alternative='two-sided')
filos.write(
'eof {:3d}: D -> {:8.4f} , pval -> {:12.3e}\n'.format(
eof, D, pval))
elif test == 'Anderson':
D, critvals, pval = stats.anderson_ksamp(
[okpc1[:, eof], okpc2[:, eof]])
filos.write(
'eof {:3d}: D -> {:8.4f} , pval -> {:12.3e}\n'.format(
eof, D, pval))
filos.close()
sys.exit()
# ORA prendo results_ref e faccio anno per anno i miniclusters e faccio il plottino
var_set, dates_set = ctl.seasonal_set(nulabs_ref, results_ref['dates'], 'DJF')
pcs_set, dates_set = ctl.seasonal_set(results_ref['pcs'], results_ref['dates'],
'DJF')
years = [da[0].year for da in dates_set]
# for numy in [6, 10, 20, 30]:
# for reg in range(4):
# figs = []
# #for va, pcs, ye in zip(var_set, pcs_set, years):
# #for cyr in np.arange(len(years))[::10][:-1]:
# for cyr in np.arange(numy/2, len(years)-numy/2-1):
# va = np.concatenate(var_set[cyr-numy/2:cyr+numy/2], axis = 0)
# pcs = np.concatenate(pcs_set[cyr-numy/2:cyr+numy/2], axis = 0)
# ye = years[cyr]-numy/2
# ye2 = years[cyr]+numy/2
#
print(mod)
whos_mod = all_mods == mod
ok_mems = np.sort(all_mems[whos_mod])
bootstraps_all = dict()
for mem in ok_mems:
print(mod, mem)
modmem = mod + '_' + mem
results[modmem]['varopt'] = ctl.calc_varopt_molt(
results[modmem]['pcs'], results[modmem]['centroids'],
results[modmem]['labels'])
results[modmem]['autocorr_wlag'] = ctl.calc_autocorr_wlag(
results[modmem]['pcs'], results[modmem]['dates'], out_lag1=True)
pcs_seas_set, dates_seas_set = ctl.seasonal_set(
results[modmem]['pcs'], results[modmem]['dates'], 'DJF')
labels_seas_set, dates_seas_set = ctl.seasonal_set(
results[modmem]['labels'], results[modmem]['dates'], 'DJF')
n_seas = len(dates_seas_set)
years_set = np.array([dat[0].year for dat in dates_seas_set])
bootstraps = dict()
for nam in allkeysss:
bootstraps[nam] = []
t0 = datetime.now()
for i in range(n_bootstrap):
#if i % 10 == 0:
print(i)
ok_yea = np.sort(np.random.choice(list(range(n_seas)), n_choice))
modnam = 'EC-Earth3P'
yearange = (2015, 2100)
cbar_range = [0., 3.]
cbar_range_notr = [-1, 1]
#############################################################################
var, coords, aux_info = ctl.read_iris_nc(ref_file, extract_level_hPa=500)
lat = coords['lat']
lon = coords['lon']
dates = coords['dates']
var, dates = ctl.sel_time_range(var, dates,
ctl.range_years(yearange[0], yearange[1]))
var_set, dates_set = ctl.seasonal_set(var, dates, 'DJF', seasonal_average=True)
years = np.array([da.year for da in dates_set])
############## PLOT GLOBAL TRENDS ######################
fig, ax = plt.subplots()
glob_mea = ctl.global_mean(var_set, lat)
g0 = glob_mea[0]
m, c = ctl.linear_regre(years, glob_mea)
ax.scatter(years, glob_mea - g0, label='Global', color='blue')
ax.plot(years, c + m * years - g0, color='blue')
var_area, lat_area, lon_area = ctl.sel_area(lat, lon, var_set, 'EAT')
eat_mea = ctl.global_mean(var_area, lat_area)
g0 = eat_mea[0]
m, c = ctl.linear_regre(years, eat_mea)
results_ssp = res['models']
yr0 = 1950
yr1 = 2005
allyr = np.arange(1950, 2005)
yr = allyr
# Erasing incomplete runs
avlen = np.median(
[len(results_hist[ke]['labels']) for ke in results_hist.keys()])
for ke in tuple(results_hist.keys()):
if len(results_hist[ke]['labels']) < avlen - 1000:
