| INTERPOLATION TECHNIQUES HELP [message #7435] |
Tue, 12 November 1996 00:00  |
![Marcos Portabella Arn[1] is currently offline](theme/default/images/xoffline.png.pagespeed.ic.XRkd1fkXye.png "Marcos Portabella Arn[1] is currently offline")
Marcos Portabella Arn[1]
Messages: 2
Registered: November 1996
|
Junior Member |
|
|
Hi,
I need some help about the idl interpolation functions. I have tried all of
them and I could not get any satisfactory result. I supose this is due to my
lack of knowledge about them and this is the reason I am writing to this
newsgroup list. I am using IDL version 4.0.1 (vms alpha). My problem is the
following:
I have three data vectors. The first two are X and Y position coordinates
(longitude and latitude) and the third one is the magnitude mesured at each
point (in my case, wind measurements). These points have not any order
(irregularly gridded points) . In order to make an interpolation of these
points in a regurlar grid (with a x and y spacing of half degree, for example)
I have tried all three idl functions that allow this type of interpolation. The
first one, the TRIGRID function, is very fast but the results are poor
(moreover, at the grid border, where it oftenly needs to extrapolate, the
results are incredibily wrong, even if I set the EXTRAPOLATE keyword). The
second one, the KRIG2D function, gives much better results but the major
problem is the execution time: the time increases exponentially with the number
of points. I do not know if changing the function parameters (A, CO, C1) may
reduce the execution time, but as far as I have tested with the idl help
examples I could no get any better result. The third and last one, the
MIN_CURVE_SURF function, has the same time restriction as the KRIG2D function.
I would like to know if there is any other interpolating function for
irregularly gridded points that gives better time results; or, if using more
addequatly the KRIG2D or MIN_CURVE_SURF functions I can get better time
results as well.
Thank you very much,
Marcos Portabella
|
|
|
|
| Re: INTERPOLATION TECHNIQUES HELP [message #7461 is a reply to message #7435] |
Tue, 19 November 1996 00:00  |
dan
Messages: 27
Registered: March 1993
|
Junior Member |
|
|
In article <56acrg$vuk@news.esrin.esa.it>, Marcos Portabella Arnus - RS/EM <marcos> writes:
|> Hi,
|>
|> I need some help about the idl interpolation functions. I have tried all of
|> them and I could not get any satisfactory result. I supose this is due to my
|> lack of knowledge about them and this is the reason I am writing to this
|> newsgroup list. I am using IDL version 4.0.1 (vms alpha). My problem is the
|> following:
|>
|> I have three data vectors. The first two are X and Y position coordinates
|> (longitude and latitude) and the third one is the magnitude mesured at each
|> point (in my case, wind measurements). These points have not any order
|> (irregularly gridded points) . In order to make an interpolation of these
|> points in a regurlar grid (with a x and y spacing of half degree, for example)
|> I have tried all three idl functions that allow this type of interpolation. The
|> first one, the TRIGRID function, is very fast but the results are poor
|> (moreover, at the grid border, where it oftenly needs to extrapolate, the
|> results are incredibily wrong, even if I set the EXTRAPOLATE keyword). The
|> second one, the KRIG2D function, gives much better results but the major
|> problem is the execution time: the time increases exponentially with the number
|> of points. I do not know if changing the function parameters (A, CO, C1) may
|> reduce the execution time, but as far as I have tested with the idl help
|> examples I could no get any better result. The third and last one, the
|> MIN_CURVE_SURF function, has the same time restriction as the KRIG2D function.
|>
|> I would like to know if there is any other interpolating function for
|> irregularly gridded points that gives better time results; or, if using more
|> addequatly the KRIG2D or MIN_CURVE_SURF functions I can get better time
|> results as well.
|>
|> Thank you very much,
|>
|> Marcos Portabella
|>
Here is a routine I wrote which maps scattered data (longitude,latitude,value)
onto a regular lat,lon grid on a sphere.
; $ID$
;+
; Name:
; INTERP_SPHERE
;
; PURPOSE:
; This function maps scattered data defined by
; (longitude,latitude,value) onto a regular, but not
; neccessarily evenly spaced, grid whose coordinates are
; also defined by longitude and latitude. The procedure searches
; for the N (default = 5) closest data points to each grid
; point and then averages these N data points weighted by
; distance^power from the grid point to the particular data point.
