Python - Read data from netCDF file with time as “

问题:

I need to extract values from a netCDf file. I am pretty new to python and even newer this file format. I need to extract time series data at a specific location (lat, lon). I have found that there is a variable (called "base_time") in UNIX time and another variable (called "time") with "seconds since 2013-20-10 00:00:00" (which is the beginning of measurement time in UTC) for now.

When i interrogate the dataset's variables, I get this:

<type 'netCDF4.Variable'>
int32 base_time()
    units: seconds since 1970-01-01 00:00:00 00:00
unlimited dimensions:
current shape = ()
filling off

<type 'netCDF4.Variable'>
float64 time(time)
    units: seconds since 2013-10-20 00:00:00 00:00
    interval(sec): 30.0
unlimited dimensions: time
current shape = (2880,)
filling off

When I read the values as arrays, like e.g.

time_var = dataset.variables['time'][:]

I can see that there are 2880 (size is 2880) values in the time variable, but just one (size is 1) in the base_time variable. I think that this answer does exactly what I need, but I have only trouble with the part where I need to convert the time. When I do:

dtime = netCDF4.num2date(time_var[:],time_var.units)

I get the error:

AttributeError: 'numpy.ndarray' object has no attribute 'units'

And I think that I need to convert the time variable (seconds since beginning of measurement) anyway, instead of converting the UNIX time (because there is only one value in the netCDf file?). I tried some variations of the datetime.dateime part but I just don't get it. I only need to convert the "seconds since 2013-10-20 00:00:00" to a readable format to be able to extract und plot the data. Thanks!

回答1:

Sorry for being verbose but this is something I've just recently run across and grasped.

In the python NetCDF4 api, which is based on numpy, there is a profound difference between a NetCDF4.Variable and the numpy data array which it contains. Your code:

time_var = dataset.variables['time'][:]

is not a NetCDF4 variable, i.e. not a time_var but just the data values, a numpy ndarray of numbers, the NetCDF variable attributes are lost, in this case units:

units: seconds since 2013-10-20 00:00:00 00:00.

What you want is:

time_var = dataset.variables['time']

Then:

dtime = netCDF4.num2date(time_var[:],time_var.units)

Should work as expected.