print('BAUUUUUUUUUUUUUUUUUUUUUUUUU')
## plots
runfreq = dict()
seasfreq = dict()
fig = plt.figure(figsize=(16, 12))
for reg in range(4):
ax = fig.add_subplot(2, 2, reg + 1)
cosi = []
for mem in okmods:
labok, datok = ctl.sel_time_range(results_hist[mem]['labels'],
results_hist[mem]['dates'],
ctl.range_years(yr0, yr1))
seasfr, yr = ctl.calc_seasonal_clus_freq(labok, datok, numclus)
seasfreq[('hist_cmip5', mem, reg)] = seasfr[reg, :]
seas20 = np.array(ctl.running_mean(seasfr[reg, :], 20))
print(mem, len(seas20))
if len(seas20) == len(yr):
ax.plot(yr, seas20)
cosi.append(seas20)
else:
print(mem, len(seas20), 'too short')
coso = np.mean(cosi, axis=0)
runfreq[('hist_cmip5', reg)] = coso
ax.plot(yr, coso, color='black', linewidth=3)
ax.set_title(reg_names_area[area][reg])
fig.savefig(cart_out + 'models_run20_{}_hist.pdf'.format(area))
reg_names = reg_names_area[area]
runfreq = dict()
seasfreq = dict()
figs = []
for tip, col in zip(alltips, colorz):
okmods = resdict[tip].keys()
fig = plt.figure(figsize=(16, 12))
for reg in range(4):
ax = fig.add_subplot(2, 2, reg + 1)
cosi = []
for mem in okmods:
seasfr, yr = ctl.calc_seasonal_clus_freq(
resdict[tip][mem]['labels'], resdict[tip][mem]['dates'],
numclus)
seasfreq[(tip, mem, reg)] = seasfr[reg, :]
seas1 = np.array(ctl.running_mean(seasfr[reg, :], 20))
runfreq[(tip, mem, reg)] = seas1
ax.plot(yr, seas1, label=tip, color=col)
ax.set_title(reg_names_area[area][reg])
#ax.legend()
fig.suptitle(tip)
figs.append(fig)
ctl.plot_pdfpages(cart_out + 'check_single_members_{}.pdf'.format(area),
figs)
fig = plt.figure(figsize=(16, 12))
for reg in range(4):
ax = fig.add_subplot(2, 2, reg + 1)
for tip, col in zip(alltips, colorz):
okmods = resdict[tip].keys()
print(okmods)
print(len(okmods))
runfreq = dict()
seasfreq = dict()
fig = plt.figure(figsize=(16, 12))
for reg in range(4):
ax = fig.add_subplot(2, 2, reg + 1)
cosi = []
for mem in okmods:
seasfr, yr = ctl.calc_seasonal_clus_freq(
results_hist[mem]['labels'], results_hist[mem]['dates'],
numclus)
seasfreq[('hist', mem, reg)] = seasfr[reg, :]
seas20 = np.array(ctl.running_mean(seasfr[reg, :], 20))
ax.plot(yr, seas20)
cosi.append(seas20)
coso = np.mean(cosi, axis=0)
runfreq[('hist', reg)] = coso
ax.plot(yr, coso, color='black', linewidth=3)
ax.set_title(reg_names_area[area][reg])
fig.savefig(cart_out + 'models_run20_{}_hist.pdf'.format(area))
reg_names = reg_names_area[area]
seasfr, yr_ref = ctl.calc_seasonal_clus_freq(results_ref['labels'],
results_ref['dates'], numclus)
for reg in range(4):
seasfreq[('hist', 'ref', reg)] = seasfr[reg, :]
okmods_hist = list(results_hist.keys())
okmods_ssp = list(results_ssp[ssp].keys())
okmods = okmods_ssp
runfreq = dict()
seasfreq = dict()
fig = plt.figure(figsize = (16,12))
for reg in range(4):
ax = fig.add_subplot(2, 2, reg+1)
cosi = []
for mem in okmods_hist:
