def ens_anom(filenames, dir_output, name_outputs, varname, numens, season,
area, extreme):
"""Ensemble anomalies.
Computation of the ensemble anomalies based on the desired value
from the input variable (it can be the percentile, mean, maximum, standard
deviation or trend)
OUTPUT: NetCDF files of ensemble mean of climatology, selected value and
anomaly maps.
"""
print('The name of the output files will be <variable>_{0}.txt'.format(
name_outputs))
print('Number of ensemble members: {0}'.format(numens))
outfiles = []
# Reading the netCDF file of 3Dfield, for all the ensemble members
var_ens = []
for ens in range(numens):
ifile = filenames[ens]
# print('ENSEMBLE MEMBER %s' %ens)
var, varunits, lat, lon, dates, _ = read_iris(ifile)
# Convertion from kg m-2 s-1 to mm/day
if varunits == 'kg m-2 s-1':
var = var * 86400 # there are 86400 seconds in a day
varunits = 'mm/day'
# Selecting a season (DJF,DJFM,NDJFM,JJA)
var_season, _ = sel_season(var, dates, season)
# Selecting only [latS-latN, lonW-lonE] box region
var_area, lat_area, lon_area = sel_area(lat, lon, var_season, area)
var_ens.append(var_area)
if varunits == 'kg m-2 s-1':
print('\nPrecipitation rate units were converted from kg m-2 s-1 '
'to mm/day')
print('The variable is {0} ({1})'.format(varname, varunits))
print('Original var shape: (time x lat x lon)={0}'.format(var.shape))
print('var shape after selecting season {0} and area {1}: '
'(time x lat x lon)={2}'.format(season, area, var_area.shape))
if extreme == 'mean':
# Compute the time mean over the entire period, for each ens member
varextreme_ens = [
np.nanmean(var_ens[i], axis=0) for i in range(numens)
]
elif len(extreme.split("_")) == 2:
# Compute the chosen percentile over the period, for each ens member
quant = int(extreme.partition("th")[0])
varextreme_ens = [
np.nanpercentile(var_ens[i], quant, axis=0) for i in range(numens)
]
elif extreme == 'maximum':
# Compute the maximum value over the period, for each ensemble member
varextreme_ens = [np.nanmax(var_ens[i], axis=0) for i in range(numens)]
elif extreme == 'std':
# Compute the standard deviation over the period, for each ens member
varextreme_ens = [np.nanstd(var_ens[i], axis=0) for i in range(numens)]
elif extreme == 'trend':
# Compute the linear trend over the period, for each ensemble member
trendmap = np.empty((var_ens[0].shape[1], var_ens[0].shape[2]))
trendmap_ens = []
for i in range(numens):
for jla in range(var_ens[0].shape[1]):
for jlo in range(var_ens[0].shape[2]):
slope, _, _, _, _ = \
stats.linregress(range(var_ens[0].shape[0]),
var_ens[i][:, jla, jlo])
trendmap[jla, jlo] = slope
trendmap_ens.append(trendmap.copy())
varextreme_ens = trendmap_ens
varextreme_ens_np = np.array(varextreme_ens)
print('Anomalies are computed with respect to the {0}'.format(extreme))
# Compute and save the anomalies with respect to the ensemble
ens_anomalies = varextreme_ens_np - np.nanmean(varextreme_ens_np, axis=0)
varsave = 'ens_anomalies'
ofile = os.path.join(dir_output,
'ens_anomalies_{0}.nc'.format(name_outputs))
# print(ofile)
print('ens_anomalies shape: (numens x lat x lon)={0}'.format(
ens_anomalies.shape))
save_n_2d_fields(lat_area, lon_area, ens_anomalies, varsave, varunits,
ofile)
outfiles.append(ofile)
# Compute and save the climatology
vartimemean_ens = [np.mean(var_ens[i], axis=0) for i in range(numens)]
ens_climatologies = np.array(vartimemean_ens)
varsave = 'ens_climatologies'
ofile = os.path.join(dir_output,
'ens_climatologies_{0}.nc'.format(name_outputs))
save_n_2d_fields(lat_area, lon_area, ens_climatologies, varsave, varunits,
