| Re: speher keyword of qhull [message #51267 is a reply to message #51174] |
Fri, 10 November 2006 09:55   |
Haje Korth
Messages: 651
Registered: May 1997
|
Senior Member |
|
|
Andy,
yes I hope one of these days, ITTVIS provide a detailed tutorial on
qhull/griddata, describing the right combination of keywords to get a
particular job done. This should benefit anyone involved in Earth
and/or space sciences, which is a pretty broad audience.
I got bit for the colinear-points problem with triangulate a few years
ago because I wanted to interpolate data on a sphere which came on a
regular grid. On regular grids the probability that three point lie
along a great circle is pretty large, thus resulting in the
colinear-point error message in triangulate.
Haje
Andy Heaps wrote:
> Hello Sebinjapan,
> an interesting question for me as I was puzzling over the
> same point last week. I don't have any real answers but the following
> should maybe provide some illumination on the subject:
>
> I have a strange grid of points in longitude and latitude on a globe:
> LATS FLOAT = Array[722, 511]
> LONGS FLOAT = Array[722, 511]
> TEMPERATURE FLOAT = Array[722,511]
> The longitudes (LONGS) go from -179.997 to 179.999 and the latitudes
> (LATS) go from -77.0105 to 89.8854. TEMPERATURE is the ocean surface
> temperature on this grid.
>
> The location of the longitudes and latitudes are very irregular so
> triangulation data is needed to pass to the CONTOUR command. The
> TRIANGULATE procedure works and CONTOUR produces a pretty decent plot.
> There is a wrapping problem, however, at the dateline seen as a
> line from the north to south poles. Adding the /SPHERE to the
> TRANGULATE call produces:
> % TRIANGULATE: Points are co-linear, no solution.
>
> No hope here then so I then used QHULL on the suggestion of a colleague:
> If I use QHULL, longs, lats, tri, /DELAUNAY, SPHERE =1
> I get:
> % QHULL: Expression must be named variable in this context: <INT
> ( 1)>.
>
> Okay, let's used a named variable:
> QHULL, longs, lats, tri, SPHERE = s
> HELP, s, tri
> S DOUBLE = Array[3, 368943]
> TRI LONG = Array[3, 733642]
>
> Using the triangulation data TRI with CONTOUR produces a perfect plot
> with no dateline missing data.
>
> So we're back to your original question - what is returned by
> SPHERE=s? THE IDL reference guide for version 6.0 QHULL says:
> DELAUNAY Performs a Delaunay triangulation and returns the vertex
> indicies of the resulting polyhedra; otherwise, the convex hull of the
> data are returned.
> SPHERE Computes the Delaunay triangulation of the points which lie on
> the surface of a sphere. The V0 argument contains the longitude, in
> degrees, and V1 contains the latitude, in degrees, of each point.
>
> It looks like the IDL documentation could be better just here. Does
> SPHERE=s imply a Delaunay triangulation and that s is the convex hull
> of the data? Printing off some values of s I get:
> PRINT, s(*, 0:3)
> PRINT, 'min, max of s are', MIN(s), MAX(s)
> 0.066654513 0.21466261 -0.97441118
> 0.064778714 0.21523610 -0.97441118
> 0.062897982 0.21579320 -0.97441118
> 0.061012460 0.21633386 -0.97441118
> min, max of s are -0.99999048 0.99999800
>
>
> Is this the sort of thing you're trying to work out as well?
>
>
> Cheers
> Andy
>
>
>
>
> David Fanning wrote:
>> sebinjapan writes:
>>
>>> Does anyone knows was is returned in the sphere variable when using:
>>> qhull, v0, v1, tr, /delaunay, sphere = s
>>
>> As I read the documentation, it seems to me SPHERE is
>> an input variable, with a value of 0 or 1. :-)
>>
>> Cheers,
>>
>> David
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
