CDO

Hi!

you could make use of the expr operator describes in the Operator News Forum:

• intro
• update

On the other hand, you could create your trend as an extra data file and just substract it. It all depends on the model behind the trend ;-)

hth
ralf

Hi Ralf,
Thanks for your reply. This kind of work (non linear trend) is totally new to me. Is there an example available to use 'expr' operator especially to remove the trend. I could see some other example in the web link given by you.

Also, how can we create our own trend as an extra data file. Do we need to use 'trend' operator for this purpose.

Thanks.
SB

SBR Climate wrote:

Also, how can we create our own trend as an extra data file. Do we need to use 'trend' operator for this purpose.

No, you just create the trend data artificially, by any model you want and then use the sub operator to remove it from the original data. trend is a diagnostics for estimation linear regression parameters.

Thanks for your reply and explanation. Still not sure how to create the trend data artificially. I am having a NetCDF file and thought that it can be done using CDO. Is there any model available in CDO to create the trend data artificially?

But you suggested a non-linear trend - what did you have in mind?

Yes, I have annual mean data for 800 years and I have prepared an area averaged time series to see the trend. I could see that the trend in the time series is non-linear. There are long term ups and down in the time series and I realised that detrend (linear fit) may not be suitable. Some documents suggest 'exponential fit' to deal with such data for which I am not aware. Thanks for following up my query.

IMO you have to change the tool for modelling your data (or its trend). numpy/scipy or the GSL might be good. Esp in the python environment there should be plenty of tools for that.

hth
ralf

Thanks for suggesting the tool to change the data. Regards, SB.