Re: AI-GEOSTATS: Kt estimation variance
Large kriging variances have nothing to do with
negative weights. The two are completely different
phenomena - especially in Universal Kriging.
In Ordinary Kriging the estimation variance can become
as high as two times the total sill of the
semi-variogram. This is a theoretical fact(?) and you
can derive the answer by simply trying to estimate the
value at one point from a single sample at or beyond
the range of influence from this location. That is, a
weight of +1!!
In Universal Kriging, the weights often become
negative because the system is trying to force the
estimated point to lie on the trend of the samples.
Sometimes it can only do this by using negative
weights. There is no problem here, unless you have
extremely high 'erratic' residuals. In this case, you
should probably resolve that problem before trying
If the variance becomes very high in Universal Kriging
it is probably because you are extrapolating into
sparsely sampled areas. Remember you are trying to
estimate the trend as well as the 'residual' value and
this contributes to higher variances. You should widen
your search radius to include more samples than with
You can see how all this works with our free kriging
game. If your data is in Geo-EAS form or a simple CSV
file, you can read it into Krigame and see the
equations and the solutions. Then vary search
parameters etc to see how they are affected. You might
want to download our Tutorial 3 which discusses
Universal Kriging with the Wolfcamp data.
All available at
http://geoecosse.bizland.com/softwares with no
restriction on use or distribution.
Want to chat instantly with your online friends? Get the FREE Yahoo!
* To post a message to the list, send it to ai-geostats@...
* As a general service to the users, please remember to post a summary of any useful responses to your questions.
* To unsubscribe, send an email to majordomo@... with no subject and "unsubscribe ai-geostats" followed by "end" on the next line in the message body. DO NOT SEND Subscribe/Unsubscribe requests to the list
* Support to the list is provided at http://www.ai-geostats.org