ofile)
outfiles.append(ofile)
ens_extreme = varextreme_ens_np
varsave = 'ens_extreme'
ofile = os.path.join(dir_output, 'ens_extreme_{0}.nc'.format(name_outputs))
save_n_2d_fields(lat_area, lon_area, ens_extreme, varsave, varunits, ofile)
outfiles.append(ofile)
return outfiles
climate_mean.append(np.mean(var_ok[mask, :, :], axis=0))
climat_std.append(np.std(var_ok[mask, :, :], axis=0))
dates_clim.append(
pd.to_datetime('2000{:02d}01'.format(mon), format='%Y%m%d'))
climate_mean = np.array(climate_mean)
climat_std = np.array(climat_std)
var_anom = []
for el, dat in zip(var, dates_pdh):
anom = el - climate_mean[dat.month - 1, :, :]
var_anom.append(anom)
var_anom = np.array(var_anom)
var_season, dates_season = sel_season(var_anom, dates, season)
dates_seas_pdh = pd.to_datetime(dates_season)
years = np.unique(dates_seas_pdh.year)
var_season_tot = []
if months[0] == 12:
years == years[1:]
for yea in years:
if months[0] == 12:
init = pd.to_datetime('{:04d}{:02d}01'.format(yea - 1, months[0]),
format='%Y%m%d')
else:
init = pd.to_datetime('{:04d}{:02d}01'.format(yea, months[0]),
format='%Y%m%d')
####################### PRECOMPUTATION #######################################
#____________run ens_anom as a module
varextreme_ens_np, vartimemean_ens, ensemble_mean, dates = ens_anom(inputs)
dates_pdh = pd.to_datetime(dates)
climatology = None
if inputs['clim_compare']:
if inputs['climat_file'] is None:
raise ValueError('climat_file not specified')
var, lat, lon, dates_clim, time_units, var_units = read3Dncfield(
inputs['climat_file'])
if season is not None:
var_season, dates_season = sel_season(var, dates_clim, season,
inputs['timestep'])
else:
var_season = var
dates_season = dates_clim
climatology_tot, _, _ = sel_area(lat, lon, var_season, inputs['area'])
climatology = np.mean(climatology_tot, axis=0)
if inputs['clim_sigma_value'] is None and inputs['climat_std'] is not None:
var, lat, lon, dates_clim, time_units, var_units = read3Dncfield(
inputs['climat_std'])
if season is not None:
var_season, dates_season = sel_season(var, dates_clim, season,
inputs['timestep'])
else:
def ens_anom(inputs):
'''
\nGOAL: Computation of the ensemble anomalies based on the desired value from the input variable
(it can be the percentile, mean, maximum, standard deviation or trend)
OUTPUT: NetCDF files of ensemble mean of climatology, selected value and anomaly maps.
'''
# User-defined packages
from read_netcdf import read3Dncfield, save_N_2Dfields, readxDncfield
from sel_season_area import sel_season, sel_area
OUTPUTdir = inputs['OUTPUTdir']
numens = inputs['numens']
name_outputs = inputs['name_outputs']
filenames = inputs['filenames']
season = inputs['season']
area = inputs['area']
varname = inputs['varname']
extreme = inputs['extreme']
timestep = inputs['timestep']
print('The name of the output files will be <variable>_{0}.txt'.format(
name_outputs))
print('Number of ensemble members: {0}'.format(numens))
#____________Reading the netCDF file of 3Dfield, for all the ensemble members
var_ens = []
for ens in range(numens):
ifile = filenames[ens]
print('ENSEMBLE MEMBER %s' % ens)
#var, lat, lon, dates, time_units, varunits = read3Dncfield(ifile)
print(ifile)
varss, lat, lon, dates, time_units, varunitsss, time_cal = readxDncfield(
ifile)
print(varss.keys())
print(varunitsss)
if len(varss.keys()) == 1:
var = varss.values()[0]
varunits = varunitsss.values()[0]
else:
try:
var = varss[varname]
varunits = varunitsss[varname]
except KeyError as ke:
print(
'Unable to find right variable among the following: {}\n'.
format(varss.keys()))
raise ke
#____________Convertion from kg m-2 s-1 to mm/day
if varunits == 'kg m-2 s-1':
var = var * 86400 #there are 86400 seconds in a day
varunitsnew = 'mm/day'
else:
varunitsnew = varunits
#____________Selecting a season (DJF,DJFM,NDJFM,JJA)
if season is not None:
var_season, dates_season = sel_season(var, dates, season, timestep)
else:
var_season = var
dates_season = dates
#____________Selecting only [latS-latN, lonW-lonE] box region
var_area, lat_area, lon_area = sel_area(lat, lon, var_season, area)
var_ens.append(var_area)
if varunitsnew == 'mm/day':
print(
'\nPrecipitation rate units are converted from kg m-2 s-1 to mm/day'
)
print('The variable is {0} ({1})'.format(varname, varunitsnew))
print('Original var shape: (time x lat x lon)={0}'.format(var.shape))
print(
'var shape after selecting season {0}: (time x lat x lon)={1}'.format(
season, var_season.shape))
print(
'var shape after selecting season {0} and area {1}: (time x lat x lon)={2}'
.format(season, area, var_area.shape))
print('Check the number of ensemble members: {0}'.format(len(var_ens)))
if extreme == 'mean':
#____________Compute the time mean over the entire period, for each ensemble member
# MEAN
varextreme_ens = [np.mean(var_ens[i], axis=0) for i in range(numens)]
elif len(extreme.split("_")) == 2:
#____________Compute the chosen percentile over the period, for each ensemble member
# PERCENTILE
q = int(extreme.partition("th")[0])
varextreme_ens = [
np.percentile(var_ens[i], q, axis=0) for i in range(numens)
]
elif extreme == 'maximum':
#____________Compute the maximum value over the period, for each ensemble member
# MAXIMUM
varextreme_ens = [np.max(var_ens[i], axis=0) for i in range(numens)]
elif extreme == 'std':
#____________Compute the standard deviation over the period, for each ensemble member
# STANDARD DEVIATION
varextreme_ens = [np.std(var_ens[i], axis=0) for i in range(numens)]
elif extreme == 'trend':
#____________Compute the linear trend over the period, for each ensemble member
# TREND
# Reshape grid to 2D (time, lat*lon) -> Y
#Y=[var_ens[i].reshape(var_ens[0].shape[0],var_ens[0].shape[1]*var_ens[0].shape[2])for i in range(numens)]
#print('Reshaped (time, lat*lon) variable: ',Y[0].shape)
trendmap = np.empty((var_ens[0].shape[1], var_ens[0].shape[2]))
trendmap_ens = []
for i in range(numens):
for la in range(var_ens[0].shape[1]):
for lo in range(var_ens[0].shape[2]):
slope, intercept, r_value, p_value, std_err = stats.linregress(
range(var_ens[0].shape[0]), var_ens[i][:, la, lo])
trendmap[la, lo] = slope
trendmap_ens.append(trendmap)
varextreme_ens = trendmap_ens
print(len(varextreme_ens), varextreme_ens[0].shape)
varextreme_ens_np = np.array(varextreme_ens)
print(varextreme_ens_np.shape)
print('\n------------------------------------------------------------')
print('Anomalies are computed with respect to the {0}'.format(extreme))
print('------------------------------------------------------------\n')
ensemble_mean = np.mean(varextreme_ens_np, axis=0)
#____________Compute and save the anomalies with respect to the ensemble
ens_anomalies = varextreme_ens_np - np.mean(varextreme_ens_np, axis=0)
varsave = 'ens_anomalies'
ofile = os.path.join(OUTPUTdir,
'ens_anomalies_{0}.nc'.format(name_outputs))
#print(ofile)
print('Save the anomalies with respect to the ensemble:')
print('ens_anomalies shape: (numens x lat x lon)={0}'.format(
ens_anomalies.shape))
save_N_2Dfields(lat_area, lon_area, ens_anomalies, varsave, varunitsnew,
ofile)
#____________Compute and save the climatology
vartimemean_ens = [np.mean(var_ens[i], axis=0) for i in range(numens)]
ens_climatologies = np.array(vartimemean_ens)
varsave = 'ens_climatologies'
ofile = os.path.join(OUTPUTdir,
'ens_climatologies_{0}.nc'.format(name_outputs))
#print(ofile)
print('Save the climatology:')
save_N_2Dfields(lat_area, lon_area, ens_climatologies, varsave,
varunitsnew, ofile)
#____________Save the extreme
ens_extreme = varextreme_ens_np
varsave = 'ens_extreme'
ofile = os.path.join(OUTPUTdir, 'ens_extreme_{0}.nc'.format(name_outputs))
#print(ofile)
print('Save the extreme:')
save_N_2Dfields(lat_area, lon_area, ens_extreme, varsave, varunitsnew,
ofile)
return varextreme_ens_np, vartimemean_ens, ensemble_mean, dates