## AI-GEOSTATS: Optimal Kriging parameters

Expand Messages
• Hi. I want to use Ordinary Kriging on an arbitrary dataset of X,Y, and Z values to estimate the Z values on a grid of arbitrary size/density. But I don t know
Message 1 of 3 , Jul 8, 2002
Hi.

I want to use Ordinary Kriging on an arbitrary dataset of X,Y, and Z
values to estimate the Z values on a grid of arbitrary size/density. But
I don't know what length and scale parameters to choose for the
semivariogram. So I need to answer the following questions. I'm looking
for guidance and resources, not necessarily definitive answers. When
answering, please keep in mind that I'm a computer programmer, not a
statistician, by education and experience. :-)

1. How does one measure the "goodness" or "badness" of a Kriging
estimate? E.g. when the bounds of the grid are fairly close to the
bounds of the dataset, I might expect the estimated surface of Z values
to have roughly the same number of "bumps" and "valleys" as the original
dataset (if discernible), and not too many flat regions. How do I
quantify such characteristics, and are there others I should be looking
for?
2. How does one arrive at the "optimal" length and scale parameters
for the semivariogram when doing ordinary Kriging, given these measures
of "goodness" and "badness"? (here's where my comp. sci education would
come in handy, if I knew the answer to #1)

Eva

[Non-text portions of this message have been removed]
• Variogram modeling is usually a pre-requisite for kriging and/or stochastic simulation. It s not usally something that you d want to automate in some sort of
Message 2 of 3 , Jul 9, 2002
Variogram modeling is usually a pre-requisite for
kriging and/or stochastic simulation. It's not
usally something that you'd want to "automate" in some
sort of computer program. Selection of a model type/range/sill
will usually be based on available sample points,
or analagous samples of the same origin as the dataset
one is looking at, complemented by a qualitative interpretation
of the spatial model. I guess one can try to "program"
this whole process from start to finish (exploratory
data analysis/variogram modeling/kriging) but this is
not at all recommended. Perhaps in some applications with
abundant data, yes, but probably not in a geoscience
setting.

You haven't told us what your applications are? Will you
be mapping some geological variable? Interpolating 6 million
pixels in an image file? Trying to gauge the distribution of a
certain species of rare tropical flower?

Syed

---- Original message ----
>Date: Mon, 8 Jul 2002 20:56:15 -0700
>From: "Eva Pierce" <logicgrrl@...>
>Subject: AI-GEOSTATS: Optimal Kriging parameters
>To: <ai-geostats@...>
>
>Hi.
>
>I want to use Ordinary Kriging on an arbitrary dataset of X,Y, and Z
>values to estimate the Z values on a grid of arbitrary size/density. But
>I don't know what length and scale parameters to choose for the
>semivariogram. So I need to answer the following questions. I'm looking
>for guidance and resources, not necessarily definitive answers. When
>answering, please keep in mind that I'm a computer programmer, not a
>statistician, by education and experience. :-)
>
>1. How does one measure the "goodness" or "badness" of a Kriging
>estimate? E.g. when the bounds of the grid are fairly close to the
>bounds of the dataset, I might expect the estimated surface of Z values
>to have roughly the same number of "bumps" and "valleys" as the original
>dataset (if discernible), and not too many flat regions. How do I
>quantify such characteristics, and are there others I should be looking
>for?
>2. How does one arrive at the "optimal" length and scale parameters
>for the semivariogram when doing ordinary Kriging, given these measures
>of "goodness" and "badness"? (here's where my comp. sci education would
>come in handy, if I knew the answer to #1)
>
>Eva
>
>
>
>

--
* 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
• Hi Eva You have your questions the wrong way round. Once you find the semi-variogram model for your particular application, the kriging system should povide
Message 3 of 3 , Jul 9, 2002
Hi Eva

You have your questions the wrong way round. Once you
find the semi-variogram model for your particular
application, the kriging system should povide you with
a measure of 'goodness' of the estimator. It is
usually called the 'kriging standard error' or
sometimes software provides the kriging variance.

book at
http://uk.geocities.com/drisobelclark/practica.html

It is only 125 A5 pages long and you can skip a lot of
RSMA article which says much the same thing in 500
words. Find this on
http://uk.geocities.com/drisobelclark/resume/Publications.html

Mind the capital P on Publications ;-)

Hope this helps
Isobel Clark

__________________________________________________
Do You Yahoo!?
Everything you'll ever need on one web page
from News and Sport to Email and Music Charts
http://uk.my.yahoo.com

--
* 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
Your message has been successfully submitted and would be delivered to recipients shortly.