Loop over netCDF datetime format and calculate mean based on month-CodePudding

I have a dataset (input_file) with the dimensions (504, 720, 500) where the first is a datetime value:

0     1979-01-15
1     1979-02-15
2     1979-03-15
3     1979-04-15
4     1979-05-15
         ...    
499   2020-08-15
500   2020-09-15
501   2020-10-15
502   2020-11-15
503   2020-12-15
Length: 504, dtype: datetime64[ns]

There is a variable with values I want to average per month. So ultimately I would like 12 values with the average of the variable based on the month in the first dimension.

I tried looping over it like such:

# empty dataframe
df = pd.DataFrame(columns = ['Month', 'Value'])

for i in range(size(df['time'])):
    month = input_file['time'][i].month # get the current month
    avg = np.average(input_file['values'][i, :, :]) # average for the month of that year

    # append to df
    df = df.append(pd.DataFrame({'Month' : month,
                                 'Value' : avg})

But up until here I am a bit lost, this doesn't work (invalid syntax) and I would still need to loop over the values again to get the average for each month seperately.

CodePudding user response：

Assuming the 2nd and 3rd dimensions are lat and lon, it seems what you are trying to do is just:

input_file.mean(dim = ['lat', 'lon'])

Then you can convert to a dataframe with .to_dataframe()