Re: [ferret_users] Gridding Scattered Data:scat2grid functions

[Ansley Manke <ansley.b.manke@xxxxxxxx>]

Hi Steve,
Regarding the last argument to the scat2gridgauss functions, we've 
changed the requirements of that argument, first requiring it as a 
necessary argument, and then after consulting further with the author of 
the underlying code, we made it unused.  If you're using a version of 
Ferret prior to v5.81 you should set the last two arguments equal to 
each other, and for v5.81 and after you can set the last argument to 
anything.

This sounds like a problem, however, that might be better solved as an 
external function, using the code that you've written to put the data 
into bins and then average it. Either do this as a Ferret external 
function or run your Fortran code separately. This is going to be slow, 
going through that much data and deciding which grid cell each point is 
in. I don't think you need to go through the data multiple times though; 
you might organize it as follows:

Input variables for scattered points lon(npts), lat(npts), var(npts)

Define the output grid; here its 0.1 degrees, with variable SUM and 
COUNT which are initialized to 0.
If the data is in the range, say, 300 to 360 LON and 22 to 65 LAT, then 
your Fortran grid could be

SUM(3000:3600, 220:650)
COUNT(3000:3600, 220:650)

loop through scattered points n = 1:npts
   iloc = INT(lon(n)*10 )
   jloc = INT(lat(n)*10 )
   SUM(iloc,jloc) = SUM(iloc,jloc) + var(n)
   COUNT(iloc,jloc) = COUNT(iloc,jloc) + 1.
end loop

and then take the average SUM/COUNT for each i,j .

This puts the value at (302.14, 29.87) would go in the bin  (3021,298). 
You'll need to be careful about how the INT works and whether you want 
the lon,lat values rounded off or truncated, but this is the idea.

Ansley

guimond@coaps.fsu.edu wrote:

> Hello,
>  I have a large set of scattered satellite data that I want to put 
> onto a regular lat-lon grid.  My data looks like this:
>
> dbz  i=1:105000
> lat  i=1:105000
> lon  i=1:105000
>
> I defined two axes of my destination grid like so:
>
> def axis/x=122:134:0.10/units=deg glon
> def axis/y=27:33:0.10/units=deg glat
>
> and tried using the scat2gridgauss_xy function call like this:
>
> let a = scat2gridgauss_xy(lon,lat,dbz,x[gx=glon],y[gy=glat],0.5,0.5,2.,2)
>
> However, the graph that results does not look right.  Does anyone know 
> what I am doing wrong here?  I made a fortran routine to smooth my 
> satellite data to a 0.10deg grid and this looks right, but takes very 
> long to cycle through the file so many times.  Is there another FERRET 
> routine that does the sort of thing I am looking for?  I really just 
> want to average all satellite points that fall within a 0.10x0.10 deg 
> box to get a regular grid, but @AVE needs lat/lon grid on source 
> data.   One further note, the last argument to scat2gridgauss_** says 
> it can be zero since it is not used.  However, I get an error when I 
> don't put something greater than zero.
>
> Thanks for your help,
>
> Steve
>
>
> =======================================================
> Stephen R. Guimond
> Graduate Research Assistant
> Center for Ocean-Atmospheric Prediction Studies (COAPS)
> Tallahassee, FL 32304
> =======================================================
>
>
> ----------------------------------------------------------------