def __init__(self,
index_name,
nc,
times_SST,
calendar_SST,
t_units_SST,
N,
season='JJA'):
""" """
index_ts = nc.variables[index_name][:]
index_annual_mean, years = calculate_annual_mean(index_ts,
times_SST,
calendar_SST,
t_units_SST,
season=season)
index_running_mean = running_mean(index_annual_mean, N)
self.index_running_mean = index_running_mean
self.years = years
# select only years 1900 onwards for ERA20C for common period with THF
if model_name == 'ERA20C':
SST_am = SST_am[years_SST >= 1900]
THF_am = THF_am[years_THF >= 1900]
years_SST = years_SST[years_SST >= 1900]
for j, region in enumerate(regions_all):
THF_region = region.get_regional_index(THF_am, lats_var, lons_var)
SST_region = region.get_regional_index(SST_am, lats, lons)
SST_region = region.get_tendency(SST_region, years_SST)
if apply_running_mean is not None:
THF_region = running_mean(THF_region, apply_running_mean)
SST_region = running_mean(SST_region, apply_running_mean)
# detrend
THF_region = detrend(THF_region)
SST_region = detrend(SST_region)
if plot_type == 'scatter':
slope, intercept, r_value, p_value, std_err = linregress(
SST_region, THF_region)
if model_name != 'ERA20C': plt.scatter(j, r_value)
else:
lo, hi = confidence_intervals(THF_region, SST_region)
plt.errorbar(j + 0.1,
r_value,
yerr=hi - r_value,
except:
# otherwise, read in SST, THF and zonal wind files to calculate regression
if i>=0:
# read in SST index data
SST_index_file_name = loading_dir + '/SST_indices_'+model_name+'_'+ML.ensemble_id[model_name]+'.nc'
nc_SST = Dataset(SST_index_file_name,'r')
times_SST = nc_SST.variables['times'][:]
calendar_SST = nc_SST.variables['times'].calendar
t_units_SST = nc_SST.variables['times'].units
NINO34, years = prep_SST_index('NINO34',nc_SST,times_SST,calendar_SST,t_units_SST,season)
IOBM2, years = prep_SST_index('IOBM2',nc_SST,times_SST,calendar_SST,t_units_SST,season)
if apply_running_mean is not None:
NINO34 = running_mean(NINO34,apply_running_mean)
IOBM2 = running_mean(IOBM2,apply_running_mean)
# detrend SST
NINO34 = detrend(NINO34)
IOBM2 = detrend(IOBM2)
# read in THF
if model_name != 'ERA20C':
data_path0 = '/network/aopp/hera/mad/patterson/CMIP6/data/piControl/'
full_path = data_path0 + 'hfls/'+ ML.model_institute[model_name] + '/' + model_name + '/hfls/'
list_of_files = [full_path+'hfls_Amon_'+model_name+'_piControl_'+ML.ensemble_id[model_name]+'_'+season+'.nc']
hfls,lats_var,lons_var,levs_var,times_var,calendar_var,t_units_var = read_in_variable(list_of_files[:],'hfls')
full_path = data_path0 + 'hfss/' + ML.model_institute[model_name] + '/' + model_name + '/hfss/'
list_of_files = [full_path+'hfss_Amon_'+model_name+'_piControl_'+ML.ensemble_id[model_name]+'_'+season+'.nc']
hfss,lats_THF,lons_THF,levs_THF,times_THF,calendar_THF,t_units_THF = read_in_variable(list_of_files[:],'hfss')
full_path = data_path0 + 'hfss/' + ML.model_institute[
model_name] + '/' + model_name + '/hfss/'
list_of_files = [
full_path + 'hfss_Amon_' + model_name + '_piControl_' +
ML.ensemble_id[model_name] + '_' + season + '.nc'
]
hfss, lats_THF, lons_THF, levs_THF, times_THF, calendar_THF, t_units_THF = read_in_variable(
list_of_files[:], 'hfss')
THF_am = hfls + hfss
#THF_am, years_THF = calculate_annual_mean(THF,times_THF,calendar_THF,t_units_THF,season=season)
if season == 'DJF':
THF_am = THF_am[::-1][1:][::-1]
if apply_running_mean is not None:
SST_am = running_mean(SST_am, apply_running_mean)
THF_am = running_mean(THF_am, apply_running_mean)
# calculate correlation map
corr_map, bottom_lat_bounds, left_lon_bounds = regress_gridpoint(
SST_am, THF_am, lats_SST, lons_SST, lats_THF, lons_THF)
# plotting
ax = plt.subplot(gs[i],
projection=ccrs.Robinson(central_longitude=180.))
