## GEOSTATS: Modeling Spatial Distribution of Landmines

Expand Messages
• In response to Edith Wilkinson s request: Edith, sounds to me like you will end up trying to characterize an almost pure spatial Poisson distribution of small
Message 1 of 6 , Feb 3, 2000
In response to Edith Wilkinson's request:

Edith, sounds to me like you will end up trying to characterize an almost
pure spatial Poisson distribution of small size samples (the mines) over an
extremely large area (if I understand the problem correctly).

Applications to sampling of spatial Poisson processes (SPP's) that I am
familiar with have used spatial statistics to model the intensity (mean) of
the Poisson process as spatial variable. If the mean number of landmines
left behind for any subdomain varies spatially, geostatistical techniques,
such as kriging, could be used to model the mean, and more elaborate models
(e.g., clustering models) can be used to describe the actual distribution of
the mines within each subdomain. SPP's with variable intensities are a form
of "Cox" processes and also referred to as "double stochastic Poisson
processes". Ideally, you may be able to find some ancillary (soft)
information that can be correlated to the intensity of the SPP and
incorporate this info through a cokriging procedure.

A good example of modeling SPP's with geostatistics is:

Caers, J., J. Gelders, A. Vervoort and L. Rombouts, 1997, Non-Conditional
and Conditional Simulation of a Spatial Point Process, in: Baafi and
Schofield (eds.) Geostatistics Wollongong '96, Kluwer Academic Publishers

There is good info on Poisson processes in the book: "An Introduction to
Stochastic Modeling" by Taylor and Karlin

If, in fact, the left over landmines are located in a purely random fashion
through the sample area, spatial statistics won't be much help and your
sampling costs to get to the required confidence of an area being clean will
be astronomical.

Sean

Sean A. McKenna
Geohydrology Department
Sandia National Laboratories
PO Box 5800 MS 0735
Albuquerque, NM 87185-0735
ph: 505 844-2450
fx: 505 844-4426

--
*To post a message to the list, send it to ai-geostats@....
*As a general service to list users, please remember to post a summary
of any useful responses to your questions.
*To unsubscribe, send email to majordomo@... with no subject and
"unsubscribe ai-geostats" in the message body.
DO NOT SEND Subscribe/Unsubscribe requests to the list!
• ... I am loath to think that landmines should follow some spatial process, other than being purely coincidental, and subject to the whims and fancies of a
Message 2 of 6 , Feb 3, 2000
>If, in fact, the left over landmines are located in a purely random fashion
>through the sample area, spatial statistics won't be much help and your
>sampling costs to get to the required confidence of an area being clean will
>be astronomical.

I am loath to think that landmines should follow some spatial
process, other than being purely coincidental, and subject
to the whims and fancies of a human being with questionable
motives (apart from blowing up the enemy). Just as with any
"geostatistical" problem, I would first attempt to get an understanding
of the genesis of such a process, i.e. why would some landmines
be planted at certain locations of a particular terrain, taking
routes, distance from munitions base, enemy population, etc.

In other words, would such a landmine distribution involve some
sort of spatial component, and if yes, should I be modeling it?

This can be as futile an exercise as the prediction of oil prices
over time, what with such variables as political instabilities,
worldwide demand, technological advances, etc. Likewise,
stock market performance..

Syed

--
*To post a message to the list, send it to ai-geostats@....
*As a general service to list users, please remember to post a summary
of any useful responses to your questions.
*To unsubscribe, send email to majordomo@... with no subject and
"unsubscribe ai-geostats" in the message body.
DO NOT SEND Subscribe/Unsubscribe requests to the list!
• Ms Wilkinson stated:
Message 3 of 6 , Feb 4, 2000
Ms Wilkinson stated:
<<
I am currently working on a project to evaluate the safety of landmine
fields. I am therefore mainly interested in the estimation of the completely
unknown "contamination" level of the field. The variable of interest is
discrete (landmines) and the populations I will study range from 1000 m2 to
100 million m2.

I am studying landmine fields after clearance. Therefore there should not be
any patterns in the "missed" landmines I am trying to evaluate. Their
density is totally unknown as their presence is a result of a clearance
process which can not be guaranteed to 100 %.
>>
Syed noted:
<
Just as with any "geostatistical" problem, I would first attempt to get an
understanding of the genesis of such a process, i.e. why would some
landmines
be planted at certain locations of a particular terrain, taking into account
munitions base, enemy population, etc.

