Re: [ferret_users] Calculating Probability Density Function

[Peter Szabo <szabpet83@xxxxxxxxx>]

Hello Archana,
First of all, i do not know what the ".des" tag is, but  I tried your calculations using my daily data set with 10800 timesteps. Though you have (29 years*12=)348 months.
I searched the archives and found that using SAMPLE* and SORT* functions you can produce PDF (* can be I, J, K, L). The only problem is that they can only work for points, but with a trick it is solvable). So with this script i saved out my variable to a new netcdf file.

use "/home/szabop/precip/HC1.nc" !for me the dims are i=1:14, j=1:32, l=1:10800
repeat/j=1:14:1 (repeat/i=1:32:1 (let tsorted_indices_20 = sortl(precip20[i=`i`,j=`j`]); let tsorted_precip20 = samplel(precip20[i=`i`,j=`j`], tsorted_indices_20);\
save/file="/home/szabop/ki/PDF.nc"/ILIMTS=1:32/JLIMITS=1:14/i=`i`/j=`j`/append/clobber tsorted_precip20))

After using this new PDF.nc you should mark out one point (with this command: plot tsorted_precip20[i=1,j=1]; I attach a GIF how it looks like for the i=1 and j=1 point without normalization) 
or as you said you can produce spatially-averaged field (easier for a non-irregular or even easier for the whole territory) and plot that standing-alone "point".

If you want to work with seasonal fields, you will only have 29 timesteps.

I do not know whether it is what you wanted or not.
Regards,

Peter Szabo

Attachment: PDF0101.gif
Description: GIF image