meaning of clamping

[wilbersa]

Dear Maxers,
I have a question for you. I projected the distribution of a couple of
species into the A1B climate scenario. and after looking carefully the
outcomes, I noticed that I got three raster layers per species. the
fires with the name of the species, then another layer with the name
of the species plus the A1B climate (I assume this is the layer that
projects the species distribution into the A1B climatic scenario), and
the third layer ends with the word clamping. here an example:

1. Ara_militaris
2. Ara_militaris_A1B climate
3. Ara_militaris_A1B climate_clamping

Now. although I have some thoughts of what the last layer could be,
those are just guesses. I would appreciate a lot if someone give me
some light of what possibly the clamping layer may mean.
Also, I reviewed the maxentresults.cvs and I just found the columns of
AUC values for testing and training. do this values correspond to the
projected distribution on the current climate conditions?, or to the
projection onto the A1B climatic scenario?.

My best wishes,

Ro

[Nora Castañeda]

Hi Ro:

I've being working with this definition of clamping:

Clamping. What is it? One problem with projecting onto a past of
future climate landscape is that there are climate conditions in the
past and future that have no analogs today. For example, some regions
may be colder and wetter than any existing spot on today’s
landscape. These conditions are outside the range represented in
the training data. Projections onto such conditions are suspect, and
for many applications, you'll want to cookie-cutter them out.
Clamping “does that” for you, by determining which areas in the
climate landscape are outside the range represented by the training
data, and showing that to you visually. Also, the maxent3.0 algorithm
will downweight those areas in the final “suitability” prediction so
they are less likely to be labeled “suitable”.

You can find it in http://macrobiodiv.googlepages.com/Lab4MacroEcoBiol.htm

Hope it will be useful for you,


Nora

[wilbersa]

Nora,
Thanks for your detailed answer, and also the link. It seems to be
very useful for people combining both softwares. However, I still have
some questions (and this goes for everybody). Now, according to the
text in the link (see above), the clamping layer gives us info about
those cells with no precedent climatic conditions (when comparing the
current climatic conditions and the projected). However, I can see it
( the clamping layer) has values from 0 to 1(like the other layers).
Which is the threshold to identify those critic cells with different
climatic conditions?. Also, I used the Maximum training sensitivity
plus specificity logistic threshold, and realized that some cells with
values near to 1 in the clamping layer, were included in the projected
distribution of the species.
Cheers,
Ro

> You can find it inhttp://macrobiodiv.googlepages.com/Lab4MacroEcoBiol.htm

[asha_mc]

I'm having the same trouble, but according to this post:

http://groups.google.com/group/Maxent/browse_thread/thread/3dcc7eabbf29f5b2/a7058c193eef606f?lnk=gst&q=clamping#a7058c193eef606f

Steven Phillips agrees that subtracting clampings values from your
projected logistic values is an adequate way of dealing with this
issue, rather than a threshold approach.

Cheers

On Feb 6, 2:29 am, wilbersa <wilbe...@daad-alumni.de> wrote:
> Nora,
> Thanks for your detailed answer, and also the link. It seems to be
> very useful for people combining both softwares. However, I still have
> some questions (and this goes for everybody). Now, according to the

> text in the link (see above), theclampinglayer gives us info about

> those cells with no precedent climatic conditions (when comparing the
> current climatic conditions and the projected). However, I can see it

> ( theclampinglayer) has values from 0 to 1(like the other layers).

> Which is the threshold to identify those critic cells with different
> climatic conditions?. Also, I used the Maximum training sensitivity
> plus specificity logistic threshold, and realized that some cells with

> values near to 1 in theclampinglayer, were included in the projected

> distribution of the species.
> Cheers,
> Ro
>
> On Feb 4, 2:10 pm, Nora Castañeda <la.guane...@gmail.com> wrote:
>
>
>
> > Hi Ro:
>
> > I've being working with this definition ofclamping:
>
> >Clamping.  What is it?  One problem with projecting onto a past of
> > future climate landscape is that there are climate conditions in the
> > past and future that have no analogs today.  For example, some regions
> > may be colder and wetter than any existing spot on today’s
> > landscape.    These conditions are outside the range represented in
> > the training data.  Projections onto such conditions are suspect,  and
> > for many applications, you'll want to cookie-cutter them out.
> >Clamping“does that” for you, by determining which areas in the
> > climate landscape are outside the range represented by the training
> > data, and showing that to you visually.  Also, the maxent3.0 algorithm
> > will downweight those areas in the final “suitability” prediction so
> > they are less likely to be labeled “suitable”.
>
> > You can find it inhttp://macrobiodiv.googlepages.com/Lab4MacroEcoBiol.htm
>
> > Hope it will be useful for you,
>

> > Nora- Hide quoted text -
>
> - Show quoted text -