbreaks)
timeseries_snht(plotdata, "%s_dep" % var, ax=axs, label='U')
axs.set_ylabel('SNHT')
axs.set_yticklabels(axs.get_yticklabels()[::2])
if post_snht:
for k, zvar in enumerate(varselection):
if (zvar == '%s_era' % var) or (zvar == var):
continue
tmpdata = plotdata.loc[:, zvar].copy().to_frame()
if verbose > 0:
print "SNHT for ", zvar
status = _1d_detection(tmpdata, zvar, 1460, 730, 50) # default parameters
axs.plot(tmpdata.index, tmpdata['%s_snht' % zvar], label=varlabel[k], alpha=0.7, ls='--')
axs.legend(loc="upper left", fontsize=plt.rcParams.get('font.size') - 2, bbox_to_anchor=(1.02, 1),
bbox_transform=axs.transAxes)
#
plt.setp(axs.get_xticklabels(), visible=False)
axt = plt.subplot2grid((nplotlevels, nbreaks), (1, 0), colspan=nbreaks, sharex=axs)
#
# DEPARTURES
#
if departures:
for jvar in varselection:
if jvar == "%s_era" % var:
continue
plotdata[jvar] = plotdata[jvar] - plotdata["%s_era" % var] # ref !!
#
if raw:
pd_mean = plotdata
else:
pd_mean = plotdata.rolling(window=window, min_periods=min_periods).mean()
#
# Figure 2 : Values
#
for j, jvar in enumerate(varselection):
if departures and jvar == '%s_era' % var:
continue
axt.plot(pd_mean.index, pd_mean[jvar], lw=2, label=varlabel[j])
axs.set_xticks(np.unique(pd_mean.index.year.astype('str'))[::2]) # ticks
axs.set_title(title + " %d hPa" % int(ilev/100))
axs.grid() # overwrite
axs.grid()
axt.set_xticks(np.unique(pd_mean.index.year.astype('str'))[::2]) # ticks
if ylim is None:
ym = int(pd_mean[var].mean())
ys = 5 # int(pd_mean[var].std())*2
ylim = (ym-ys, ym+ys)
axt.set_ylim(ylim)
axt.set_yticks(axt.get_yticks()[::2])
if departures:
axt.set_ylabel("%s [%s]" % (var.upper(),unit))
else:
axt.set_ylabel("%s dep. [%s]" % (var.upper(),unit))
axt.grid()
# axt.legend(loc="upper center", ncol=5, fontsize=plt.rcParams.get('font.size') - 2) # bbox_to_anchor=(1, 1))
axt.legend(loc="upper left", fontsize=plt.rcParams.get('font.size') - 2, bbox_to_anchor=(1.02, 1),
bbox_transform=axt.transAxes)
[axt.axvline(x=ib, color='k') for ib in breaks]
[axt.text(ib, pd_mean[var].min(), "B%02d" % (nbreaks - i), color='k') for i, ib in enumerate(breaks)]
#
#
# Figure 3: Histograms for each breakpoint
#
for ibreak in range(nbreaks):
if ibreak == 0:
ax1 = plt.subplot2grid((nplotlevels, nbreaks), (2, ibreak))
axi = ax1
else:
axi = plt.subplot2grid((nplotlevels, nbreaks), (2, ibreak), sharey=ax1)
# Reference
if ibreak + 1 < nbreaks:
iref = slice(breaks[ibreak], breaks[ibreak + 1])
else:
iref = slice(breaks[ibreak], None)
# Biased Sample
if ibreak > 0:
ibiased = slice(breaks[ibreak - 1], breaks[ibreak])
else:
ibiased = slice(None, breaks[ibreak])
#
axi.hist(plotdata.ix[iref, var], bins=bins, normed=1, label='BB')
if "%s_era_adj" % var in varselection:
axi.hist(plotdata.ix[iref, "%s_era_adj" % var], bins=bins, normed=1, label='EA', alpha=0.6)
axi.hist(plotdata.ix[ibiased, var], bins=bins, normed=1, label='AB', alpha=0.6)
#
if ibreak == 0:
axi.set_yticks(axi.get_yticks()[::2])
axi.grid()
axi.set_title("B%02d" % (nbreaks - ibreak), y=0.75)
if ibreak > 0:
plt.setp(axi.get_yticklabels(), visible=False)
#
# Figure 4: Correction Histograms for each breakpoint
#
if correction_present:
axj = plt.subplot2grid((nplotlevels, nbreaks), (3, ibreak), sharey=ax1, sharex=axi)
plt.setp(axi.get_xticklabels(), visible=False)
alpha=1
for j, jvar in enumerate(varselection):
if 'cor' in jvar:
axj.hist(plotdata.ix[ibiased, jvar], bins=bins, normed=1, label=varlabel[j], alpha=(alpha-0.1*j))
axj.grid()
axj.set_xlabel('%s [%s]' % (var.upper(), unit))
if ibreak > 0:
plt.setp(axj.get_yticklabels(), visible=False)
axi.legend(loc='upper left', fontsize=plt.rcParams.get('font.size') - 2, bbox_to_anchor=(1.05, 1))
if correction_present:
axj.legend(loc='upper left', fontsize=plt.rcParams.get('font.size') - 2, bbox_to_anchor=(1.05, 1))
def timeseries_anomaly(data, var, dvar, p=None, ax=None, levels=None, legend=True, name=None, freq=None,
resample_func='mean', add_zero=False, logy=True, breakvar=None, verbose=0, **kwargs):
"""
Parameters
----------
data
var
dvar
p
ax
levels
legend