; Default is power=-1 which weights the points inversely by
; distance. All distances are along great circles on a sphere
; (the shortest distance between two points along the
; surface of a sphere).
;
; CATEGORY:
; Interpolation?
;
; CALLING SEQUENCE:
; grid = INTERP_SPHERE(lat,lon,data)
;
; INPUTS:
;
; lat: The latitudes on the grid where interpolated
; values are desired (in degrees)
;
; lon: The longitudes on the grid where interpolated
; values are desired (in degrees)
;
; data: An array (3,ndata) where ndata is the number of
; data points, and can be any number larger than N.
; each row of data should contain a longitude, a
; latitude, and a value to be interpolated.
;
; KEYWORD PARAMETERS:
;
; N: The number of closest data points to be used
; for each grid point interpolation. Default = 5
;
; power: The exponent for the distance weighting function.
; Default = -1 (weighting inversely by distance).
; An input of power=-.5 would weight inversely by the
; square root of the distance.
;
; latwt: The weighting for the interpolation in the meridional
; (North-South) direction. For negative power,
; latwt > 1 produces a weighting with less latitude
; influence. Default = 1
;
; mask: Mask for calculating grid values
;
;
; OUTPUTS:
;
; grid: An array of interpolated data values. It has dimensions
; (nlon,nlat) where nlon is the number of entries in the
; input lon, and nlat is the number of entries in the input
; lat.
;
; EXAMPLE:
;
; MODIFICATION HISTORY:
;
; written by: Dan Bergmann dbergmann@llnl.gov 11/10/94
;-
FUNCTION INTERP_SPHERE,lat,lon,data,n=n,power=power,latwt=latwt
nlat = (size(lat))(1)
nlon = (size(lon))(1)
grid = fltarr(nlon,nlat)
if (not(keyword_set(n))) then n = 5
if (not(keyword_set(power))) then power = -1
if (not(keyword_set(latwt))) then latwt = 1
if (not(keyword_set(mask))) then begin
mask = intarr(nlon,nlat)
mask(*,*) = 1
endif
dtr = !pi / 180.
; convert lat and lon to radians
latr = dtr * lat
lonr = dtr * lon
; convert the lat and lon of the data to radians
dlatr = dtr * data(1,*)
dlonr = dtr * data(0,*)
; calculate the cartesian coordinates of the data points
; assuming a unit sphere.
xdata = cos(dlatr) * sin(dlonr)
ydata = cos(dlatr) * cos(dlonr)
zdata = sin(dlatr)
for x=0,nlon-1 do begin
sinlonr = sin(lonr(x))
coslonr = cos(lonr(x))
for y=0,nlat-1 do begin
; check to see if this grid should be calculated
if (mask(x,y) ne 0) then begin
; calculate the cartesian coordinates of this particular
; grid point.
xorig = cos(latr(y)) * sinlonr
yorig = cos(latr(y)) * coslonr
zorig = sin(latr(y))
; calculate the length squared of the cords connecting this grid
; point to all the data points and then sort the data points by
; these values.
corddistsq = (xorig-xdata)^2+(yorig-ydata)^2+((zorig-zdata)*latwt)^2
sortdist = (sort(corddistsq))(0:n-1)
; if a data point lies directly on top of this grid point, then
; assign that value to the grid point.
; Otherwise calculate the n great circle distances and do a weighted
; average of the data values.
if ((corddistsq(sortdist))(0) eq 0) then begin
grid(x,y) = data(2,(sortdist)(0))
endif else begin
grcirdis = asin(sqrt(corddistsq(sortdist))/2.)
grid(x,y) = (total(data(2,sortdist) * grcirdis^power)) / total(grcirdis^power)
endelse
endif
endfor
endfor
return,grid
end
--
************************************************************ ***
** Dan Bergmann dbergmann@llnl.gov **
** Atmospheric Science Division fax (510) 422-5844 **
** Lawrence Livermore National Lab human (510) 423-6765 **
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
