def plot_histogram_maps(hist_filename1, hist_filename2, runtitle,
plotname_root):
"""
Plot histogram maps
"""
ppn_hist_cube1 = make_hist_maps.read_data_cube(hist_filename1)
ppn_hist_cube2 = make_hist_maps.read_data_cube(hist_filename2)
avg_rain_bins_a, avg_rain_bins_frac_a = make_hist_maps.calc_rain_contr(
ppn_hist_cube1)
avg_rain_bins_b, avg_rain_bins_frac_b = make_hist_maps.calc_rain_contr(
ppn_hist_cube2)
# (optional) Define how you want to lump the bins together (below is the default)
all_ppn_bounds = [(0.005, 10.), (10., 50.), (50., 100.), (100., 3000.)]
# Plot as actual contributions for specific region, e.g. 60 to 160E,10S to 10N
plotname = '{}_actual_contributions.png'.format(plotname_root)
plot_hist_maps.plot_rain_contr(avg_rain_bins_a,
avg_rain_bins_b,
plotname,
runtitle,
'Timescale 1',
'Timescale 2',
all_ppn_bounds,
region=[60.0, -10.0, 160.0, 10.0])
# Plot as fractional contributions
plotname = '{}_fractional_contributions.png'.format(plotname_root)
plot_hist_maps.plot_rain_contr(avg_rain_bins_frac_a,
avg_rain_bins_frac_b,
plotname,
runtitle,
'Timescale 1',
'Timescale 2',
all_ppn_bounds,
region=[60.0, -10.0, 160.0, 10.0],
frac=1)
return
def plot_metric(filename1, filename2, dataname1, dataname2, season, timescale,
dates, maskfile, plotname):
"""
Args:
* filename1,filename2:
files containing histogram counts for two datasets, which must be on the same grid.
filename2 data are the baseline (e.g. obs) against which data in filename1
will be compared.
* dataname1,dataname2:
names of datasets in filename1, filename2 (e.g. a model name and obs name,
or two model names)
* season:
string descriptor of season for which data in filenames apply (e.g. 'JJA')
* timescale:
string descriptor of time frequency of original data (e.g. '3-hourly')
* dates:
string descriptor of dates to which histograms apply
* maskfile:
filename for land/sea fraction dataset (on same grid as both filenames)
with values ranging from 0.0 = no land, to 1.0 = all land.
* plotname:
filename for plot
Example usage:
index, indexl, indexs, indextrop, indexnhml, indexshml=plot_metric(filename1,filename2,dataname,season,'3-hourly','1990-2014',maskfile,plotname)
"""
seasonc = season.upper()
ppn_hist_cube1 = read_data_cube(filename1)
ppn_hist_cube2 = read_data_cube(filename2)
tot_rain_bins_seas_a, tot_rain_bins_frac_a = calc_rain_contr(
ppn_hist_cube1)
tot_rain_bins_seas_b, tot_rain_bins_frac_b = calc_rain_contr(
ppn_hist_cube2)
# Limit region to 60S-60N to fit with use of GPM-IMERG observations
# Could be omitted or made optional according to which two datasets are being compared.
ce = iris.coords.CoordExtent('longitude', 0.0, 360.0)
ce2 = iris.coords.CoordExtent('latitude', -59, 59)
tot_rain_bins_frac_a.coord('longitude').circular = True
if tot_rain_bins_frac_a.coord('longitude').bounds is None:
tot_rain_bins_frac_a.coord('longitude').guess_bounds()
tot_rain_bins_frac_a.coord('latitude').guess_bounds()
tot_rain_bins_frac_a1 = tot_rain_bins_frac_a.intersection(ce, ce2)
tot_rain_bins_frac_b.coord('longitude').circular = True
if tot_rain_bins_frac_b.coord('longitude').bounds is None:
tot_rain_bins_frac_b.coord('longitude').guess_bounds()
tot_rain_bins_frac_b.coord('latitude').guess_bounds()
tot_rain_bins_frac_b1 = tot_rain_bins_frac_b.intersection(ce, ce2)
minf = tot_rain_bins_frac_a1.copy()
minf.data = np.minimum(tot_rain_bins_frac_a1.data,
tot_rain_bins_frac_b1.data)
skill = minf.collapsed('precipitation_flux', iris.analysis.SUM, mdtol=1)
# Get mask (currently total rain <1 mm/day)
bin_mid2 = np.exp(
np.log(0.005) + np.sqrt(
np.linspace(0.5, 98.5, 99) *
((np.square(np.log(120.) - np.log(0.005))) / 59.)))