In other words, would such a landmine distribution involve some sort of
spatial component, and if yes, should I be modeling it?
>

This location of remaining landmines appears to be the the sum of three
processes: 1) putting in the mines, 2) their detonation, and 3) their
removal.

1) would seem to be a combination of military doctrine (strategy - where to
put them and desired local pattern) - and tactical result (actual locations,
methods used to place the mines.)

2) would perhaps have eroded the outer portions of (uniform) areas. More
recent data are presumably what you have to work from.

3) The strategy here may have fewer options, but the tactics may be better
known, and there should be more information. Do (did) the clearers see any
patterns during their work?

It may be feasible to set up a classification of areas, based on information
of 1 & 3, that relate to parameters of the distribution.

--------
Amateurs discuss strategy, professionals discuss logistics. - anon.
--------

--
*To post a message to the list, send it to ai-geostats@....
*As a general service to list users, please remember to post a summary
of any useful responses to your questions.
*To unsubscribe, send email to majordomo@... with no subject and
"unsubscribe ai-geostats" in the message body.
DO NOT SEND Subscribe/Unsubscribe requests to the list!
• However, in one sence, the placement of the mines could be a function of terrain or geology as I suggested earlier. Let s face it, landmines are used to kill
Message 4 of 6 , Feb 4, 2000
However, in one sence, the placement of the mines could be a function of
terrain or geology as I suggested earlier. Let's face it, landmines are
used to kill people, so therefore, there is likely a higher probablity that
they will be placed in areas where people are likely travel. For example,
near roads or along terrain that is easily traversed. They might be less
likely in areas where bedrock is near the surface or areas of heavy brush,
wetlands, or rocky terrain. Using parameters such as this, you could define
areas of probable mine locations based on a number of independent variables
(as previously stated). Geostatistical analysis could certainly be done
with reasonable certainty for topographic elevation and near surface
soils/geology etc. For example, would not areas that are relatively flat,
between a road and say a river have a higher probability of having landmines
than say a rocky area with heavy brush away from main traffic zones???? If
you feel this is a feasible approach, then a variation of coKriging relating
a number of independent variables with possibly a stochastic approach might
be a reasonable way of determining high "danger" areas.

Hope this helps. Let me know how it goes.
Regards,
Frank
*************************************

Frank Breen
Breen GeoScience Management, Inc.
25116 W. Catherine Drive
Plainfield, Il. 60544

phone/voicemail: 815.254.3834
fax/voicemail: 800.471.4855

fab@...

*************************************

> -----Original Message-----
> From: owner-ai-geostats@...
> [mailto:owner-ai-geostats@...]On Behalf Of srahman@...
> Sent: Thursday, February 03, 2000 11:58 PM
> To: McKenna, Sean A
> Cc: 'Wilkinson Ms E'; 'ai-geostats@...'
> Subject: Re: GEOSTATS: Modeling Spatial Distribution of Landmines
>
>
>
> >If, in fact, the left over landmines are located in a purely
> random fashion
> >through the sample area, spatial statistics won't be much help and your
> >sampling costs to get to the required confidence of an area
> being clean will
> >be astronomical.
>
> I am loath to think that landmines should follow some spatial
> process, other than being purely coincidental, and subject
> to the whims and fancies of a human being with questionable
> motives (apart from blowing up the enemy). Just as with any
> "geostatistical" problem, I would first attempt to get an understanding
> of the genesis of such a process, i.e. why would some landmines
> be planted at certain locations of a particular terrain, taking
> routes, distance from munitions base, enemy population, etc.
>
> In other words, would such a landmine distribution involve some
> sort of spatial component, and if yes, should I be modeling it?
>
> This can be as futile an exercise as the prediction of oil prices
> over time, what with such variables as political instabilities,
> worldwide demand, technological advances, etc. Likewise,
> stock market performance..
>
> Syed
>
>
> --
> *To post a message to the list, send it to ai-geostats@....
> *As a general service to list users, please remember to post a summary
> of any useful responses to your questions.
> *To unsubscribe, send email to majordomo@... with no subject and
> "unsubscribe ai-geostats" in the message body.
> DO NOT SEND Subscribe/Unsubscribe requests to the list!
>