del results_hist[ke]
elif len(results_hist[ke]['labels']) > avlen + 100:
# there is some duplicated year
labs, dats = ctl.seasonal_set(results_hist[ke]['labels'],
results_hist[ke]['dates'], None)
pcs, dats = ctl.seasonal_set(results_hist[ke]['pcs'],
results_hist[ke]['dates'], None)
yeas = np.array([da[0].year for da in dats])
labs_ok = []
dats_ok = []
pcs_ok = []
for ye in np.arange(1950, 2005):
okse = np.where(yeas == ye)[0][0]
labs_ok.append(labs[okse])
dats_ok.append(dats[okse])
pcs_ok.append(pcs[okse])
results_hist[ke]['labels'] = np.concatenate(labs_ok)
results_hist[ke]['dates'] = np.concatenate(dats_ok)
results_hist[ke]['pcs'] = np.concatenate(pcs_ok)
for ise, seas in enumerate(seasons):
y1 = 1958
if seas == 'FM': y1 = 1959
print(indexname, seas, y1)
cart_out = cart_out_ind + seas + '/'
if not os.path.exists(cart_out): os.mkdir(cart_out)
# plt.ion()
fig = plt.figure(figsize=(16, 12))
for reg in range(4):
ax = fig.add_subplot(2, 2, reg + 1)
#freq_ok, dates_ok = ctl.sel_season(results_ref['monthly_freq']['freq'][reg], results_ref['monthly_freq']['dates'], seas)
freq_seas, dates_seas = ctl.seasonal_set(
results_ref['monthly_freq']['freq'][reg],
results_ref['monthly_freq']['dates'],
seas,
dates_range=ctl.range_years(y1, 2014))
freq_seas = np.mean(freq_seas, axis=1)
years = np.array([da.year for da in yrdates])
yealen = np.arange(len(years))
# freq = np.array(ctl.running_mean(freq_ok, 15))
freq = np.array(ctl.running_mean(freq_seas, n_yr))
oks = ~np.isnan(freq)
freq = freq[oks]
#amvc = np.array(ctl.running_mean(amv_ref_djf, 15))
print(len(freq), len(years))
years = years[oks]
kwar['detrended_eof_calculation'] = False
kwar['detrended_anom_for_clustering'] = False
kwar['nrsamp_sig'] = 500
results_ref = cd.WRtool_core(var_season,
lat,
lon,
dates_season,
area,
heavy_output=True,
**kwar)
kwar['ref_solver'] = results_ref['solver']
kwar['ref_patterns_area'] = results_ref['cluspattern_area']
kwar['use_reference_eofs'] = True
var_era_set, dates_era_set = ctl.seasonal_set(var_season, dates_season, season)
n_seas = len(var_era_set)
years_set = np.array([dat[0].year for dat in dates_era_set])
# LEGGO SPHINX E LANCIO IL BOOTSTRAP
var_mod = dict()
dates_mod = dict()
all_res = dict()
#for mod in ['base', 'stoc', 'era']:
# print(mod)
# if mod in ['base', 'stoc']:
# var_mod = []
# dates_mod = []
# for i in range(3):
# file_in = cart_in + filenam[mod].format(i)
results_hist[mod] = results_hist['EC-Earth3_r1i1p1f1']
del ece_ssp['EC-Earth3_r1i1p1f1']
results_ssp.update(ece_ssp)
for mod in results_ssp.keys():
print(mod)
if mod not in results_hist:
print('skipping ' + mod)
continue
bau = results_hist[mod]['var_dtr']
bauda = np.arange(1965, 2015)
if len(bau) != len(bauda):
dates_set, _ = ctl.seasonal_set(results_hist[mod]['dates'],
results_hist[mod]['dates'],
'NDJFM',
seasonal_average=False)
bauda = np.array([da[0].year for da in dates_set])
gigi = results_ssp[mod]['var_dtr']
gigida = np.arange(2015, 2100)
if len(gigi) != len(gigida):
dates_set, _ = ctl.seasonal_set(results_ssp[mod]['dates'],
results_ssp[mod]['dates'],
'NDJFM',
seasonal_average=False)
gigida = np.array([da[0].year for da in dates_set])
annette = np.concatenate([bauda, gigida])
