Splitting the domain into zones with individual SRFs


We've talked about this, but I figured it should be documented.

So what has been proposed is that the user supplies information on how they want the domain to be split into distinct zones and the parameters for each zone's SRF. The RFG will create a random field for the domain by assigning each zone their own random field respective to their own SRF. This can either be done:
  • in MAD: the RFGD is passed a zone's coordinates, returns a RF, MAD assigns that RF to the correct locations in the domain, and then repeats for the remaining zones.
  • in the RFGD: the whole domain is passed to the RFG and the splitting of zones is handled by R.
The method is up to you, Carlos.

For now, the classifying of zones is up to the user. They will provide a text file with the following format:

col     row     layer     zone
1       1       1         1
2       1       1         2
1       2       1         2
2       2       1         1
MAD will need to figure out how many zones there are given the values in the 'zone' column and prompt the user for that many structural models. I am unsure about the mode of entry for these model parameters.
It would be nice if we could do it via text file(s). I'm open to suggestions. Maybe:

zone     model     mean     sill     range     nugget
1        exp       -2       *        10        0
2        exp       -8       *        10        0
with * indicating that this is a random variable and a prior is needed.


frystacka wrote Nov 14, 2014 at 12:46 AM

Can I get an update if this has been considered yet? Should I provide more information, possibly an example input file?

carosoisu wrote Nov 14, 2014 at 3:08 AM

I'm changing the structural parameters GUI to support the split function.. It was designed to support only one covariance model..