Kriging

Someone is working on adding Kriging for GRASS, but I forgot
who they were. I just joined this list and am looking for updated
information on this project.

Darrell

Hi Darrell and others...

I am at least one "Someone" out there working on putting
Kriging into GRASS...there are probably others and perhaps
someone at the USGS has already done so....

Rather than respond directly via email to Darrell, I thought I'd put
this out to the list in the hope that there actually is another
"someone" out there or at least some interested potential users...

The Kriging subroutine will be finished by the end of March
(or my supervisor will have my "guts for garters"...). At that time
I will deliver it to the appropriate GRASS ftp site/dir for all
to discover that I am NOT a programmer! Excuse the fortran like C!

The program is using s.surf.idw (written by M. Shapiro I think) as
the template. It is a Universal Block Kriging routine, i.e. you can
include any knowledge of trends that you may have and it is primarily
for raster applications.

The basic guts of the routine is as follows:

Variogram models:
     1. spherical
     2. power
     3. exponential
     4. logarithmic
     5. gaussian
     6. hole-effect (haven't got this one working yet!)
     7. nested models (same state as 6.)

Trend removal:
    1. linear
    2. quatratic

Distance calculations are currently lat-long on a spherical
model of the earth, until I can get G_distance working that is.

Outputs 2 raster maps:
   1. output_z....kriged estimate of any value z
   2. output_varz....kriged estimate of the variance of value z

Input required:
   1. input site list
   2. output_z map
   3. output_var map
   4. interpolation pts.
   5. variogram range
   6. variogram sill
   7. variogram nugget
   8. trend info (format to-be-decided)

I will include all the software I've written for massaging the data
into variograms and determining the appropriate model, but this
is written in fortran and requires the gf graphics package, so users
without gf will need to supply their own graphics routines or use a
commercial package such as SURFER (from Golden Software).

I would appreciate feedback, advise (humour appreciated!),wishes,
wants...or any discussion regarding the general problems associated
with Kringing or geostatistics in general.

Most of my algorithms are from a book by Issaks and Srivastava,
"An Intro to Applied Geostatistics" and the matrix algebra stuff
is entirely from Numerical Recipes in C.

Cheers,
Chris

P.S. Is it true that we won't have access to any .mil sites
      from downunder anymore? So says the rumour mill down
      here....

Variogram models:
     1. spherical
     2. power
     3. exponential
     4. logarithmic
     5. gaussian
     6. hole-effect (haven't got this one working yet!)
     7. nested models (same state as 6.)

8. linear ?

Outputs 2 raster maps:
   1. output_z....kriged estimate of any value z
   2. output_varz....kriged estimate of the variance of value z

block kriging, huh? Would it be difficult to add punctual Kriging
and output site lists?

into variograms and determining the appropriate model, but this
is written in fortran and requires the gf graphics package, so users
without gf will need to supply their own graphics routines or use a
commercial package such as SURFER (from Golden Software).

Is this just for plotting points and variograms? If so, perhaps it
could be modified to use GNUPLOT or something that's available to
most everyone. At one time, I put together a *crude* hack that
used the GRASS display. This would probably be the way to go
(is anyone listening? how easily could this be done?)

I would appreciate feedback, advise (humour appreciated!),wishes,
wants...or any discussion regarding the general problems associated
with Kringing or geostatistics in general.

I started to do just what you are doing one summer, but gave up.
I'm am really looking forward to seeing what you put together.
If fact, I have several experiments planned that will utilize
your stuff. If I can be of any help, let me know.

Darrell