ds = ds.stack(realiz=['year', 'number'])
#compute seasonal means: ASO and SON
ninio3_aso = ds.sel(**{'month': slice(8, 10)}).mean(dim='month')
ninio3_son = ds.sel(**{'month': slice(9, 11)}).mean(dim='month')
ninio3_seasonal = xr.concat([ninio3_aso, ninio3_son], dim='season')
#select upper and lower quartile
ninio3_season_lower = ninio3_aso.quantile(0.25,
dim='realiz',
interpolation='linear')
ninio3_season_upper = ninio3_aso.quantile(0.75,
dim='realiz',
interpolation='linear')
ninio3_monthly_lower = ds.quantile(0.25, dim='realiz', interpolation='linear')
ninio3_monthly_upper = ds.quantile(0.75, dim='realiz', interpolation='linear')
[lamb, v, PC] = eofdata.eofdata(ds.sst.values, 4)
print(v[:, 0])
sst_monthly_index = -PC[0, :]
[lamb, v, PC] = eofdata.eofdata(ninio3_seasonal.sst.values, 8)
sst_seasonal_index = PC[0, :]
ds_new = xr.Dataset({
'ninio3_mon': xr.DataArray(sst_monthly_index),
'ninio3_seas': xr.DataArray(sst_seasonal_index)
})
ds_new.to_netcdf('~/datos/data/ninio3_index.nc')
ds.reset_index('realiz').to_netcdf('~/datos/data/ninio3_monthly.nc')
ninio3_seasonal.reset_index('realiz').to_netcdf(
'~/datos/data/ninio3_seasonal.nc')
RUTA = '~/datos/ERAI/'
hgt = []
for Y in np.arange(1981, 2019):
if Y != 2002:
ds = xr.open_dataset(RUTA + 'HGT50_erai_' + str(Y) + '.grib',
engine='cfgrib',
backend_kwargs={'indexpath': ''})
ds = ds.drop(['number', 'isobaricInhPa', 'step'])
hgt.append(ds)
hgt = xr.concat(hgt, dim='time')
hgt.time.values = np.arange(0, len(hgt.time.values))
hgt.to_netcdf('~/datos/data/hgt_erai_50.nc4')
hgt = hgt.sel(**{
'latitude': slice(-60, -90)
}).mean(dim=['longitude', 'latitude'])
hgt.to_netcdf('~/datos/data/PV_monthly_erai.nc4')
hgt.time.values = hgt.valid_time.values
hgt = hgt.sel(**{'time': slice('1981-01-01', '2017-12-31')})
PV_ason = hgt.sel(
time=np.logical_and(hgt['time.month'] >= 8, hgt['time.month'] <= 11))
PV = np.reshape(PV_ason.z.values, [36, 4])
PV = (PV - np.mean(PV, axis=0))
[lamb, v, PC] = eofdata.eofdata(PV.T, 3)
print(v[:, 0])
PV_monthly_index = -1 * PC[0, :]
ds_new = xr.Dataset({'PV_mon': xr.DataArray(PV_monthly_index)})
ds_new.to_netcdf('~/datos/data/PV_index_erai.nc4')
#compute monthly el ninio 3.4 index
import numpy as np
import xarray as xr
import eofdata
import os
os.environ['HDF5_USE_FILE_LOCKING'] = 'FALSE'
RUTA = '~/datos/data/'
ds = xr.open_dataset(RUTA + 'monthly_hgt50_aug_feb.nc')
ds = ds.sel(**{
'latitude': slice(-60, -90)
}).mean(dim=['longitude', 'latitude']).compute()
[lamb, v, PC] = eofdata.eofdata(ds.z.values, 3)
print(v[:, 0])
SPV_monthly_index = -1 * PC[0, :]
ds_new = xr.Dataset({'SPV_index': xr.DataArray(SPV_monthly_index)})
ds_new.to_netcdf(RUTA + 'fogt/SPV_index.nc')
#plt.savefig('./figures/scatter_PCs_SSTs.jpg')
#
##open era intreim data a compute el ninio index
#
sst_erai = xr.open_dataset('~/datos/data/fogt/sst_erai.nc4')
sst_erai.time.values = sst_erai.valid_time.values
#compute ninio 3.4 index
ninio34_erai = sst_erai.sel(
**{
'time': slice('1981-08-01', '2018-02-01'),
'latitude': slice(5, -5),