--
*To post a message to the list, send it to ai-geostats@....
*As a general service to list users, please remember to post a summary
of any useful responses to your questions.
*To unsubscribe, send email to majordomo@... with no subject and
"unsubscribe ai-geostats" in the message body.
DO NOT SEND Subscribe/Unsubscribe requests to the list!
• ... used to kill people, so therefore, there is likely a higher probablity that they will be placed in areas where people are likely travel.
Message 5 of 6 , Feb 4, 2000
>Let's face it, landmines are
used to kill people, so therefore, there is likely a higher probablity that
they will be placed in areas where people are likely travel.<

Actually, killing people is seldom a purpose of landmines. Most doctrines
(again, for the armies advanced enough to follow doctrines) suggest that an
injury is preferable to death. This is primarily because an injury is more
of an immediate burden on the enemy's logistical system (takes two soldiers
to carry one stretcher, etc.). Furthermore, mines are often used as an
area denial weapon. Armies can put mines out where they don't want anyone
(including themselves) to go. In most cases, mine warfare doctrine opens
with something like "Mines are a weapon that attacks the enemy's ability to
maneuver..." Antitank mines go after vehicles and antipersonnel mines both
protect the antitank mines and go after walking personnel.

Your fundamental observation that we should be able to assign probability
based on known placement tactics is, however, a good one. In this case,
domain knowledge should include mine warfare doctrine of the armies (or
terrorists, or para-military, etc.) that placed the mines.

Good luck,
Steve Riese

SteveRiese@...
--
*To post a message to the list, send it to ai-geostats@....
*As a general service to list users, please remember to post a summary
of any useful responses to your questions.
*To unsubscribe, send email to majordomo@... with no subject and
"unsubscribe ai-geostats" in the message body.
DO NOT SEND Subscribe/Unsubscribe requests to the list!
• Well in response to ignorance regarding the use of landmine, I am happy to say I have had no experience with mines. I do, however, find the problem an
Message 6 of 6 , Feb 4, 2000
Well in response to ignorance regarding the use of landmine, I am happy to
say I have had no experience with mines. I do, however, find the problem an
interesting one and hope you can find a reasonable approach to solving this
problem.

Regards,
Frank

*************************************

Frank Breen
Breen GeoScience Management, Inc.
25116 W. Catherine Drive
Plainfield, Il. 60544

phone/voicemail: 815.254.3834
fax/voicemail: 800.471.4855

fab@...

*************************************

> -----Original Message-----
> From: Stephen R. Riese [mailto:SteveRiese@...]
> Sent: Friday, February 04, 2000 2:43 PM
> To: Frank Breen
> Cc: Stephen R. Riese; McKenna, Sean A; [unknown]; [unknown]; 'Wilkinson
> Ms E'
> Subject: RE: GEOSTATS: Modeling Spatial Distribution of Landmines
>
>
> >Let's face it, landmines are
> used to kill people, so therefore, there is likely a higher
> probablity that
> they will be placed in areas where people are likely travel.<
>
> Actually, killing people is seldom a purpose of landmines. Most doctrines
> suggest that an
> injury is preferable to death. This is primarily because an
> injury is more
> of an immediate burden on the enemy's logistical system (takes
> two soldiers
> to carry one stretcher, etc.). Furthermore, mines are often used as an
> area denial weapon. Armies can put mines out where they don't want anyone
> (including themselves) to go. In most cases, mine warfare doctrine opens
> with something like "Mines are a weapon that attacks the enemy's
> ability to
> maneuver..." Antitank mines go after vehicles and antipersonnel
> mines both
> protect the antitank mines and go after walking personnel.
>
> Your fundamental observation that we should be able to assign probability
> based on known placement tactics is, however, a good one. In this case,
> domain knowledge should include mine warfare doctrine of the armies (or
> terrorists, or para-military, etc.) that placed the mines.
>
> Good luck,
> Steve Riese
>
> SteveRiese@...
>

--
*To post a message to the list, send it to ai-geostats@....
*As a general service to list users, please remember to post a summary
of any useful responses to your questions.
*To unsubscribe, send email to majordomo@... with no subject and
"unsubscribe ai-geostats" in the message body.
DO NOT SEND Subscribe/Unsubscribe requests to the list!
Your message has been successfully submitted and would be delivered to recipients shortly.