From: Arjen Markus on
On 7 jan, 16:09, Dieter Britz <dieterbr...(a)yahoo.com> wrote:
> Arjen Markus wrote:
> > On 7 jan, 13:52, Dieter Britz <dieterbr...(a)yahoo.com> wrote:
> >> Can someone point me at a free subroutine in Fortran for 2D interpolation
> >> that is more recent that the 1980's?
> >> --
> >> Dieter Britz (dieterbritz<at>yahoo.com)
>
> > netlib.org is the authoritive source of course, but why the need for
> > something more recent than the 1980's? Do you expect a new algorithm
> > has popped up?
>
> > Regards,
>
> > Arjen
>
> Point taken. What I should have asked is, what is recommended? I have
> tried out one, but it doesn't seem to work very well. I'll check to
> see what netlib has.
>
> Dieter
> --
> Dieter Britz (dieterbritz<at>yahoo.com)

Hm, that is very hard to answer in general. If your data points are
randomly spread, then a method like Renka-Cline's (triangulate and
interpolate linearly) could do the job nicely.

If your data points lie on irregular tracks, triangular meshes are
probably not very good, unless the mesh generation is aware of
that.

Simple methods are of course available too: data points weighted
by inverse distance or inverse distance squared. But they tend
to create somewhat unsmooth results (near the data points the
contour lines get distorted).

Or some Kriging method ...

What kind of data points do you have?

Regards,

Arjen
From: Ron Shepard on
In article <hi4tot$fol$1(a)news.net.uni-c.dk>,
Dieter Britz <dieterbritz(a)yahoo.com> wrote:

> Point taken. What I should have asked is, what is recommended? I have
> tried out one, but it doesn't seem to work very well. I'll check to
> see what netlib has.

If your data is on a 2D grid, then look for "bilinear interpolation"
routines. The simple versions work on regular grids with equally
spaced x and y values, but there are other versions that work on
irregularly space grid points too. If your data is not on a grid
but on randomly located (x_i,y_i) points, then this is a more
difficult problem.

$.02 -Ron Shepard
From: AnotherSquid on
On Jan 7, 8:49 am, Ron Shepard <ron-shep...(a)NOSPAM.comcast.net> wrote:
>
> If your data is on a 2D grid, then look for "bilinear interpolation"
> routines.

Here at NCAR (National Center for Atmospheric Research, Boulder) we
have occasion to do interpolation on 2D regular and Gaussian grids.
Lots of people here think that bilinear interpolation is fine. We also
use bi-cubic spline interpolation (two continuous derivatives in each
direction). This algorithm was programmed here by Alan Kaylor Cline in
the 1980s and is still, I believe, considered state-of-the-art. If you
want the program, let me know.

Andy