Change dimensions of time-concatenated dataset

[Richard Dixon]

Hi there

I have a dataset that is lat, lon, wind, time. The time axis is very large because each member of an ensemble has been concatenated along the time axis.

So rather having a file of

X=180, Y=90, T=51000

Am looking to change the dimensions to

X=180, Y=90, T=100, Z=51

where Z is the ensemble member.

I've been looking for worked examples but unable to find something similar.

Thanks, Richard

[henry Butowsky]

Hi Richard,
Its abit fiddly with ncap2 but possible.
Try the following script

/*** new.nco ****/
defdim("Tnw", 100);
defdim("Z", 51);
Tnw[Tnw]=T(::51)
windnw[$T, $Z, $Y, $X]=0.0;
windnw=wind;
/*******/

Then run the above script with the following command.

ncap2 -C -v -S new .nco in.nc out.nc

Then use ncrename to rename the variables "Tnw" and "windnw". and the dimension "Tnw". back to their original names

...Henry

[Richard Dixon]

Hi Henry - thanks for the quick reply, will give this a try, have not used any nco scripting on ncap2 before.

Just to clarify - I presume the $X and $Y are for the names of the X and Y co-ordinates, respectively? (I've used X and Y above but they're actually Lon and Lat).

Richard

[henry Butowsky]

Hi Richard,
$X and $Y are the lon dimension and the lat dimension.
So I guess you'll have to replace:
$X with $Lon
$Y with $Lat
X with Lon
Y with Lat

...Henry

[Richard Dixon]

Thanks, I will give it a try later. And sorry for using the wrong forum - please move to the Help one, if you can...!

[Richard Dixon]

Hi there - afraid I'm none the wiser on the above as it's all pretty new to me - maybe I need to provide more detail to help the script as it's crashing with a

ew.nco line 1, column 20: unexpected character '

A bit more detail on the data as I was a bit vague above: the contents of my netcdf file given by print (g.variables.keys()) is:

odict_keys(['time', 'lon', 'lat', '10fg'])

In terms of the shapes of the variables:

10fg (wind data) is (10965, 181, 360) - so (Time, Latitude, Longitude)
Lat is (181)
Lon is (360)
Time is (10965) - but time is just a repeating loop of 24 to 5160 in repeated 51 times over.

I would there like the time array to be 215, and a new dimension being Ensemble (51) that just counts upwards from 1 to 51.

i.e. 10fg would be (215, 51, 181, 360)

Hope this is a bit clearer (but doesn't make it harder)...

Thanks, Richard

[henry Butowsky]

HI Richard,
An error on the first line ?
Thi s is the revised script - the wind variable "10fg" has to be quoted in ncap2 so that it can be distinguished from a plain number. This assumes that the lattiude coordinate variable is 'Lat' and the longitude coordinate variable is 'Lon'

defdim("time_nw", 215);
time_nw[$time_nw]=time(0:214);

defdim("Ensemble", 51);
Ensemble=array(1.0,1.0, $Ensemble);

wind_nw[$time_nw, $Ensemble, $Lat, $Lon]=0.0; 
wind_nw='10fg';

....Henry

[Richard Dixon]

Hi Henry

Many thanks for this - nearly there - one last thing is that I seem to lose the lat and lon variables in the process - the output file is:

odict_keys(['time_nw', 'Ensemble', 'wind_nw'])

Presumably I just need lines in there that are:

lat_nw[$lat_nw]=lat(0:180) and the like (I thought I'd check first up!)

Many thanks again
Richard

[henry Butowsky]

Hi Richard,
run the command line agin but omit the '-C' switch so all the used coordinate variables are pulled accross

ncap2 -v -S new .nco in.nc out.nc

Alternativly ammend the script as follows.

defdim("time_nw", 215);
time_nw[$time_nw]=time(0:214);

defdim("Ensemble", 51);
Ensemble=array(1.0,1.0, $Ensemble);

wind_nw[$time_nw, $Ensemble, $Lat, $Lon]=0.0; 
wind_nw='10fg';

lat=lat;
lon=lon;

...Henry

[Richard Dixon]

Hi Henry - just wanted to thank you for all the help on this - all working!