12 Feb 2009 05:05

### Re: AI-GEOSTATS: theoretical sill

```Figen-
Your description is a bit confusing without pictures, but a few points might be worth saying more generally.

First, if your data are covariance stationary, then it is true that the sill is the population variance.
However, the sample variance is usually a biased (downward) estimate of the population variance.  Thus,
you should almost always find the sill to be larger than the sample variance.  Despite the bias of the sample
variance, it is asymptotically unbiased as the sample domain gets larger.  So whether or not you should
consider the sample variance comes down to an educated choice, "do you think your domain is large enough so
that the bias of the sample variance is small?"

Second, It sounds, however, like there is some sort of trend in your data, thus the data  are not even
covariance stationary, and there won't be a theoretical variance.  Of course, you can always calculate
the sample variance, but this is meaningless if the data are not covariance stationary.

You could spend a lot of effort trying to get the zonal anisotropy modeled well, with the sill possibly
different along the major and minor axes (by picking large ranges as you suggest), but if you are
interested in the predictions,  then the most critical part is not to model the sill, but to properly model
the semivariogram at small distances.

Nicholas

Nicholas N. Nagle, Assistant Professor
Department of Geography
UCB 260, Guggenheim 110
Boulder, CO 80309-0260
phone: 303-492-4794

---- Original message ----
>Date: Wed, 11 Feb 2009 12:31:51 -0600
```

11 Feb 2009 19:31

### AI-GEOSTATS: theoretical sill

```Dear List,

This is a bit long email. Couldn't figure out how to ask shorter.
I have couple questions about the sample variance and modeling 3D
variograms. In literature I found that some writers strongly recommend
that we have to use sample variance as a theoretical sill and anything
above will indicate a trend. However some papers claim that sample
variance isn't actually representing the total population's variance
so  shouldn't be considered. I am working on 3D modeling. I have
finally had a chance to have a software (GSLIB) that I can check
anisotropy. In GISLIB theoretical sill is optional. Now my questions
are how important do you think theoretical sill? What if you have a
structure that decreases with the distance as it is supposed to be but
the sill is above the theoretical sill?
I have read Isobel Clark's tutorials and followed her and GISLIB guide
and created some variograms for my data:
0, 45, 90 and 135 degrees horizontal,
0, 45, 90, 135, 180, 225, 270 and 315 at a dip of 45 down from horizontal
for 0,45,90 and 135 at dip 90
the vertical variogram clearly shows some trend after 8m which I am
not interested in and also the pair numbers after 8m is very less. The
horizontal variogram is almost flat with a little decrease in gamma at
the beginning and at the end. It is totally above the theoretical
sill. Also it is above the theoretical sill. At this point I am
confused. Could this be the effect of trend in vertical direction? In
45 degree dip down from horizontal the gamma values also increased
above theoretical sill and as the distance increased, the values
decreased??? Without considering the theoretical sill I can say that
vertical variogram shows trend and there is geometric and zonal
```

9 Feb 2009 16:42

```
Dear all,

I agree with those of you that have replied to me. The access to the book is
likely to be illegal.

without thinking twice.

I will inform the authors about the issue.

I apologize for that...

Gregoire

-----Original Message-----
From: Lena Szymanek [mailto:lena@...]
Sent: Monday, February 09, 2009 3:18 PM
To: Gregoire DUBOIS (European Commission)

Dear Gregoire,

I think this is not a legal way to get this book, unless the authors
confirm this action.

Best regards

Lena Szymanek