plt.title(model_name, fontsize=20)
cs = plt.pcolormesh(left_lon_bounds,
bottom_lat_bounds,
corr_map,
cmap='RdBu_r',
vmin=-0.9,
nc = Dataset(regression_file_name, 'r')
pcs = nc.variables['pcs'][:]
variance_fraction = nc.variables['variance_fraction'][:]
print('Read in %s' % (regression_file_name))
# read in SST data
SST_index_file_name = loading_dir + '/SST_indices_' + model_name + '_' + ML.ensemble_id[
model_name] + '.nc'
nc_SST = Dataset(SST_index_file_name, 'r')
times_SST = nc_SST.variables['times'][:]
calendar_SST = nc_SST.variables['times'].calendar
t_units_SST = nc_SST.variables['times'].units
PDO, years = prep_SST_index('NINO34', nc_SST, times_SST, calendar_SST,
t_units_SST, season)
if apply_running_mean is not None:
PDO = running_mean(PDO, apply_running_mean)
# make sure that all indices have the same sign and add EOF maps to array
if model_name in ['CNRM-ESM2-1', 'CNRM-CM6-1', 'INM-CM5-0']:
pc_number = 1 # these models have the first 2 EOFs the other way round
else:
pc_number = 0
U = nc.variables['regress_U'][pc_number, :, :]
U_pc, sign_changed = change_index_sign(pcs[:, pc_number],
U,
lats_ua,
lons_ua,
bounds=[25, 35, 100, 160],
text='U')
if sign_changed == True: U = -U
SST = nc_SST.variables['reg_coeff_NINO34'][:]
# calculate annual means of fields
psl_am, years = calculate_annual_mean(psl_data,
times_psl,
calendar_psl,
t_units_psl,
season=season)
pr_am, years = calculate_annual_mean(pr_data,
times_pr,
calendar_pr,
t_units_pr,
season=season)
# perform low pass time filtering on fields
N = 10
halfN = int(N / 2)
psl_rm = running_mean(psl_am, N)
pr_rm = running_mean(pr_am, N)
# prepare SST indices
PDO_object = prep_index_regress_save('PDO',
nc_SST,
times_SST,
calendar_SST,
t_units_SST,
N,
season=season)
IOD_object = prep_index_regress_save('IOD',
nc_SST,
times_SST,
calendar_SST,
t_units_SST,
# read in SST data
SST_index_file_name = loading_dir + '/SST_indices_' + model_name + '_' + ensemble_id[
model_name] + '.nc'
nc_SST = Dataset(SST_index_file_name, 'r')
times_SST = nc_SST.variables['times'][:]
calendar_SST = nc_SST.variables['times'].calendar
t_units_SST = nc_SST.variables['times'].units
IOBM2, years = prep_SST_index('IOBM2', nc_SST, times_SST,
calendar_SST, t_units_SST, season)
NINO34, years = prep_SST_index('NINO34', nc_SST, times_SST,
calendar_SST, t_units_SST, season)
# apply running mean
if apply_running_mean is not None:
IOBM2 = running_mean(IOBM2, apply_running_mean)
NINO34 = running_mean(NINO34, apply_running_mean)
# regress out other index
IOBM2_no_NINO34 = regress_out_index(IOBM2, NINO34)
NINO34_no_IOBM2 = regress_out_index(NINO34, IOBM2)
# read in circulation indices
data_path0 = '/network/aopp/hera/mad/patterson/CMIP6/data/piControl/ua/'
data_path1 = '/' + var_name + '/'
full_path = data_path0 + model_institute[
model_name] + '/' + model_name + data_path1
list_of_files = [
full_path + var_name + '_Amon_' + model_name + '_piControl_' +
ensemble_id[model_name] + '_' + season + '.nc'
]
variable2_data = variable2_data[time_mask2]
# calculate annual means
variable1_JJA, years = calculate_annual_mean(variable1_data,
times1,
calendar1,
t_units1,
season='JJA')
variable2_JJA, years = calculate_annual_mean(variable2_data,
times2,
calendar2,
t_units2,
season='JJA')
# perform low pass time filtering
variable1_rm = running_mean(variable1_JJA, N)
variable2_rm = running_mean(variable2_JJA, N)
# only use variable in region over East Asia for MCA
lon_mask_EAsia = (lons1 >= lon_min) & (lons1 <= lon_max)
lat_mask_EAsia = (lats1 >= lat_min) & (lats1 <= lat_max)
variable1_rm_EAsia = variable1_rm[:, lat_mask_EAsia, :][:, :,
lon_mask_EAsia]
# only take variable 2 in tropics
lon_mask_tropics = (lons2 >= lon_min_tropics) & (lons2 <=
lon_max_tropics)
lat_mask_tropics = (lats2 >= lat_min_tropics) & (lats2 <=
lat_max_tropics)
variable2_rm_tropics = variable2_rm[:,
lat_mask_tropics, :][:, :,
clevs_mean,
title='ERAI EOF2 %.f%%' % (100 * var_frac[1]))
ax = plt.subplot(gs[1, 2], projection=ccrs.PlateCarree(central_longitude=180.))
cs = plot_mode(ax,
-reg_pc3,
np.mean(psl_am_ERAI, axis=0),
lats_ERAI,
lons_ERAI,
clevs,
clevs_mean,
title='ERAI EOF3 %.f%%' % (100 * var_frac[2]))
# do the same with a running mean applied
print('Calculating running means')
N = 10
psl_am_HadGEM_rm = running_mean(psl_am_HadGEM, N)
psl_am_ERAI_rm = running_mean(psl_am_ERAI, N)
clevs = 0.2 * clevs
ax = plt.subplot(gs[2, :])
make_plots.colorbar(ax, cs, orientation='horizontal')
var_frac, reg_pc1, pval_pc1, reg_pc2, pval_pc2, reg_pc3, pval_pc3 = calculate_EAsia_rm_eofs(
psl_am_HadGEM_rm, lats_HadGEM, lons_HadGEM)
ax = plt.subplot(gs[3, 0], projection=ccrs.PlateCarree(central_longitude=180.))
cs = plot_mode(ax,
reg_pc1,
np.mean(psl_am_HadGEM, axis=0),
lats_HadGEM,
lons_HadGEM,
clevs,
