AI-GEOSTATS: Dealing with Universal Kriging

[Rubens Caldeira Monteiro]

Dear all,   We are trying to apply Universal Kriging to “High Plains” Aquifer in Kansas (OLEA, 1999) for land surface elevation (LSE), using its 317 data points. The purpose of this application is just for didactic ends. Our first step was to filter a prominent 1st degree drift. The way we did it was using Surfer 8.0 (Golden Software) and obtaining the residuals, in such way that summing the residuals by the 1st degree trend we obtain the original data. Obviously the values of this new variable (residuals of LSE, i.e., RLSE) are much lower than the original one. Calculating the experimental variogram, modeling it and kriging the variable is possible to obtain the RLSE (residuals of land surface elevation) map. Summing this map to the drift (calculated in a deterministic way) we obtain a map that we suppose that represents a map for the original variable (LSE). But what about the standard deviation? We did a little test and it seems that the standard deviation map for the residuals (RLSE) represents the std. dev. map for the original variable (LSE).   Is this a correct conclusion and procedure? If we have less data, will it work? If we use a 2nd degree drift the standard deviation could be wrong?   Thanks for your attention,   Rubens   ========================== Rubens Caldeira Monteiro # ICQ 106157533 São Paulo State University at Rio Claro  - UNESP/Rio Claro – PhD candidate on Geosciences & Environmental Sci. São Paulo University at Piracicaba - ESALQ-USP/Piracicaba – Undergrad. on Environmental Management University of North Carolina at Chapel Hill - UNC-CH URL: http://www.unc.edu/~rubenscm/CASEhome.html