Hi Joaquim,
I still use GMT’s Classic Mode. That’s probably the legacy of over 20 years of GMT and code snippets for everything imaginable.
My first thought was also to use gmtselect and calculate the statistics from there. However, writing so many files put me off. With several hundred coordinates per time step and five raster data sets for the statistics, 100 time steps per scenario and 64 scenarios, that’s quite a lot of files that need to be written. Do you think the julia approach could be faster than grdfilter? The grdfilter commands are very easy to parallelize, but as far as I understand, your calculation does not require writing temporary files because the data remains in RAM?
Our IT department blocks julia codes on Windows computers, but I could try running julia - gmt on the Linux cluster if it pays off in terms of performance.
Thanks for your solution suggestion
Jörg