How to split data information (YYYMMDD) ino YYYY MM DD from a NetCDF file in python?

Question

I am quite new to python and I read some previous related questions, but those does not answer exactly my question:

I have an already sliced .nc file for precipitation for my case study. In python I do the processing, and I need the time in 3 different columns [2015 01 31], though is given me to in one column YYYYMMDD [20150101 20150131 .... 20171221]

I have tried using the module netCDF4 and numpy to process and split the data as I required it.

#Script for reading .nc files using the netcdf4 module in Python

import numpy as np
import netCDF4 as nc


os. chdir('/mnt/lustre02/work/ch0636/g260150/sims/validation/selectedmodelRCPs/RCP2.6/Model1/')

#extract time in a single array to append it to the whole dly. weather file

data=nc.Dataset('precipitation.nc','r')
time=data.variables['time'][:]
b=np.zeros((time.size,1))
b[:,0]=time

#Extracting time sections: year, month, day

year=[]
month=[]
day=[]

for i in range(len(b)):
    year.append(b[i][0:4])
    month.append(b[i][4:6])
    day.append(b[i][6:])

print(day)  
print(month)
print(year)

Running this part gives that day is the whole date including year, month and day. Month and day are empty arrays. I am not really sure how I could actually approach this in a proper way, that helps me getting what I need for my further processing.

Answer 1

To get the arrays of year, month and day using your solution, I suggest modification like this:

year=[]
month=[]
day=[]

for i in range(len(b)):
    yval=np.floor(b[i]/10000.)
    mval=np.floor((b[i]-yval*10000.)/100.)
    dval=np.floor(b[i]-yval*10000.-mval*100.)    
    year.append(int(yval))
    month.append(int(mval))
    day.append(int(dval))

print(day)  
print(month)
print(year)

Nevertheless, I can also suggest a code like this, where we first convert the time to date values and then extract the years, months and days:

# other way:
import datetime
def get_date(datevalin):
    fraction_of_day=datevalin-np.floor(datevalin)
    d0=datetime.datetime.strptime(str(int(np.floor(datevalin))),'%Y%m%d')
    dval=d0+datetime.timedelta(seconds=fraction_of_day*24.*3600)
    return dval

dates=[get_date(val) for val in time];
year=[val.year for val in dates];
month=[val.month for val in dates];
day=[val.day for val in dates];