[Sgems-users] strange result on kriging and cokriging variances

[M. N. H. <mn_...@ya...>]

Dear SGeMS Users,
 
I would like to share the strange result on kriging and cokriging variances in SGeMS. I am using SGeMS v2.1 and sometimes compared it to version v2.0 and GSLIB.
When I did kriging and have a look into the kriging variance, SGeMS v2.1 showed much larger values compared to GSLIB i.e. the mean is up to 4 times larger.
If we want to use the kriging variance to measure the accuracy on estimation result, this will be a serious problem, eventhough their kriged values are similar.
 
Also the cokriging variance obtained by SGeMS v2.1 is larger than GSLIB, and their cokriged values is bit different i.e. statistics variance of cokriged values is lower in GSLIB.
Moreover, strange result was seen in cokriging variance in SGeMS v2.1 where the values are very small close to 0 for the block where samples exist. 
When I tried to check the cokriging variance in SGeMS v2.0, it was not identified (-999) for the block where samples exist. I think GSLIB produced more realistic ones.
 
Looking forward to any suggestions and comments.
 
Thanks for kind attention.
 
Best regards,
---------------
M. Nur Heriawan
Research Group of Earth Resources Exploration 
Faculty of Mining and Petroleum Engineering
Institut Teknologi Bandung (ITB)
Jl. Ganesha 10 Bandung 40132 INDONESIA