name
freq
resample_func
add_zero
verbose
kwargs
"""
import matplotlib.pyplot as plt
funcid = "[plot TANO] "
if not isinstance(var, str):
raise ValueError("Variable is a string")
if not isinstance(dvar, str):
raise ValueError("Variable is a string")
pressure_levels, dates, plevs, test, label = timeseries_preprocessor(data, [var, dvar], p=p,
add_vars=breakvar, funcid=funcid)
title = kwargs.pop('title', 'Radiosonde Anomaly Timeseries')
yticklabels = kwargs.pop('yticklabels', None)
xlabel = kwargs.pop('xlabel', 'Time')
if ax is None:
f, ax = plt.subplots(figsize=kwargs.pop('figsize', (10, 6)))
if name is None:
name = 'Anomaly (%s - %s)' % (dvar, var)
breaks = None
if breakvar in label:
if pressure_levels:
breaks = np.unique(dates[(test[breakvar] != 0).all(1)])
else:
breaks = np.unique(dates[(test[breakvar] != 0)])
#
# Resample
#
if freq is not None:
test = (test[dvar] - test[var]).resample(freq, axis=0).apply(resample_func)
dates = test.index
if pressure_levels:
plevs = test.columns
test = test.values
xlabel += " %s (%s)" % (freq, resample_func)
else:
test = (test[dvar] - test[var]).values
#
# Plot (timeseries or contour)
#
if not pressure_levels:
ax.plot(dates, test, **kwargs)
ax.set_ylabel(name)
if add_zero:
ax.axhline(y=0, color='k')
else:
# PCOLOR MESH VERSION
# test = np.ma.masked_where(np.isnan(test),test)
# cmap = plt.get_cmap(cmap)
# norm = BoundaryNorm(levels, ncolors=cmap.N, clip=True)
# CS = ax.pcolormesh(data.major_axis.to_pydatetime(), data.minor_axis.values,
# test.T,norm=norm,cmap=cmap)# ,levels=levels,extend='both')
if 'cmap' not in kwargs.keys():
kwargs['cmap'] = 'RdBu_r'
cs = ax.contourf(dates, plevs, test.T, levels=levels, **kwargs)
if logy:
ax.set_yscale('log')
ax.set_yticks(plevs, minor=True)
if yticklabels is not None:
yticklabels = np.asarray(yticklabels) # can not calc on list
ax.set_yticks(yticklabels)
ax.set_yticklabels(np.int_(yticklabels / 100.))
else:
ax.set_yticks(plevs[::2])
ax.set_yticklabels(np.int_(plevs[::2] / 100.))
ax.set_ylabel('p Levels')
ax.invert_yaxis()
if legend:
cbar = plt.colorbar(cs, ax=ax, orientation="vertical", fraction=0.1)
cbar.ax.set_ylabel(name)
if add_zero:
cszero = ax.contour(dates, plevs, test.T, (0,), colors='k')
ax.grid(True)
ax.set_xlabel(xlabel)
if breaks is not None:
[ax.axvline(x=ib, color='k') for ib in breaks]
ax.set_title(title)
if pressure_levels and not legend:
return cs
def timeseries_var(data, var, p=None, ax=None, levels=None, legend=True, name=None, freq=None,
resample_func='mean', logy=True, verbose=0, **kwargs):
"""
Parameters
----------
data
var
p
ax
levels
legend
name
freq
resample_func
verbose
kwargs
"""
import matplotlib.pyplot as plt
funcid = "[plot TVAR]"
if not isinstance(var, str):
raise ValueError("Variable is a string")
pressure_levels, dates, plevs, test, label = timeseries_preprocessor(data, var, p=p,
add_vars='%s_breaks' % var, funcid=funcid)
title = kwargs.pop('title', 'Radiosonde Timeseries')
yticklabels = kwargs.pop('yticklabels', None)
xlabel = kwargs.pop('xlabel', 'Time')
if ax is None:
f, ax = plt.subplots(figsize=kwargs.pop('figsize', (10, 6)))
if name is None:
name = 'Variable %s' % var
breaks = None
if '%s_dep_breaks' % var in label:
if pressure_levels:
breaks = np.unique(dates[(test["%s_breaks" % var] != 0).all(1)])
else:
breaks = np.unique(dates[(test["%s_breaks" % var] != 0)])
#
# Resample
#
if freq is not None:
test = test[var].resample(freq, axis=0).apply(resample_func)
dates = test.index
if pressure_levels:
plevs = test.columns
test = test.values
xlabel += " %s (%s)" % (freq, resample_func)
else:
test = test[var].values
#
# Plot (timeseries or contour)
#
if not pressure_levels:
ax.plot(dates, test, **kwargs)
ax.set_ylabel(name)
else:
if 'cmap' not in kwargs.keys():
kwargs['cmap'] = 'RdYlBu_r'
cs = ax.contourf(dates, plevs, test.T, levels=levels, **kwargs)
if logy:
ax.set_yscale('log')
ax.set_yticks(plevs, minor=True)
if yticklabels is not None:
yticklabels = np.asarray(yticklabels) # can not calc on list
ax.set_yticks(yticklabels)
ax.set_yticklabels(np.int_(yticklabels / 100.))
