Re: using ind and num function instead of IF statement

Hi Debasish,

I am not confident that my algorithm is correct, so I have not attached it.

Here is how I would do it in 3 steps.

  1. Create a binary event -- that is, map your values to 1 when x > 10 and 0
     otherwise
                                       ___ ___ ___
  2. count number of delta functions __| |____| |____| |__
  3. select items in 2, which conform to your specification of duration

Step 1 is simple, and you can customize it for your need:
example invocation for your problem 1 :: xcat=categorize_var(x,"gt",10)
function categorize_var(var,sign,crit)
begin
  if sign.eq."gt"
    return(where(var.gt.crit,1,0))
  end if
  if sign.eq."lt"
    return(where(var.lt.crit,1,0))
  end if
end

Here is Dave Allured's code for item 2 and maybe 3 too..
I had to write one in Fortran to run the multi dimensional case, such as
yours. If you need the fortran version, let me know. However I hope you
can extend the NCL function below so that it works for the 2-dimensional
case.

function numBinOneRuns ( a[*]:integer, nCrit:integer)
; Dave Allured: CU/CIRES Climate Diagnostics Center (CDC)
; Return the number of runs [sequences] of 1s
local inds, ni, deltas, starts, ends, lengths
begin
   dim_nCrit = dimsizes(nCrit)
   if (dim_nCrit.gt.2) then
       print("numBinOneRuns: nCrit size must be .le. 2: nCrit="+dim_nCrit)
       return(-999)
   end if
   inds = ind (a .eq. 1)
   ni = dimsizes (inds)
   deltas = new (ni+1, integer)
   deltas = 0
   deltas(1:ni-1) = (inds(1:ni-1) - inds(0:ni-2))
   starts = inds(ind (deltas(0:ni-1).ne.1 .and. deltas(1:ni).eq.1))
   ends = inds(ind (deltas(0:ni-1).eq.1 .and. deltas(1:ni).ne.1))
   lengths = ends - starts + 1
   if (dim_nCrit.eq.1) then
       return(num (lengths.ge.nCrit) )
   else
       return(num (lengths.ge.nCrit(0) .and. lengths.le.nCrit(1)) )
   end if
end

cheers!
saji

--
* Debasish PaiMazumder <debasish@sca.uqam.ca> [2009-11-11 16:19:56 -0500]:
> Sorry for posting again... Last post has some numerical errors
> 
> Hello,
> 
> I need help regarding a problems mentioned below
> 
> I have a variable x with dimension month, latitude and longitude
> 
> x(month,lat, lon)     month=360, lat= 172, lon=180
> 
> I need to find out following things
> 
> 1) Y (lat,lon)= number of 3 to 9 month events having x exceeding 10
>   That means I need to calculate how many times x > 10 for consecutive
> 3 to 9 months at a particular gridcell
> 
> 2) suppose Y1(lat, lon)=number of at least 10 month and more events
> having x exceeding 20
> i.e. means I need to calculate how many time x > 20 for consecutive 10
> months and more at particular gridcell
> 
> How do I calculate Y and Y1 without using if loop in NCL
> 
> Thanks
> 
> Debasish
> 
> Debasish PaiMazumder, PhD
> Postdoctoral Researcher
> Analysis and Simulation of Regional-Scale Climate
> Department of Earth and Atmospheric Sciences
> University of Québec at Montréal (UQÀM)
> 550, Sherbrooke west, 19th floor, west tower,
> Montréal, QC, Canada H3A 1B9
> Tel: (514) 282-6464-343
>      (514) 937-0042 (R)
> email: debasish@sca.uqam.ca
> 
> 
> 
> 
> 
> _______________________________________________
> ncl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
-- 
Saji N. Hameed
APEC Climate Center          				
1463 U-dong, Haeundae-gu,                               +82 51 745 3951
BUSAN 612-020, KOREA                    		saji@apcc21.net
Fax: +82-51-745-3999
_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk