Hi all,
I am trying to generate using kriging a metamodel of a sedimentary soil. I am assuming that the soil properties are symmetric along the vertical axis. Not sure how can I do that. Any help more than welcome.
Best
Toni
If the objective is to spare computations on the other side of the data you could run the kriging on one side of the data then try to use the ExpertMixture to assemble a metamodel that combines the original metamodel and the symmetrized metamodel to cover the whole domain.
Or just do a kriging on the whole domain with a larger discretization.
Thanks @schueller for your prompt response.
I think that I was not clear in my intention. I am trying to represent a sedimentary soil. That soil is formed by the superposition of layers that sediment vertically. For that reason, a reasonable assumption would be that the scale parameters are the same in the horizontal plane and differ from the one in the vertical axis.
I attach a simple synthetic example in which from some boreholes we perform a geotechnical tests at different depths. The example tries to represent a soft upper layer, underlaid by an intermediate very stiff layer and finishing with a less stiff layer. I am using an ot.SquaredExponential as covariance model. I am getting these scale factors [18.033, 20.9073, 4]. As expected the scale factor along x and y are very similar. However, I would like to know if there is a way to force them to be the same.
Thanks a lot.
Toni
Ground_krigging_3D.py (4.7 KB)
we have the IsotropicCovarianceModel for that, but I I dont know if its possible to add the z component from there