[R-sig-eco] kernel HR size compared to MCP

Mon Dec 8 14:49:59 CET 2008

Hi Roy,

This is a pretty typical behavior.  Because the kernel is a probability
model, it extends the home range boundaries beyond the edge of the data
cloud.  In contrast, the MCP is just a summary of the sampled points.
The size of the difference in estimates depends on the smoothing
parameter used in the kernel estimate (bigger = more extension), and on
the degree to which the MCP includes unused interior space. The latter
is a common problem with MCPs, and one of the main reasons that the
estimation literature almost universally recommends against their use.
You can find more details in Seaman and Powell's work, and the
subsequent lit on home range estimation.

-----Original Message-----
From: r-sig-ecology-bounces at r-project.org
[mailto:r-sig-ecology-bounces at r-project.org] On Behalf Of
r-sig-ecology-request at r-project.org
Sent: Saturday, December 06, 2008 6:00 AM
To: r-sig-ecology at r-project.org
Subject: R-sig-ecology Digest, Vol 9, Issue 6

Send R-sig-ecology mailing list submissions to
	r-sig-ecology at r-project.org

To subscribe or unsubscribe via the World Wide Web, visit
	https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
or, via email, send a message with subject or body 'help' to
	r-sig-ecology-request at r-project.org

You can reach the person managing the list at
	r-sig-ecology-owner at r-project.org

When replying, please edit your Subject line so it is more specific
than "Re: Contents of R-sig-ecology digest..."

Today's Topics:

   1. adehabitat: kernel HR size compared to MCP (Roy Sanderson)
   2. (no subject) (calenge at biomserv.univ-lyon1.fr)

----------------------------------------------------------------------

Message: 1
Date: Fri, 05 Dec 2008 16:08:41 +0000
From: Roy Sanderson <r.a.sanderson at newcastle.ac.uk>
Subject: [R-sig-eco] adehabitat: kernel HR size compared to MCP
To: r-sig-ecology at r-project.org
Message-ID:

<EMEW-kB4G4a96008269d3afa8a0b78958257142122a-1228493321.18734.186.camel@
clarinet>

Content-Type: text/plain

Hello SIG-Ecology

Could anyone shed any information on why the area of home ranges as
estimated in adehabitat can be several times higher when the kernelUD \
kernel.area functions are used compared to estimates from the mcp
function.  I would have expected them to be roughly the same, rather
than up to an order of magnitude different.  The feature is obvious even
in the example puechabon dataset that comes with the adehabitat package.

Using R version 2.8.0, adehabitat 1.7.1 on Ubuntu linux 8.04

Many thanks
Roy

-- 
Roy Sanderson,
Institute for Research on Environment and Sustainability (IRES),
Devonshire Building,
Newcastle University,
Newcastle upon Tyne NE1 7RU
r.a.sanderson at newcastle.ac.uk
0191 246 4835

------------------------------

Message: 2
Date: Fri, 05 Dec 2008 17:57:36 +0100
From: calenge at biomserv.univ-lyon1.fr
Subject: [R-sig-eco] (no subject)
To: r.a.sanderson at newcastle.ac.uk
Cc: r-sig-ecology at r-project.org
Message-ID: <1228496256.49395d8051d0b at webmail.univ-lyon1.fr>
Content-Type: text/plain; charset=ISO-8859-1

Hello Roy,

> Could anyone shed any information on why the area of home ranges as
> estimated in adehabitat can be several times higher when the kernelUD
\
> kernel.area functions are used compared to estimates from the mcp
> function.  I would have expected them to be roughly the same, rather
> than up to an order of magnitude different.  The feature is obvious
even
> in the example puechabon dataset that comes with the adehabitat
package.

Because the two methods differ: the MCP computes the smallest polygon
encompassing a given percentage of the relocations, while the kernel
method
tries to estimate the bivariate probability density function to relocate
the
animal at a given place (the utilization distribution, or UD). The
kernel home
range is estimated from the UD as the smallest area on which the
probability to
relocate the animal is equal to a given probability (e.g. 0.95, 0.7,
etc.).

Because the two methods do not rely on the same mathematical bases they
are not
expected to return the same results. Very often, the kernel home-range
estimate
is /smaller/ than the MCP (because the MCP may include large areas not
actually
used by the animal, while this is not the case for the kernel home
range). When
it is higher, it may be because the smoothing parameter h is
misspecified (as it
is the case for the "puechabon" dataset). Alternatively, it can be just
a
characteristic of your data.

Indeed, the ad-hoc or so-called "reference" method to estimate the
smoothing
parameter, i.e. the most widely used method in the ecological literature
(and
the default in adehabitat, but it can be changed with the parameter "h"
of
kernelUD), tends to overestimate the home-range size, as this method
supposes
that the UD is bivariate normal, which is actually rarely the case.
There is a
large literature on this issue. You may begin with:

Worton, B. 1995. Using Monte Carlo simulation to evaluate kernel-based
home
range estimators, Journal of Wildlife Management, 59, 794-800
Worton, B. 1989. Kernel methods for estimating the utilization
distribution in
home range studies. Ecology, 70, 164-168.

Hope this helps,

Cl?ment.

-- 
Cl?ment CALENGE
Office national de la chasse et de la faune sauvage
Saint Benoist - 78610 Auffargis
tel. (33) 01.30.46.54.14

------------------------------

_______________________________________________
R-sig-ecology mailing list
R-sig-ecology at r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology

End of R-sig-ecology Digest, Vol 9, Issue 6