bin_mid = np.zeros(100)
bin_mid[1:] = bin_mid2[0:99]
bin_mid3 = bin_mid.reshape(100, 1, 1)
lshista = iris.load_cube(filename1)
tot_events = lshista.collapsed('precipitation_flux', iris.analysis.SUM)
tot_events.data = np.ma.masked_where(tot_events.data == 0.0,
tot_events.data)
tot_rain_bins = lshista * bin_mid3
tot_rain_boxa = tot_rain_bins.collapsed('precipitation_flux',
iris.analysis.SUM) / tot_events
lshistb = iris.load_cube(filename2)
lshistb.coord('latitude').coord_system = lshista.coord(
'latitude').coord_system
lshistb.coord('longitude').coord_system = lshista.coord(
'longitude').coord_system
tot_events = lshistb.collapsed('precipitation_flux', iris.analysis.SUM)
tot_events.data = np.ma.masked_where(tot_events.data == 0.0,
tot_events.data)
tot_rain_bins = lshistb * bin_mid3
tot_rain_boxb = tot_rain_bins.collapsed('precipitation_flux',
iris.analysis.SUM) / tot_events
tot_rain_boxa.data = np.ma.masked_where(tot_rain_boxa.data < 1.0,
tot_rain_boxa.data)
tot_rain_boxb.data = np.ma.masked_where(tot_rain_boxb.data < 1.0,
tot_rain_boxb.data)
tot_events = tot_rain_boxa * tot_rain_boxb
# Extract same area as skill is calculated for
tot_events.coord('longitude').circular = True
if tot_events.coord('longitude').bounds is None:
tot_events.coord('longitude').guess_bounds()
tot_events.coord('latitude').guess_bounds()
tot_events1 = tot_events.intersection(ce, ce2)
# Apply land/sea mask to get sea or land only
lsmask = iris.load_cube(maskfile)
lsmask.coord('latitude').coord_system = lshista.coord(
'latitude').coord_system
lsmask.coord('longitude').coord_system = lshista.coord(
'longitude').coord_system
lsmask.coord('longitude').circular = True
if lsmask.coord('longitude').bounds is None:
lsmask.coord('longitude').guess_bounds()
lsmask.coord('latitude').guess_bounds()
lsmasks = lsmask.intersection(ce, ce2)
lsmaskl = lsmask.intersection(ce, ce2)
lsmasks.data = np.ma.masked_where(lsmasks.data >= 0.5, lsmasks.data)
lsmaskl.data = np.ma.masked_where(lsmaskl.data < 0.5, lsmaskl.data)
lsmasksea = lsmasks * tot_events1
lsmaskland = lsmaskl * tot_events1
# Mask the skill cube and calculate index as rms over region
skill.data.mask = tot_events1.data.mask
grid_areas = iris.analysis.cartography.area_weights(skill)
index = skill.collapsed(['latitude', 'longitude'],
iris.analysis.RMS,
weights=grid_areas)
skilltrop = skill.extract(
iris.Constraint(latitude=lambda cell: -15 <= cell <= 15))
grid_areas = iris.analysis.cartography.area_weights(skilltrop)
indextrop = skilltrop.collapsed(['latitude', 'longitude'],
iris.analysis.RMS,
weights=grid_areas)
skillnhml = skill.extract(
iris.Constraint(latitude=lambda cell: 30 <= cell <= 60))
grid_areas = iris.analysis.cartography.area_weights(skillnhml)
indexnhml = skillnhml.collapsed(['latitude', 'longitude'],
iris.analysis.RMS,
weights=grid_areas)
skillshml = skill.extract(
iris.Constraint(latitude=lambda cell: -60 <= cell <= -30))
grid_areas = iris.analysis.cartography.area_weights(skillshml)
indexshml = skillshml.collapsed(['latitude', 'longitude'],
iris.analysis.RMS,
weights=grid_areas)
skill = minf.collapsed('precipitation_flux', iris.analysis.SUM, mdtol=1)
skill.data.mask = lsmasksea.data.mask
skilltst = skill.extract(
iris.Constraint(latitude=lambda cell: -30 <= cell <= 30))
grid_areas = iris.analysis.cartography.area_weights(skilltst)
indexs = skilltst.collapsed(['latitude', 'longitude'],
iris.analysis.RMS,
weights=grid_areas)
skill = minf.collapsed('precipitation_flux', iris.analysis.SUM, mdtol=1)
skill.data.mask = lsmaskland.data.mask
skilltst = skill.extract(
iris.Constraint(latitude=lambda cell: -30 <= cell <= 30))
grid_areas = iris.analysis.cartography.area_weights(skilltst)
indexl = skilltst.collapsed(['latitude', 'longitude'],
iris.analysis.RMS,
weights=grid_areas)
print(
dataname1, dataname2,
'Index = %0.2f, Index (land) = %0.2f, Index (sea) = %0.2f, Index (tropics) = %0.2f, Index (NH mid-lat) = %0.2f, Index (SH mid-lat) = %0.2f'
% (index.data, indexl.data, indexs.data, indextrop.data,
indexnhml.data, indexshml.data))
skill = minf.collapsed('precipitation_flux', iris.analysis.SUM, mdtol=1)
skill.data.mask = tot_events1.data.mask
cf = qplt.pcolormesh(skill, vmin=0.5, vmax=0.95)
ax = plt.gca()
gl = ax.gridlines(draw_labels=True)
gl.xlabels_top = False
ax.coastlines()
plt.title(dataname1 + ' vs ' + dataname2 + ' ' + timescale + ' ' +
seasonc + ' ' + dates + '\nIndex = %0.2f' % (index.data))
plt.savefig(plotname)
plt.close()
return index, indexl, indexs, indextrop, indexnhml, indexshml
def plot_1dhist(plotname,
region,
filenames,
runtitles,
plottitle,
timescale=None,
filenames_obs=None,
runtitles_obs=None,
frac=None,
col_offset=None,
log=None):
"""
Function to plot 1d histograms. Can overplot several (e.g. at different timescale or resolution,
or plot differences between one or more pairs.
