average (DJF)

以下のようにしてmonthly dataからDJF(12月～2月)平均データを求めることができる．

import numpy as np

import xarray as xr

ds = xr.open_dataset(fn) # fn is file name of netcdf data of geopotential height (zg).

# Read data at 500 hPa

z500=ds['zg'].sel(plev=50000, \

time=slice('1950-12-01','1952-12-31'), \

lon=slice(0.4, 0.6), lat=slice(-89.6, -89.4))

# Seasonal average (DJF

# Use time.season, which returns DJF, MAM, JJA, or SON,

# and mask (set as nan) non-DJF seasons

z500_djf1 = z500.where(z500['time.season'] == 'DJF')

# 3-month running average, which results in averaged data is located only

# time position of January.

z500_djf2 = z500_djf1.rolling(min_periods=3, center=True, time=3).mean()

# Take yearly averaged, which just extract January data (since other month data

# is nan, and convert time coordinate into year.

z500_djf3 = z500_djf2.groupby('time.year').mean('time')

# Remove first year data, which is nan. Note that if you take MAM average,

# first year dat is not nan and should be kept.

z500_djf4=z500_djf3[1:,:,:]

# In the above example, for the convenience of readers, different names

# are used for z500_djf1, ...4, but in real application it is better

# to use the same name for the efficient use of memory.