seasfr, yr = ctl.calc_seasonal_clus_freq(results_hist[mem]['labels'], results_hist[mem]['dates'], numclus)
seasfreq[('hist', mem, reg)] = seasfr[reg, :]
seas10 = np.array(ctl.running_mean(seasfr[reg, :], yr10))
ax.plot(yr, seas10)
cosi.append(seas10)
coso = np.mean(cosi, axis = 0)
runfreq[('hist', reg)] = coso
ax.plot(yr, coso, color = 'black', linewidth = 3)
ax.set_title(reg_names_area[area][reg])
fig.savefig(cart_out + 'models_freq10_{}_hist.pdf'.format(area, ssp))
trend_ssp = dict()
residtime_ssp = dict()
for ssp in allssps:
fig = plt.figure(figsize = (16,12))
for reg in range(4):
if 'freq_clus_seasonal' not in resu_ref:
resu_ref['freq_clus_seasonal'], resu_ref[
'freq_clus_seasonal_years'] = ctl.calc_seasonal_clus_freq(
resu_ref['labels'], resu_ref['dates'], 4)
#gigi_obs = xr.DataArray(resu_ref['freq_clus_seasonal'], dims = ('reg', 'time'), coords = {'reg': [0, 1, 2, 3], 'time': resu_ref['freq_clus_seasonal_years']})
fig = plt.figure(figsize=(16, 12))
axes = []
for num, patt in enumerate(patnames[seas]):
ax = plt.subplot(2, 2, num + 1)
for ke, col in zip(keall, colorz):
timeseries[ke].sel(reg=num).plot(ax=ax, color=col, linewidth=0.2)
rupi = ctl.running_mean(timeseries[ke].sel(reg=num).values, 10)
ax.plot(timeseries[ke].time,
rupi,
color=col,
linewidth=3,
label=ke)
ax.set_title(patt, fontsize=16)
axes.append(ax)
ax.grid()
if num in [2, 3]:
ax.set_xlabel('year')
else:
ax.set_xlabel(None)
if num == 1:
ax.legend()
# plot resid times w std dev
for mod in all_res:
axes = []
fig = plt.figure()
for j in range(kwar['numclus']):
ax = fig.add_subplot(2, 2, j + 1)
ax.set_title(pattnames[j])
histomea = np.mean(
[coso['histo_resid_times'][j] for coso in all_res[mod]], axis=0)
histostd = np.std(
[coso['histo_resid_times'][j] for coso in all_res[mod]], axis=0)
ax.bar(numarr, histomea, alpha=0.5, label='mean', color='indianred')
cosomea = ctl.running_mean(histomea[:-1], 3)
ax.bar(numarr, histostd, alpha=0.5, label='stddev', color='steelblue')
cosostd = ctl.running_mean(histostd[:-1], 3)
ax.plot(numarr[:-1], cosomea, color='indianred')
ax.plot(numarr[:-1], cosostd, color='steelblue')
ax.legend()
ax.set_xlim(0, max_days + 2)
tics = np.arange(0, max_days + 2, 5)
labs = ['{}'.format(ti) for ti in tics[:-1]]
labs.append('>{}'.format(max_days))
ax.set_xticks(tics, minor=False)
ax.set_xticklabels(labs, size='small')
ax.set_xlabel('Days')
ax.set_ylabel('Frequency')
ax.set_ylim(0, 0.2)
spect_mean = np.mean(allcose, axis=0)
spect_q1 = np.percentile(allcose, 25, axis=0)
spect_q3 = np.percentile(allcose, 75, axis=0)
if ind == 'enso':
zuc = np.sum(spect_mean)
spect_mean /= zuc
spect_q1 /= zuc
spect_q3 /= zuc
if ind == 'enso':
ruwi = 3
else:
ruwi = 5
spect_mean = ctl.running_mean(spect_mean, ruwi)
spect_q1 = ctl.running_mean(spect_q1, ruwi)
spect_q3 = ctl.running_mean(spect_q3, ruwi)
#due giri di smoothing
spect_mean = ctl.running_mean(spect_mean, ruwi)
