Problem in extracting data from netcdf

I am trying to collocate two datasets. My netcdf file is 113x101x24 means I have data for 24 hours. There's another text file that contains data. In this text file time is present in two columns (HH MM). I have extracted the time from this text file in to one column vector so, for example 4 hours 25 minutes is now 425.

Below is time from netcdf file. It's like 0, 60, 120, 180, 240...

time     
           Size:       24x1
           Dimensions: time
           Datatype:   int32
           Attributes:
                       long_name      = 'time'
                       units          = 'minutes since 2017-06-05 00:30:00'
                       time_increment = 10000
                       begin_date     = 20170605
                       begin_time     = 3000
                       vmax           = 999999986991104
                       vmin           = -999999986991104
                       valid_range    = [-999999986991104  999999986991104]
                       origname       = 'time'
                       fullnamepath   = '/time'

My question is how can I extract the data from netcdf at 4 hrs ? As I am doing collocation so. e.g. If from txt file my time is 425 (4 hrs 25 minutes) how to extract the 4 hrs time from netcdf file ? Or second case, if it's 435 from text file, it should be 5 from netcdf?

Any help is highly appreciated

Answer 1

[Cris LaPierre]

You haven't shared your file or a link to the source, so here's my best guess.

Based on what I see in the time attributes, the first sheet of your data corresponds to 2017/06/05 00:30:00. The begin_time is llsted as 3000 and the time_increment as 10000. This must mean 00:30:00 and 1:00:00 (1 hour). There is no data, then, that corresponds to 4:00:00. You can find data for 3:30:00 and 4:30:00.

To extract a portion of the data, specify the startLoc and count (see this example).

vardata = ncread(source,varname,start,count)

So to extract the data for 3:30:00, I could do this (untested)

data330 = double(ncread(ncFileName,'varNm',[1 1 3],[inf,inf,1]))

and for 4:30:00, I could do this

data430 = double(ncread(ncFileName,'varNm',[1 1 4],[inf,inf,1]))

[Zhou Ci]

@Cris LaPierre I tried above line of code and it gave me this error:

Error using internal.matlab.imagesci.nc/read (line 613)
START has incorrect number of elements (3). The variable has 1 dimensions.

Error in ncread (line 66)
vardata = ncObj.read(varName, varargin{:});

So, I changed it to

data330 = double(ncread(fname_MERRA2,'time',[3],[1]))
data330 =

   120

My file is too large that's why I didn't attach it. Secondly, my question is about comparing the netcdf time with text file time. It's fine if there comes a difference of few minutes. Lastly I have large number of netcdf files whose time I want to compare with time in text file (extracted it and made a column vector). 500 means 5 hrs and 00 minutes. 535 means 5 hrs and 35 minutes.

I read netcdf time:

   24×1 int32 column vector

      0
     60
    120
    180
    240
    300
    360
    420
    480
    540
    600
    660
    720
    780
    840
    900
    960
   1020
   1080
   1140
   1200
   1260
   1320
   1380

[Cris LaPierre]

It all depends what variable you are trying to load with the code I shared. It was intended to be used for loading the 113x101x24 data, not 'time'. It works on a netcdf file I have access to, but may have to adjust it a little for it to work for your specific file.

You need to specify a start and count for each dimension. of your (see the links above). Your code is starting at the 3rd row of 'time' and reading in 1 number (the [3],[1] inputs). To extract all the times, I would do one of the following based on what I needed

% as datetimes
t = datetime(2017,06,05,0,30,0) + minutes(double(ncread(fname_MERRA2,'time')));
% as elapsed time
t = duration(0,30,0) + minutes(double(ncread(fname_MERRA2,'time')));

The challenge is converting your text file time to the same format. You can find other posts here about it. I would probably read the text file using readmatrix and then convert to a duration doing something like this.

tft = readmatrix('filename.txt');
h = floor(tft/100);
m = rem(tft,100);
t_dur = duration(h,m,0)

Once you have both times captured, you might be able to find the corresponding sheet index in your data for each text tile time using ismembertol with the following syntax

[LIA,LocB] = ismembertol(___)

Here, LocB is an array of the index location in B for each element in A that is a member of B. You just have the find the correct tolerance to use.

EDIT: ismembertol only accepts single and double datatypes. That means any duration or datetime datatype will need to be converted to one of those. I would use the minutes function to convert durations to equivalent minutes (a double). That would allow you to specify the tolerance using meaningful values.