Rich Shepard wrote:
material. There is no interpolation algorithm in GRASS currently which
can
handle that sort of data well.
So what is needed is a political algorithm. 
That's actually right: given the presence of n different
layer types in the vicinity of an empty voxel, the algorithm
would need to decide by some sort of "majority vote"
which type to assign to that voxel.
Kidding aside, I suspect that a fuzzy interpolation algorithm would solve
the problem.
How? You could make the interpolated value depend on a
fuzzy set member function, I suppose, but the situation
here is actually so well defined that I think a probabilistic
approach would be preferable. Since each voxel can only
store one value, a second output map could store the
classification probability. That may be very useful
for visualization (you could show voxels with little
probability hazier).
Ben
Rich
_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user
--
Benjamin Ducke
Geospatial Consultant
Oxford Archaeology Digital
Janus House
Osney Mead
OX2 0ES
Oxford, U.K.
Tel: +44 (0)1865 263 800 (switchboard)
Tel: +44 (0)1865 980 758 (direct)
Fax :+44 (0)1865 793 496
benjamin.ducke@oadigital.net
http://oadigital.net
------
Files attached to this email may be in ISO 26300 format (OASIS Open Document Format). If you have difficulty opening them, please visit http://iso26300.info for more information.
On Fri, 8 Jan 2010, Benjamin Ducke wrote:
That's actually right: given the presence of n different layer types in
the vicinity of an empty voxel, the algorithm would need to decide by some
sort of "majority vote" which type to assign to that voxel.
Ben,
I'm not a geologist so I don't know how best to extrapolate from well
logs. I've used water well logs to determine the presence of impervious
layers and aquifer surfaces, but not tried to imcorporate those data into a
GRASS analysis. But your comment above makes me wonder if calculating
Veronoi diagram of these various layers would provide that majority vote.
Rich
It seems to me that this is a 3D interpolation problem with categorical variables.
Maybe the Bayesian Maximum Entropy approach could help. There are some interesting publications around also for geology and soil sciences, and they can deal with categorical data as well. Look for example here: http://www.enge.ucl.ac.be/staff/curr/Bogaert/biblioBME/BMEbibsubject.html#Soil%20Science
Or maybe you can have a look at SGeMS (http://sgems.sourceforge.net), a tool for 3D geostatistics.
None of them is available through GRASS, but the algorithms are freely available (I think open-source, but not verified).
I am not a geologist, so please forgive if it is not adequate...
Christian Kaiser
On 8 janv. 2010, at 11:04, Benjamin Ducke wrote:
Rich Shepard wrote:
material. There is no interpolation algorithm in GRASS currently which
can
handle that sort of data well.
So what is needed is a political algorithm.
That's actually right: given the presence of n different
layer types in the vicinity of an empty voxel, the algorithm
would need to decide by some sort of "majority vote"
which type to assign to that voxel.
Kidding aside, I suspect that a fuzzy interpolation algorithm would solve
the problem.
How? You could make the interpolated value depend on a
fuzzy set member function, I suppose, but the situation
here is actually so well defined that I think a probabilistic
approach would be preferable. Since each voxel can only
store one value, a second output map could store the
classification probability. That may be very useful
for visualization (you could show voxels with little
probability hazier).
Ben
Rich
_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user
--
Benjamin Ducke
Geospatial Consultant
Oxford Archaeology Digital
Janus House
Osney Mead
OX2 0ES
Oxford, U.K.
Tel: +44 (0)1865 263 800 (switchboard)
Tel: +44 (0)1865 980 758 (direct)
Fax :+44 (0)1865 793 496
benjamin.ducke@oadigital.net
http://oadigital.net
------
Files attached to this email may be in ISO 26300 format (OASIS Open Document Format). If you have difficulty opening them, please visit http://iso26300.info for more information.
_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user