spect_q1 = ctl.running_mean(spect_q1, ruwi)
spect_q3 = ctl.running_mean(spect_q3, ruwi)
# if ind == 'enso':
# thres = 10
# else:
# thres = 50
# ax.fill_between(frme, spect_q1, spect_q3, color = col, alpha = 0.3)
# ax.plot(frme, spect_mean, color = col, lw = 2)
cb.set_label('Global net TOA (W/m2)')
fig.savefig(cart_out + 'amvperc_vs_tasytoa_50.pdf')
fig, axs = plt.subplots(2, 2, figsize=(16, 12))
for ru, ax in zip(allru, axs.flatten()):
piuz = enso[ru]['tos'].groupby('time.year').mean()
data = piuz.values.flatten()
ps = np.abs(np.fft.fft(data))**2
freqs = np.fft.fftfreq(data.size, 1)
idx = freqs > 0
ax.plot(1 / freqs[idx], ps[idx], linewidth=0.5)
#ps_low = ctl.butter_filter(ps, 20)
ps_low = ctl.running_mean(ps, 20)
ax.plot(1 / freqs[idx], ps_low[idx], linewidth=2)
#ps_low = ctl.butter_filter(ps, 50)
ps_low = ctl.running_mean(ps, 50)
ax.plot(1 / freqs[idx], ps_low[idx], linewidth=2)
ax.set_title(ru)
if ax in axs[1, :]:
ax.set_xlabel('period (yr)')
#ax.set_ylabel('amplitude')
ax.set_xlim(2, 50)
#ctl.adjust_ax_scale(axs.flatten())
fig.savefig(cart_out + 'amv_spectra_all500.pdf')
#frbins = [0, 5, 10, 20, 30, 50, 100, 200]
fig = plt.figure()
for ens in stoc_ens:
sig = np.array([results[(ens,area,ran)]['significance'] for ran in yr_ranges])
plt.plot(cyea, sig, label = ens)
# plt.scatter(cyea, sig, color = col, marker = sym, label = ens)
plt.legend()
plt.ylabel('Significance')
plt.xlabel('central year of 30yr period')
plt.title(area+' - stoc runs')
fig.savefig(cartsig+'Sig_{}_stoc_5yr.pdf'.format(area))
fig = plt.figure()
for ens in base_ens:
sig = np.array([results[(ens,area,ran)]['significance'] for ran in yr_ranges])
sig = ctl.running_mean(sig, 5)
plt.plot(cyea, sig, label = ens)
# plt.scatter(cyea, sig, color = col, marker = sym, label = ens)
plt.legend()
plt.ylabel('Significance')
plt.xlabel('central year of 30yr period')
plt.title(area+' - base runs (25 yr smooth)')
fig.savefig(cartsig+'Sig_{}_base_20yr_smooth.pdf'.format(area))
fig = plt.figure()
for ens in stoc_ens:
sig = np.array([results[(ens,area,ran)]['significance'] for ran in yr_ranges])
sig = ctl.running_mean(sig, 5)
plt.plot(cyea, sig, label = ens)
# plt.scatter(cyea, sig, color = col, marker = sym, label = ens)
plt.legend()
ref_solver=ref_solver,
ref_patterns_area=ref_patterns_area,
detrended_eof_calculation=True,
detrended_anom_for_clustering=False,
heavy_output=False)
freq1 = ctl.calc_seasonal_clus_freq(resu1['labels'], resu1['dates'])
freq2 = ctl.calc_seasonal_clus_freq(resu2['labels'], resu2['dates'])
years = np.unique(pd.to_datetime(resu1['dates']).year)[:-1]
patnames = ['NAO +', 'Blocking', 'NAO -', 'Atl. Ridge']
fig = plt.figure()
plt.ylim(15., 35.)
plt.title('detr.EOFs + detrended anomalies')
for clu, clunam in enumerate(patnames):
smut = ctl.running_mean(freq1[:, clu], wnd=30)
plt.plot(years, smut, label=clunam)
fig = plt.figure()
plt.ylim(20., 30.)