```

9 Feb 2009 14:40

```Dear all,

You may find the following book interesting for you, especially because you

Richard Webster, Margaret A. Oliver. "Geostatistics for Environmental
Scientists"
(Statistics in Practice)
Wiley | 2007-12 | ISBN: 0470028580 | 332 pages | PDF | 5 Mb

http://www.ebooksz.cn/2009/02/geostatistics-for-environmental.html

Best regards

Gregoire

PS: I hope this message will not arrive 3 times to you

++++++++++++++++++++
Gregoire Dubois (PhD)

DG Joint Research Centre - European Commission
Institute for Environment and Sustainability
Global Environment Monitoring Unit
Monitoring Of Natural resources for DEvelopment (MONDE)

http://gem.jrc.ec.europa.eu
http://www.ai-geostats.org

Via Fermi, TP 440,  I-21027 Ispra (VA), ITALY
```

23 Jan 2009 17:16

### AI-GEOSTATS: Special Issue on Information Semantics and its Implications for Geographical Analysis

```Hello,

Several on this list may be interested in this call. We are especially
interested in publishing work employing spatial analytical methods, including
geostatistics, on spatial semantic data.

---------------------------------------------------------------------------
2nd Call for Papers – Int. Journal of Geographical Information Science
---------------------------------------------------------------------------

Special Issue on
Information Semantics and its Implications for Geographical Analysis
http://www.cogsci.uni-osnabrueck.de/~isga08/SpecialIssue-IJGIS.pdf

Editors:
Ola Ahlqvist, The Ohio State University, U.S.A.
Martin Raubal, University of California, Santa Barbara, U.S.A.
Angela Schwering, University of Osnabrueck, Germany
Ashton Shortridge, Michigan State University, U.S.A.

Ontology and information semantics are central to research on enhanced
interoperability between geographic information systems, services, and data
sets. It is increasingly understood that deeper insights into information
semantics also have consequences for geographic analysis in general.
array of analytical methods for categorical data than those traditionally
employed. Existing examples that demonstrate this potential are fuzzy
accuracy assessment, map similarity estimates, and semantic versions of the
standard variogram. An explicit recognition of the importance of context in
semantic assessments also poses interesting questions for a quantitative
```

14 Jan 2009 09:31

### AI-GEOSTATS: Geostats 2008, Santiago, presentations

```Dear list,

some (most) of the presentations of the Geostats 2008 conference, Santiago
1-5 Dec 2008, are now available on-line as pdf's,
http://www.geostats2008.com/cierre2008/

Peter

-----------------------------------------------------
Peter Bossew

European Commission (EC)
Joint Research Centre (JRC)
Institute for Environment and Sustainability (IES)

TP 441, Via Fermi 1
21020 Ispra (VA)
ITALY

Tel. +39 0332 78 9109
Fax. +39 0332 78 5466
Email: peter.bossew@...

www:  http://rem.jrc.ec.europa.eu

"The views expressed are purely those of the writer and may not in any
circumstances be regarded as stating an official position of the European
Commission."

+
```

12 Jan 2009 13:53

### AI-GEOSTATS: search gam mex sources to use MATLAB Kriging Toolbox on Linux operating system

Dear all,
In order to use MATLAB Kriging Toolbox on Linux operating system, I need gam(s) and ksone mex sources interfacing GSLIB to matlab for compilation.
If anybody can help me to get mex sources, I would be happy.
Happy new year

Frédéric HUGUET

9 Jan 2009 11:52

### AI-GEOSTATS: stratifying the spatial domain on the basis of dependent variable

Dear All,

I'm experiencing some doubt about the logic consistency of the procedure I'm following to analyse my data.
The dataset I'm dealing with, comprises 360 sampling points; for each one of this points I have a measure of nitrate concentration in groundwater.
What I'm trying to do is to assess the groundwater vulnerability to nitrate contamination.
The first step has been to interpolate the contamination data from the point observations.
My idea was to approach the problem using regression kriging (given the availabilty of numerous candidate covariates, such as: land use, rainfall, clay content in soils, etc.).

However, after a preliminary look at the data I've found that my data is bimodal. This was not a big concern as the residuals of linear regression are still normal, although strongly heteroskedastic (but this won't affect the estimates of the coefficients, if I recall correctly my statistics courses).