'longitude': slice(190, 240)
}).mean(dim=['longitude', 'latitude'])
ninio34_erai = ninio34_erai.sel(time=np.logical_or(
ninio34_erai.time.values <= np.datetime64('2002-07-31'),
ninio34_erai.time.values >= np.datetime64('2003-08-01')))
ninio34_erai = ninio34_erai.sel(time=np.logical_or(
ninio34_erai['time.month'] >= 8, ninio34_erai['time.month'] <= 2))
ninio34_erai = np.reshape(ninio34_erai.sst.values, [36, 7])
[lamb, v, PC] = eofdata.eofdata(np.transpose(ninio34_erai), 4)
print(v[:, 0])
sst_monthly_index = -PC[0, :]
ds = xr.Dataset({'ninio34_index': xr.DataArray(sst_monthly_index)})
ds.to_netcdf('~/datos/data/fogt/ninio34_erai_index.nc4')
#compute seasona means: ASO and SON
PV_aso = ds.sel(**{'month':slice(8,10)}).mean(dim='month')
PV_son = ds.sel(**{'month':slice(9,11)}).mean(dim='month')
PV_seasonal = xr.concat([PV_aso, PV_son], dim='season')
#select upper and lower quartile
PV_aso_lower = PV_aso.quantile(0.25, dim='realiz', interpolation='linear')
PV_aso_upper = PV_aso.quantile(0.75, dim='realiz', interpolation='linear')
PV_son_lower = PV_aso.quantile(0.25, dim='realiz', interpolation='linear')
PV_son_upper = PV_aso.quantile(0.75, dim='realiz', interpolation='linear')
PV_monthly_lower = ds.quantile(0.25, dim='realiz', interpolation='linear')
PV_monthly_upper = ds.quantile(0.75, dim='realiz', interpolation='linear')
ds.reset_index('realiz').to_netcdf('./data/PV_monthly2.nc')
[lamb, v, PC] = eofdata.eofdata(ds.z.values[0:4, :], 3)
print(v[:, 0])
PV_monthly_index = PC[0, :]
[lamb, v, PC] = eofdata.eofdata(PV_seasonal.z.values, 2)
print(v[0,:])
PV_seasonal_index = -1 * PC[0, :]
ds_new = xr.Dataset({'PV_mon': xr.DataArray(PV_monthly_index), 'PV_seas': xr.DataArray(PV_seasonal_index)})
ds_new.to_netcdf('./data/PV_index.nc')
PV_aso.reset_index('realiz').to_netcdf('./data/PV_aso.nc')
PV_son.reset_index('realiz').to_netcdf('./data/PV_son.nc')
PV_aso_lower.to_netcdf('./data/PV_aso_lower.nc')
#compute ninio indexes for erai and define ninio and ninia years
import numpy as np
import xarray as xr
import matplotlib.pyplot as plt
import pandas as pd
import eofdata
import matplotlib.pyplot as plt
import os
os.environ['HDF5_USE_FILE_LOCKING'] = 'FALSE'
RUTA = '~/datos/data/fogt/'
ds = xr.open_dataset(RUTA + 'sst_ninio34_aug_feb.nc4')
ninio34_monthly_lower = ds.quantile(0.25, dim='realiz', interpolation='linear')
ninio34_monthly_upper = ds.quantile(0.75, dim='realiz', interpolation='linear')
[lamb, v, PC] = eofdata.eofdata(ds.sst.values, 4)
print(v[:, 0])
ninio34_monthly_index = -PC[0, :]
ds_new = xr.Dataset({'ninio34_index': xr.DataArray(ninio34_monthly_index)})
ds_new.to_netcdf('~/datos/data/fogt/ninio34_monthly.nc4')
#plot scatter PC against SST. This is done to check how representative this index is
plt.figure(1)
plt.scatter(sst_monthly_index, sst_seasonal_index, s=80)
plt.xlabel('PC monthly')
plt.ylabel('PC seasonal')
plt.savefig('./figures/scatter_PCs.jpg')
plt.figure(2, (12, 16))
plt.subplot(421)
plt.scatter(sst_monthly_index, ds.sst.values[0, :], s=80)