plt.title('detr.EOFs + non-detrended anomalies')
for clu, clunam in enumerate(patnames):
smut = ctl.running_mean(freq2[:, clu], wnd=30)
plt.plot(years, smut, label=clunam)
trans_matrix = ctl.calc_regime_transmatrix(1, resu1['labels'], dates_season)
print(trans_matrix)
filter = resu1['dist_centroid'] < np.percentile(resu1['dist_centroid'], 70)
trans_pcs = ctl.find_transition_pcs(1,
resu1['labels'][filter],
dates_season[filter],
cose['fdNAO50'] = freqs[(ssp, mod, 'tot50')][0] - freqs[
('hist', mod, 'tot50')][0]
cose['fdSBL50'] = freqs[(ssp, mod, 'tot50')][1] - freqs[
('hist', mod, 'tot50')][1]
cose['fdAR50'] = freqs[(ssp, mod, 'tot50')][2] - freqs[
('hist', mod, 'tot50')][2]
cose['fdNAOneg50'] = freqs[(ssp, mod, 'tot50')][3] - freqs[
('hist', mod, 'tot50')][3]
cose['fdNAO50_divDT'] = cose['fdNAO50'] / cose['deltaT']
cose['fdSBL50_divDT'] = cose['fdSBL50'] / cose['deltaT']
cose['fdAR50_divDT'] = cose['fdAR50'] / cose['deltaT']
cose['fdNAOneg50_divDT'] = cose['fdNAOneg50'] / cose['deltaT']
gigi = ctl.running_mean(seasfreq[(ssp, mod, 0)],
yr10,
remove_nans=True)
pio = mk.original_test(gigi)
cose['trendNAO_pval'] = pio.p
if pio.h:
trendo = pio.trend
nutrend = pio.trend
ima = -1
while nutrend == trendo:
gigi = ctl.running_mean(seasfreq[(ssp, mod, 0)][:ima],
yr10,
remove_nans=True)
pio = mk.original_test(gigi)
nutrend = pio.trend
linewidth=1.5,
color=col[exp][1])
axtot2.scatter(x[50:], y[50:], label=exp, color=col[exp][0], s=2)
axtot2.scatter(x[:50], y[:50], color=col[exp][0], s=8, marker='*')
axtot2.plot(xlin,
xlin * linefirst[0] + linefirst[1],
label='slope = {:8.3f} +/- {:5.2f}'.format(
linefirst[0], linefirst[2]),
linewidth=1.5,
color=col[exp][1])
ax1 = fig2.add_subplot(1, 2, i + 1)
print(i)
ax1.scatter(time, x, label='tas', color='lightsteelblue', s=3)
coso = ctl.running_mean(x, 10)
ax1.plot(time, coso, color='steelblue')
axes1.append(ax1)
ax1.set_title(exp)
ax2 = ax1.twinx()
ax2.scatter(time, y, label=var2, color='lightcoral', s=3)
coso = ctl.running_mean(y, 10)
ax2.plot(time, coso, color='indianred')
axes2.append(ax2)
#for i, (xli, line) in enumerate(zip(xlins, lines)):
# ax.plot(xli, xli*line[0]+line[1], label='y{} = {:9.2e} x + {:9.2e}'.format(i, line[0], line[1]), linewidth = 1.5)
ax1.legend(fontsize='small', loc=3)
ax2.legend(fontsize='small', loc=4)
#axes.append(ax)
oksiarea = oksi * areaok[np.newaxis, :]
seaicearea = np.nansum(oksiarea, axis=1)
dates = np.array(gigi.time.data)
okmarch = np.array([da.month == 3 for da in dates])
oksept = np.array([da.month == 9 for da in dates])
resdict[(ru, varnam, 'glomean', 'mar')] = seaicearea[okmarch]
resdict[(ru, varnam, 'glomean', 'sep')] = seaicearea[oksept]
if ru != 'pi':
yeaok = np.array([da.year for da in dates])
else:
yeaok = np.array([da.year - 2256 + 2015 for da in dates])
