CDO

Dear Sri,

I am not quite sure, what you mean by 'interannual variability'. There come to ways to my mind:

1. You can just calculate the yearsum of precipitation and take its variance:

cdo -timvar -yearsum precipitation_file output_file

cdo -timstd -yearsum precipitation_file output_file

2. you could use the time series filter to compute interannual variability. Using a lowpass with fmax=1, you would suppress the variability on timescales shorter than a year, and leave the interannual fluctuations. In the end, the variance of this time series could be taken as a measure for the interannual variability.

cdo -lowpass,1 precipitation_file output_file

Please note, that if you use the filter, it has for finite number of time-steps not a strictly rectangular response which means, that you could use some variability although the frequency fmax=1 would be included. Consequently I would suggest, to use a value fmax slightly larger than one.

Cedrick

Hello Cedrik,

Many sincere thanks for the reply and commands. I tried this, but still cannot seem to get it work.

Are all the 3 commands above doing the same thing or one follows the other ?? a bit confused here...

Thanks,

Sri

Hi Cedrik and Sri,

I have come accross the same problem as Sri. I have monthly data but I want it in a yearly timescale (i.e to have one value of precip for each year). So then, by plotting that, I want to be able to view the interannual variability of rainfall in my region.

How can I do this?

Cedrik, how does CDO know when you set fmax =1 that 1 stands for 1 year and not 1 month?

Thank you,

Duarte

Hi,

if you want the yearly precipitation sum, cdo yearsum is what you're looking for. Similarly cdo yearstd and cdo yearvar give you the year-to-year standard deviation respectively variance.
The commands cdo ymonstd or ymonvar would give you the year-to-year variability for each month seperately.

If you are having problems could you please provide examples? Some files and a statement on what kind of command you used would be very helpful in order to help you.

Frequencies for band-pass filters are assumed to be given in years. If you want the frequency of a month, this would be 12. The time axis info from the input file is used to determine the reference timestep, which is always the first one in the datafile. This is also documented in the online help which you can call by cdo -h highpass. This also includes some hints how to effectively use the filters.

Cheers,
Cedrick