cosette = np.concatenate([bau, gigi])
coeffs, covmat = np.polyfit(annette, cosette, deg=3, cov=True)
mem = 'r1'
if na == 'ssp585': mem = 'r4'
fils = np.concatenate(
[glob.glob(filna.format(na, mem, miptab, var)) for var in allvars_2D])
kose = xr.open_mfdataset(fils, use_cftime=True)
kose = kose.drop_vars('time_bnds')
for var in allvars_2D:
print(var)
if var not in kose:
continue
yeamean[(ru, var, 'max')] = ctl.seasonal_set(kose[var],
season='year',
seasonal_stat='max')
if var != 'pr':
cosoye = kose[var].groupby("time.year").mean().compute()
yeamean[(ru, var, 'mean')] = cosoye
yeamean[(ru, var, 'min')] = ctl.seasonal_set(kose[var],
season='year',
seasonal_stat='min')
else:
yeamean[(ru, var, 'sum')] = ctl.seasonal_set(kose[var],
season='year',
seasonal_stat='sum')
pickle.dump(yeamean, open(cart_out + 'bottino_yeastat_tas.p', 'wb'))
allruK = allru
enso_std50 = dict()
enso_abs50 = dict()
enso_yr = dict()
for ru in allruK:
if ind not in ['nam', 'sam']:
# if ru == 'b990':
# years = np.reshape(enso[ru].time.values, (-1, 12))[:, 0].astype(int)
# piuz = np.reshape(enso[ru]['tos'].values, (-1, 12)).mean(axis = 1)
# piuz = xr.DataArray(data = piuz, dims = ['year'], coords = [years])
# else:
piuz = enso[ru].groupby('time.year').mean()
elif ind == 'nam':
piuz = ctl.seasonal_set(enso[ru],
season='NDJFM',
seasonal_stat='mean')
elif ind == 'sam':
piuz = ctl.seasonal_set(enso[ru],
season='MJJAS',
seasonal_stat='mean')
# if ind == 'amv' and ru == 'pi':
# # remove 200yr oscillation
# piuzlow = ctl.lowpass_butter(piuz, 150)
# piuz = piuz-piuzlow
enso_yr[ru] = piuz
enso_std50[ru] = []
enso_abs50[ru] = []
for mod in ece_ssp.keys():
results_hist[mod] = results_hist['EC-Earth3_r1i1p1f1']
del ece_ssp['EC-Earth3_r1i1p1f1']
del ece_ssp_rebase['EC-Earth3_r1i1p1f1']
results_ssp.update(ece_ssp)
results_ssp_rebase.update(ece_ssp_rebase)
# Erasing incomplete runs
for ke in tuple(results_ssp.keys()):
if len(results_ssp[ke]['labels']) < 12000:
del results_ssp[ke]
elif len(results_ssp[ke]['labels']) > 13000:
# there is some duplicated year
for cosone in [results_ssp, results_ssp_rebase]:
labs, dats = ctl.seasonal_set(cosone[ke]['labels'],
cosone[ke]['dates'], None)
pcs, dats = ctl.seasonal_set(cosone[ke]['pcs'],
cosone[ke]['dates'], None)
yeas = np.array([da[0].year for da in dats])
labs_ok = []
dats_ok = []
pcs_ok = []
for ye in np.arange(2015, 2100):
okse = np.where(yeas == ye)[0][0]
labs_ok.append(labs[okse])
dats_ok.append(dats[okse])
pcs_ok.append(pcs[okse])
cosone[ke]['labels'] = np.concatenate(labs_ok)
cosone[ke]['dates'] = np.concatenate(dats_ok)
cosone[ke]['pcs'] = np.concatenate(pcs_ok)
print(mem)
fils = glob.glob(filna.format(exp, mem, miptab, var))
if len(fils) == 0:
print('NO data for {} {}'.format(var, exp, mem))
continue
memok.append(mem)
kose = xr.open_mfdataset(fils, use_cftime=True)
kose = kose.drop_vars('time_bnds')
cosoye = kose[var].groupby("time.year").mean().compute()
yeamean[(exp, mem, var)] = cosoye
for sea in allseasons:
seamean[(exp, mem, var,
sea)] = ctl.seasonal_set(kose[var],
season=sea,
seasonal_stat='mean')
cosoye = np.mean([yeamean[(exp, mem, var)] for mem in memok], axis=0)
yeamean[(exp, 'ensmean', var)] = cosoye