sima10 = ctl.running_mean(seaicearea[okmarch], 10)
sise10 = ctl.running_mean(seaicearea[oksept], 10)
axs[ii, 0].plot(yeaok[okmarch],
sima10,
linestyle='solid',
marker='None',
color=col,
label=ru,
linewidth=2)
axs[ii, 1].plot(yeaok[oksept],
sise10,
linestyle='solid',
marker='None',
color=col,
label=ru,
for indexname, file_ref, file_mod, filogen, reg_names, n_yr in zip(
indexes, file_refs, file_mods, filogens, reg_names_all, n_yrs):
cart_out_ind = cart_out_all + indexname + '/'
if not os.path.exists(cart_out_ind): os.mkdir(cart_out_ind)
results, results_ref = pickle.load(open(filogen, 'rb'))
amv_ref, dates = read_file_index(file_ref)
amv_ref, dates = ctl.sel_time_range(amv_ref, dates,
ctl.range_years(1957, 2013))
# Yearly amv index
amv_ref_yr, yrdates = ctl.yearly_average(amv_ref, dates)
amv_ref_yr = np.squeeze(amv_ref_yr)
# amv_ref_djf, dates_djf = ctl.sel_season(amv_ref, dates, seas)
# amv_ref_djf = np.squeeze(amv_ref_djf)
amv_nyr = np.array(ctl.running_mean(amv_ref_yr, n_yr))
amvc = amv_nyr[~np.isnan(amv_nyr)]
fig_all = plt.figure(figsize=(16, 12))
ax_all = []
for i in range(4):
ax_all.append(fig_all.add_subplot(2, 2, i + 1))
fig_all_ul = plt.figure(figsize=(16, 12))
ax_all_ul = []
for i in range(4):
ax_all_ul.append(fig_all_ul.add_subplot(2, 2, i + 1))
for ise, seas in enumerate(seasons):
y1 = 1958
if seas == 'FM': y1 = 1959
lats_N = np.stack(lats_N)
lats_S = np.stack(lats_S)
percs = [10, 25, 50, 75, 90]
lindict = dict()
for prc in percs:
lindict[('N', prc)] = np.percentile(lats_N, prc, axis=1).squeeze()
lindict[('S', prc)] = np.percentile(lats_S, prc, axis=1).squeeze()
lons = mapa['lon'].values
for pol in ['N', 'S']:
for prc in percs:
lindict[(pol, prc)] = ctl.running_mean(lindict[(pol, prc)],
5,
cyclic=True)
lindict[(pol, prc)], lonu = cutil.add_cyclic_point(lindict[(pol,
prc)],
coord=lons)
# spd_N = np.array(spd_N)
# spd_S = np.array(spd_S)
# lats_N_smo[spd_N < 5] = np.nan
# lats_S_smo[spd_S < 5] = np.nan
for pol in ['N', 'S']:
ax.fill_between(lonu,
lindict[(pol, 10)],
lindict[(pol, 90)],
n_color_levels=21,
draw_contour_lines=False,
n_lines=5,
color_percentiles=(0, 100),
bounding_lat=30,
plot_margins='EAT',
add_rectangles=None,
draw_grid=True,
plot_type='filled_contour',
verbose=False,
lw_contour=0.5)
figs.append(fig)
ctl.plot_pdfpages(cart_out_orig + 'check_climate_mean_allmem.pdf', figs)
fig = plt.figure()
noreb = ctl.running_mean(resssp_noreb['EC-Earth3_r1i1p1f1']['pcs'][:, 0], 10)
reb = ctl.running_mean(resssp['EC-Earth3_r1i1p1f1']['pcs'][:, 0], 10)
plt.plot(noreb, label='rebase_ssp')
plt.plot(reb, label='rebase_hist')
plt.plot(reb - noreb, label='diff')
fig.savefig(cart_out_orig + 'check_first_eof_vs_rebase.pdf')
fig = plt.figure()
noreb = ctl.running_mean(resssp_noreb['EC-Earth3_r1i1p1f1']['pcs'][:300, 0],
