Dear Gregoire,

>
> Still, part of the question remained unanswered. We all agree that in
> ....
> The real issue was however the fact that the investigated dataset was
> not considered to present a pure nugget efffect!
> ....
> 
> What are thus the requirements for not observing this decrease and how
> (linearly?) are both correlations (spatial correlation of observed
> values versus the correlation between observations and residuals)
> linked?
>

Lets tackle the problem with pure theory (for simplicity with a second order 
stationary field with simple kriging):

Let us start from the simple equation :

Z(x)= R(x)+K(x)

Where Z(x) is the value of the field, K(x) is the kriging prediction and R(x) 
is the Residual. 

The kriging error is defined as var(R(x))

For the Simple Kriging predictor we have according to Gauss-Markov Theorem:
cov(R(x),K(x))=0 

And thus 
cov(Z(x),R(x))=cov(R(x)+K(x),R(x))=cov(R(x),R(x))+cov(K(x),R(x))=
=cov(R(x),R(x))=var(R(x))=Kriging Error

Thus the correlation is given by

cor(Z(x),R(x))=var(R(x))/sqrt(var(R(x))var(Z(x))=
sqrt(var(R(x))/sqrt(var(Z(x)))=
= Kriging Standard deviation / Standard deviation of the Field

Thus in simplified theory the  correlation is given by the square root of the 
ratio of the kriging error devided by the sill: 

cor(Z(x),R(x))=sqrt( Kriging error / sill of variogramm)

Does this formula explain your findings?

The important parameter would  thus not be directly the spatial correlation, 
but the kriging error resulting from it and the spatial arrangements of 
measurements. 


Best regards,
Gerald 

-- 
-------------------------------------------------
Prof. Dr. K. Gerald v.d. Boogaart
Professor als Juniorprofessor fuer Statistik
http://www.stat.boogaart.de
http://www.math-inf.uni-greifswald.de/statistik/  

B�ro: Franz-Mehring-Str. 48, 1.Etage rechts
e-mail: [EMAIL PROTECTED]
phone:  00+49 (0)3834/86-4621
fax:    00+49 (0)3834/86-4615   (Institut)

paper-mail:
Ernst-Moritz-Arndt-Universitaet Greifswald
Institut f�r Mathematik und Informatik
Jahnstr. 15a
17487 Greifswald
Germany
--------------------------------------------------

+
+ To post a message to the list, send it to [email protected]
+ To unsubscribe, send email to majordomo@ jrc.it with no subject and 
"unsubscribe ai-geostats" in the message body. DO NOT SEND 
Subscribe/Unsubscribe requests to the list
+ As a general service to list users, please remember to post a summary of any 
useful responses to your questions.
+ Support to the forum can be found at http://www.ai-geostats.org/

Reply via email to