Anyway, after a second look at the data I've tried to separate the two "populations" considering them as a mixture of two normal distributions, the outcome of this clustering shows that each one of these "populations" have a different spatial distribution, as one tends to occupy the northern part of my study area and the other the southern. This difference is not due to anyone of the covariates available, or any factor I can think of...

So the next step I tried was to treat these areas as two different strata fitting a mixed model instead of the simple linear one. This has effectively increased the accuracy of the prediction, because the two groups seem to have different coefficients beside a different intercept.

However what puzzles me is the logic of the procedure. Using the outcome of a process as an input in the estimation of the same outcome sounds like circular logic to me... is this procedure correct?
I know that similar problems are addressed by latent class analysis, but I couldn't find anything about this kind of problems in spatial statistics.
If anybody have dealt with similar problems (or knows of some literature about), I'll gladly hear any suggestion.

Cristiano Ballabio

Environmental Sciences Dept. (DISAT)

University of Milano-Bicocca (UNIMIB)

P.za della Scienza, 1

20126 MILANO

Italy

Phone office:      +39 02 64482881

+39 02 64482876

Fax:                  +39 02 64482895

8 Jan 2009 10:53

### AI-GEOSTATS: New year

Hi Gregoire

I replied some days ago but the mail returned back...some problems with the server ...

So, happy new year "ai-geostats"!
I guess that this year the mailing list will be less quiet...

I'm also working with lidar data and using spatial continuity indexes to describe and eventually classify topographic surface.
I think that high resolution topographic data, given the complexity of morphogenetic processes and consequently the complexity
of spatial patterns, could be really useful for a SIC exercise.

Bye
Sebastiano

Dear all,

In contrast to previous years, AI-GEOSTATS has been very quiet in 2008 and I hope 2009 will see more discussions on the list. May I remind everyone to reply as much as possible to the mailing list rather than only to the author of a question?

I was planning to organize by the end of 2008 a new SIC (spatial interpolation comparison) exercise focusing on classification but didn
8 Jan 2009 09:31

### AI-GEOSTATS: Problems with emails

Dear all,

2009 is starting with a few mail problems (mails sent to the list are distributed 3-4 times to a number of subscribers). I hope this will be fixed soon.

I apologize for the inconvenience.

Best regards

Gregoire (moderator)

++++++++++++++++++++

Gregoire Dubois (PhD)

DG Joint Research Centre - European Commission

Institute for Environment and Sustainability

Global Environment Monitoring Unit

Monitoring Of Natural resources for DEvelopment (MONDE)

http://gem.jrc.ec.europa.eu

http://www.ai-geostats.org

Via Fermi, TP 440,  I-21027 Ispra (VA), ITALY

Tel : +39 0332 786360

Fax : +39 0332-789960

"The views expressed are purely those of the writer and may not in any circumstances be regarded as stating an official position of the European Commission." (Disclaimer required under the terms and conditions of use of the internet and electronic mail from Commission equipment)"

7 Jan 2009 17:28

### AI-GEOSTATS: 3D Kriging neighborhood size

```All,

First of all a happy 2009 to everyone!

I have a few (beginner?) questions about the neighborhood size (number of points) for Kriging, in
particular in 3D:

1) Firstly, I would just like to hear some user experiences - what number have you used in the past? Was that
3D? What range of numbers would you normally test?

2) If I understand correctly, Kriging weights can become negative, but I get the impression that normally
the large majority of the weights are positive. Could I therefore assume that if I use 100 points, then the
smallest weights are likely to be (much) smaller than 0.01?

3) I understand that (except for simple Kriging), it can be usefull to use a larger search neighborhood than
the variogram range. What about the opposite, if you have relatively dense sampling, and there are many
points within, say, one tenth of the range?

Many thanks,
Greg

+
+ To post a message to the list, send it to ai-geostats@...
+ To unsubscribe, send email to majordomo <at>  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/

```

Gmane