cosoye = np.std([yeamean[(exp, mem, var)] for mem in memok], axis=0)
yeamean[(exp, 'ensstd', var)] = cosoye
yeamean[(exp, 'members')] = memok
pickle.dump(
[yeamean, seamean],
open(cart_out + 'bottino_yeamean_3_{}_{}.p'.format(exp, var),
'wb'))
cart_lui = cart + 'Results_v5_rebase/{}_NDJFM/'.format(area)
freqs, residtimes, patterns = pickle.load(
open(cart_lui + 'allresults_dicts_{}_v3.p'.format(area), 'rb'))
okmods = [
ke[1] for ke in freqs
if 'ssp585' in ke and 'tot50' in ke and 'all' not in ke and 'rel' not in ke
]
print(okmods)
#['BCC-CSM2-MR_r1i1p1f1', 'CanESM5_r1i1p1f1', 'CESM2-WACCM_r1i1p1f1\', 'CNRM-CM6-1_r1i1p1f2', 'CNRM-ESM2-1_r1i1p1f2', 'EC-Earth3_r1i1p1f1', 'FGOALS-g3_r1i1p1f1', 'INM-CM4-8_r1i1p1f1', 'INM-CM5-0_r1i1p1f1', 'IPSL-CM6A-LR_r1i1p1f1', 'MIROC6_r1i1p1f1', 'MPI-ESM1-2-HR_r1i1p1f1', 'MRI-ESM2-0_r1i1p1f1', 'UKESM1-0-LL_r1i1p1f2']
for nu in range(4):
for mod in okmods:
var, dat = ctl.seasonal_set(
results_ssp['EC-Earth3_r1i1p1f1']['pcs'][:, 0],
results_ssp['EC-Earth3_r1i1p1f1']['dates'],
'DJF',
seasonal_average=True)
years = np.array([da.year for da in dat])
m, c, err_m, err_c = ctl.linear_regre_witherr(years, var)
daynum = np.arange(len(results_ssp['EC-Earth3_r1i1p1f1']['dates']))
m, c, err_m, err_c = ctl.linear_regre_witherr(
daynum, results_ssp['EC-Earth3_r1i1p1f1']['pcs'])
cartmon_hist = '/data-hobbes/fabiano/CMIP6/historical_mon_zg/'
cartmon_ssp = '/data-hobbes/fabiano/CMIP6/ssp585_mon_zg/'
filssp = 'zg_Amon_ssp585_{}_{}_2015-2100.nc'
filhist = 'zg_{}_mon.nc'
print(seas)
y1 = 1957
if seas == 'FM': y1 = 1958
print(indexname, seas, y1)
cart_out = cart_out_ind + seas + '/'
if not os.path.exists(cart_out): os.mkdir(cart_out)
# plt.ion()
fig = plt.figure(figsize=(16, 12))
for reg in range(4):
ax = fig.add_subplot(2, 2, reg + 1)
#freq_ok, dates_ok = ctl.sel_season(results_ref['monthly_freq']['freq'][reg], results_ref['monthly_freq']['dates'], seas)
freq_seas, dates_seas = ctl.seasonal_set(
results[modmem]['monthly_freq']['freq'][reg],
results[modmem]['monthly_freq']['dates'],
seas,
dates_range=ctl.range_years(y1, y2))
freq_seas = np.mean(freq_seas, axis=1)
if len(freq_seas) < len(amv_ref_yr):
print('aaaaaaaaaaaa', len(freq_seas), len(amv_ref_yr))
continue
years = np.array([da.year for da in yrdates])
yealen = np.arange(len(years))
# freq = np.array(ctl.running_mean(freq_ok, 15))
freq = np.array(ctl.running_mean(freq_seas, n_yr))
oks = ~np.isnan(freq)
freq = freq[oks]
#fils = np.concatenate([glob.glob(filna.format(na, mem, miptab, var)) for var in allvars_2D[:-1]])
fils = glob.glob(filna.format(na, mem, miptab, var))
if len(fils) == 0:
print('NO data for {} {}'.format(var, ru))
continue
kose = xr.open_mfdataset(fils, use_cftime=True)
kose = kose.drop_vars('time_bnds')
cosoye = kose[var].groupby("time.year").mean().compute()
yeamean[(ru, var)] = cosoye
for sea in allseasons:
seamean[(ru, var, sea)] = ctl.seasonal_set(kose[var],
season=sea,
seasonal_stat='mean')
pickle.dump([yeamean, seamean],
open(cart_out + 'bottino_yeamean_3_{}.p'.format(var), 'wb'))
# 3D vars
for var in allvars_3D:
yeamean = dict()
seamean = dict()
print(var)
for na, ru, col in zip(allnams, allru, colors):
print(ru)
mem = 'r1'
if na == 'ssp585': mem = 'r4'