10)
reb = ctl.running_mean(resssp['EC-Earth3_r1i1p1f1']['pcs'][:300, 0], 10)
plt.plot(noreb, label='rebase_ssp')
plt.plot(reb, label='rebase_hist')
plt.plot(reb - noreb, label='diff')
fig.savefig(cart_out_orig + 'check_first_eof_vs_rebase_zoom.pdf')
print(okmods)
print(len(okmods))
runfreq = dict()
seasfreq = dict()
fig = plt.figure(figsize=(16, 12))
for reg in range(4):
ax = fig.add_subplot(2, 2, reg + 1)
cosi = []
for mem in okmods:
seasfr, yr = ctl.calc_seasonal_clus_freq(
results_hist[mem]['labels'], results_hist[mem]['dates'],
numclus)
seasfreq[('hist', mem, reg)] = seasfr[reg, :]
seas10 = np.array(ctl.running_mean(seasfr[reg, :], yr10))
ax.plot(yr, seas10)
cosi.append(seas10)
coso = np.mean(cosi, axis=0)
runfreq[('hist', reg)] = coso
ax.plot(yr, coso, color='black', linewidth=3)
ax.set_title(reg_names_area[area][reg])
fig.savefig(cart_out + 'models_freq10_{}_hist.pdf'.format(area, ssp))
trend_ssp = dict()
residtime_ssp = dict()
for ssp in allssps:
fig = plt.figure(figsize=(16, 12))
for reg in range(4):
ctl.plot_pdfpages(figure_file, figures)
plt.close('all')
figure_file = cart_out + 'global_diff_net_fluxes.pdf'
figures = []
axes = []
fig = plt.figure()
ax = plt.subplot(1, 1, 1)
plt.title('Net flux at TOA')
for lett, tim in zip(['a', 'f'], ['_hist', '_fut']):
for cosone, ls in zip(['stoc', 'base'], [':', '--']):
coso = radclim[('global', cosone + tim, 'toa_balance')]
ax.plot(years[lett], coso, linewidth=0.5, color='grey', linestyle=ls)
rollcoso = ctl.running_mean(coso, wnd=10)
ax.plot(years[lett], rollcoso, label=cosone + tim, linewidth=2.0)
plt.xlabel('Year')
plt.ylabel('Rad. forcing (W/m^2)')
plt.legend()
plt.grid()
figures.append(fig)
axes.append(ax)
# fig = plt.figure()
# ax = plt.subplot(1,1,1)
# cosone = 'diff s-b'
# coso = radclim[('global', 'stoc', 'toa_balance')]-radclim[('global', 'base', 'toa_balance')]
# plt.title('Net flux at TOA - ({})'.format(cosone))
# ax.plot(years, coso, label = 'base', linewidth = 0.5, color = 'grey')
# rollcoso = ctl.running_mean(coso, wnd = 10)
cosopi = yeamean[('pi', var)]
# cosohist = yeamean[('hist', var)]
# cosossp = yeamean[('ssp585', var)]
cosohist = yeamean[('historical_mean', var)]
cosossp = yeamean[('ssp585_mean', var)]
#cosohistssp = xr.concat([cosohist, cosossp], dim = 'year')
cosohistssp = np.concatenate([cosohist, cosossp], axis=0)
yeahissp = np.arange(1970, 2101)
fact = 1
if varnam == 'pr':
fact = 60 * 60 * 24 * 365
pimap = fact * cosopi.mean('year').values
pirun = ctl.running_mean(cosopi, 50)
pistd = (cosopi - pirun).std('year').values
if varnam == 'pr':
thres = pimap < 50.
fig_patt_trans = []
fig_patt_stab = []
fig_abs_start = []
fig_abs_end = []
fig_ratio = []
fig_std = []
for ru in allru[1:]:
coso = yeamean[(ru, var)]