Also allows histograms from a different type of dataset (here it's from observations, assuming the
default is models) to be overplotted in the same colours but with dashed lines, for comparison.
NOTE that plotting differences excludes overplotting alternative datasets.
Args:
* plotname:
filename for resulting plot.
* region:
[W,S,E,N] limits of region to average over
* filenames:
list of filenames for files of histograms (output from make_hist_maps)
OR list of LIST PAIRS OF filenames (e.g. [[file1,file2],[file3,file4]]) for differencing (b-a)
in which case filenames_obs etc should be OMITTED.
* runtitles:
list of runtitles for legend (must have title for all filenames)
NOTE this could be dataset names, timescales, whatever is appropriate.
OR list of LIST PAIRS OF runtitles (e.g. [[title1,title2],[title3,title4]]) for differencing, matching the filename pairs (b-a).
* plottitle:
main title for plot
Optional arguments:
* timescale:
If set, adds the timescale to the plot subtitle (best left blank if you are putting timescales in as runtitles)
* filenames_obs:
list of obs filenames for files of histograms (output from make_hist_maps)
* runtitles_obs:
list of obs runtitles (must have title for all obs filenames)
* frac:
If frac is not None, calculates fractional contribution, rather than absolute
* log:
If log is not None, uses log scale for X axis only
* col_offset:
If set, offsets start of colour set by whatever this is set to
"""
# Check that a 4-element region has been entered, if required
if region is None or len(region) != 4:
raise Exception('Please enter a 4-element region!' + str(region))
# Check that all filenames have a runtitle
if len(runtitles) != len(filenames):
raise Exception('Not all run filenames have a runtitle!')
if isinstance(filenames[0], list):
if (not isinstance(runtitles[0],
list)) or (len(runtitles[0]) != len(filenames[0])):
raise Exception('Not all run filenames have a runtitle!')
if filenames_obs is not None:
if runtitles_obs is None:
raise Exception('Please supply runtitles for obs filenames!')
elif len(runtitles_obs) != len(filenames_obs):
raise Exception('Not all obs filenames have a runtitle!')
# Go through filenames reading in cube_of_hist from file and calculating regional averages of the
# histograms of contributions to average rainfall
nice_cmap = plt.get_cmap('brewer_Paired_12')
num_colors = 12
color = iter(nice_cmap(np.linspace(0, 1, num_colors)))
if col_offset is not None:
for j in range(0, col_offset):
c = next(color)
for i, filename in enumerate(filenames):
c = next(color)
if isinstance(filename, str):
cube_of_hist = iris.load_cube(filename)
avg_rain_bins, avg_rain_bins_frac = calc_rain_contr(cube_of_hist)
if frac is not None:
hist1d = calc_1d_contr(avg_rain_bins_frac, region)
else:
hist1d = calc_1d_contr(avg_rain_bins, region)
else:
cube_of_hist_a = iris.load_cube(filename[0])
cube_of_hist_b = iris.load_cube(filename[1])
avg_rain_bins_a, avg_rain_bins_frac_a = calc_rain_contr(
cube_of_hist_a)
avg_rain_bins_b, avg_rain_bins_frac_b = calc_rain_contr(
cube_of_hist_b)
if frac is not None:
hist1da = calc_1d_contr(avg_rain_bins_frac_a, region)
hist1db = calc_1d_contr(avg_rain_bins_frac_b, region)
else:
hist1da = calc_1d_contr(avg_rain_bins_a, region)
hist1db = calc_1d_contr(avg_rain_bins_b, region)
hist1d = hist1db - hist1da
if type(runtitles[i]) is list:
iplt.plot(hist1d[1:90],
label='{} minus {}'.format(runtitles[i][1],
runtitles[i][0]),
color=c,
linewidth=2.5)
else:
iplt.plot(hist1d[1:90], label=runtitles[i], color=c, linewidth=2.5)
# Now add plot titles and legend
a0 = str(int(round(region[0])))
a2 = str(int(round(region[2])))
a1 = str(int(round(region[1])))
a3 = str(int(round(region[3])))
if frac is not None:
if type(filenames[0]) is list:
method = 'DIFFERENCE in Frac'
else:
method = 'Fractional'
units = ''
else:
if isinstance(filenames[0], list):
method = 'DIFFERENCE in Avg'
else:
method = 'Average'
units = 'mm/day'
if timescale is None:
timescale = ''
plt.title(
'{} precip contribution from {} events \n {} \n {} - {}E {} - {}N'.