else:
ax.set_yticks(plevs[::2])
ax.set_yticklabels(np.int_(plevs[::2] / 100.))
ax.set_ylabel('p Levels')
ax.invert_yaxis()
if legend:
cbar = plt.colorbar(cs, ax=ax, orientation="vertical", fraction=0.1)
cbar.ax.set_ylabel(name)
ax.grid(True)
ax.set_xlabel(xlabel)
if breaks is not None:
[ax.axvline(x=ib, color='k') for ib in breaks]
ax.set_title(title)
if pressure_levels and not legend:
return cs
def timeseries_snht(data, var, p=None, borders=None, ax=None, legend=True, levels=None, logy=True,
**kwargs):
""" Creates a timeseries plot of the SNHT
1. time x pressure levels
2. time at a specified pressure level
Parameters
----------
data DataFrame/Panel Radiosonde Data
var str Variable name that has an snht variable as well
p int/float Pressure level
borders int days befre and after breakpoint to highlight
ax Axes
legend bool
kwargs **
"""
import matplotlib.gridspec as gridspec
import matplotlib.pyplot as plt
funcid = '[plot TSNHT] '
if not isinstance(var, str):
raise ValueError("Variable is a string")
pressure_levels, dates, plevs, test, label = timeseries_preprocessor(data,
[var, '%s_breaks' % var, '%s_snht' % var],
p=p, funcid=funcid)
title = kwargs.pop('title', 'Radiosonde SNHT')
yticklabels = kwargs.pop('yticklabels', None)
xlabel = kwargs.pop('xlabel', 'Time')
if borders is not None:
title += " B(%d days)" % borders
if ax is None:
if pressure_levels:
f = plt.figure(figsize=kwargs.pop('figsize', (10, 6)))
gs = gridspec.GridSpec(2, 1, height_ratios=[1, 4])
ax = plt.subplot(gs[0])
axy = plt.subplot(gs[1], sharex=ax)
else:
f, ax = plt.subplots(figsize=kwargs.pop('figsize', (10, 6)))
#
# Plot summed SNHT
#
if pressure_levels:
breaks = dates[(test["%s_breaks" % var] != 0).all(1)]
tmp = data["%s_snht" % var].sum(1)
ax.plot(dates, tmp) # Kwargs ar going to contour
if 'cmap' not in kwargs.keys():
kwargs['cmap'] = 'Reds'
else:
breaks = dates[(test["%s_breaks" % var] != 0)]
tmp = test["%s_snht" % var]
ax.plot(dates, tmp, **kwargs) # Plot
#
ax.grid(True)
[ax.axvline(x=ib, color='k') for ib in breaks]
#
if borders is not None:
for ib in breaks:
# tmp = test.ix['%s_snht' % var, (ib-pd.DateOffset(days=borders)):(ib+pd.DateOffset(days=borders)), :]
x = tmp.ix[(ib-pd.DateOffset(days=borders)):(ib+pd.DateOffset(days=borders))]
ax.fill_between(x.index, 0, x.values, color='m', alpha=0.6)
ax.set_title(title)
if not pressure_levels:
ax.set_xlabel(xlabel)
return
ax.set_xlabel('')
ax.set_ylabel(r'$\sum$ SNHT')
ax.set_yticks(ax.get_yticks()[::2])
#
# Plot Contour Plot
#
area = data['%s_snht' % var]
area = np.where(area < 10, np.nan, area)
cs = axy.contourf(dates, plevs, area.T, levels=levels, **kwargs)
if logy:
axy.set_yscale('log')
axy.set_yticks(plevs, minor=True)
if yticklabels is not None:
yticklabels = np.asarray(yticklabels) # can not calc on list
axy.set_yticks(yticklabels)
axy.set_yticklabels(np.int_(yticklabels/100.))
else:
axy.set_yticks(plevs[::2])
axy.set_yticklabels(np.int_(plevs[::2]/100.))
axy.set_ylabel('p Levels')
axy.set_xlabel(xlabel)
axy.grid(True)
[axy.axvline(x=ib, color='k') for ib in breaks]
axy.invert_yaxis()
if legend:
cbar = plt.colorbar(cs, ax=axy, orientation="horizontal", fraction=0.05)
cbar.ax.set_ylabel('SNHT')
else:
return cs