format(method, timescale, plottitle, a0, a2, a1, a3),
fontsize=12)
plt.ylabel('{} Precip contribution {}'.format(method, units), fontsize=10)
# Now add dashed lines for another type of data (e.g. obs) if supplied
# ONLY if we haven't plotted differences in the above code (indicated by single, not paired, filenames)
if isinstance(filenames[0], str) and filenames_obs is not None:
nice_cmap = plt.get_cmap('brewer_Greys_09')
color = iter(nice_cmap([8, 6, 4, 2]))
linestyle = iter(['dashed', 'dotted', 'solid', 'dashed'])
for i, (v, w) in enumerate(zip(filenames_obs, runtitles_obs)):
c = next(color)
c2 = next(linestyle)
filename = v
runtitleo = w
cube_of_hist = iris.load_cube(filename)
avg_rain_bins, avg_rain_bins_frac = calc_rain_contr(cube_of_hist)
if frac is not None:
hist1d = calc_1d_contr(avg_rain_bins_frac, region)
else:
hist1d = calc_1d_contr(avg_rain_bins, region)
iplt.plot(hist1d[1:90],
label=runtitleo,
color=c,
linewidth=2.5,
linestyle=c2)
# Label axes and add legends
plt.xlabel('Precip bin mm/day', fontsize=10)
if log is not None:
plt.xscale('log')
plt.xlim((0.1, 1000))
plt.legend(ncol=2, fontsize=9, loc='upper left')
else:
plt.xlim((0.1, 400))
plt.legend(ncol=2, fontsize=9, loc='upper right')
# Save the figure to the supplied plot filename
plt.savefig(plotname)
plt.clf()
def plot_histogram_maps(hist_filenames, plotname_root, wk_dir, region, ext,
settings):
"""
Plot histogram maps
"""
hist_filename1 = hist_filenames[0]
hist_filename2 = hist_filenames[1]
ppn_hist_cube1 = make_hist_maps.read_data_cube(hist_filename1)
ppn_hist_cube2 = make_hist_maps.read_data_cube(hist_filename2)
avg_rain_bins_a, avg_rain_bins_frac_a = make_hist_maps.calc_rain_contr(
ppn_hist_cube1)
avg_rain_bins_b, avg_rain_bins_frac_b = make_hist_maps.calc_rain_contr(
ppn_hist_cube2)
ppn_names = make_runtitle([hist_filename1, hist_filename2], settings)
ppn1_name = ppn_names[hist_filename1].replace("_", " ")
ppn2_name = ppn_names[hist_filename2].replace("_", " ")
names = make_runtitle([hist_filename1, hist_filename2],
settings,
model_only=True)
runtitle = "{0} vs {1}".format(names[hist_filename1].replace("_", " "),
names[hist_filename2].replace("_", " "))
# (optional) Define how you want to lump the bins together (below is the default)
all_ppn_bounds = [(0.005, 10.), (10., 50.), (50., 100.), (100., 3000.)]
# Plot as actual contributions for specific region, e.g. 60 to 160E,10S to 10N
desc = {}
plotname = '{0}_actual_contributions{1}'.format(plotname_root, ext)
plotname = os.path.join(wk_dir, plotname)
plot_hist_maps.plot_rain_contr(avg_rain_bins_a,
avg_rain_bins_b,
plotname,
runtitle,
ppn1_name,
ppn2_name,
all_ppn_bounds,
region=region)
desc.update({
os.path.relpath(plotname, start=wk_dir): {
"description":
"Actual contribution of each timescale for region {0}".format(
region)
}
})
# Plot as fractional contributions
plotname = '{0}_fractional_contributions{1}'.format(plotname_root, ext)
plotname = os.path.join(wk_dir, plotname)
plot_hist_maps.plot_rain_contr(avg_rain_bins_frac_a,
avg_rain_bins_frac_b,
plotname,
runtitle,
ppn1_name,
ppn2_name,
all_ppn_bounds,
region=region,
frac=1)
desc.update({
os.path.relpath(plotname, start=wk_dir): {
"description":
"Fractional contribution of each timescale for region {0}".format(
region)
}
})
update_json("plots", desc, wk_dir + "/output.json")
return