low auc score

[Jenna Hamlin]

Hi all,

I am trying to develop SDMs for two sister species. One of those species has a pretty widespread distribution while the other species distribution is found within the distribution of the wider ranging species.

I have successfully followed the Hijmans and Elith turtorial using R with the dismo package.
For the species with the wide distribution, the auc value does not go beyond .6 however; a quick run following the same procedure for the other species, with the narrow distribution, generates an auc score of about .8

I have removed highly correlated environmental variables, which leaves 6 BioClim variables that I am working with, which are: 2, 4, 7, 8, 15, & 18.

Am stuck on what is causing such a low auc score for the widely distributed species and how I might go about improving it.

Thanks!

Jenna

[Mike Treglia]

Hi Jenna,

Have you tried just looking at scatter plots of the presence data with individual variables? You might see that those variables tours using might not have any clear or obvious relationship to the species' presence. Are you using just bioclim vars? If so, maybe the species just isn't that sensitive to climate compared to other variables...? 

Hope that helps a bit at least...

Mike

--
You received this message because you are subscribed to the Google Groups "Maxent" group.
To unsubscribe from this group and stop receiving emails from it, send an email to maxent+un...@googlegroups.com.
To post to this group, send email to max...@googlegroups.com.
Visit this group at http://groups.google.com/group/maxent.
For more options, visit https://groups.google.com/groups/opt_out.

[Francisco Rodriguez Sanchez]

Hi Jenna,

The low AUC for a widely distributed species is expected, e.g. see
http://onlinelibrary.wiley.com/doi/10.1111/j.1466-8238.2007.00358.x/abstract.
That will happen regardless the environmental predictors you use. If the
species happens nearly everywhere, models can hardly discriminate
between absences and presences (which AUC tries to convey). It would be
easy to increase AUC for this species just increasing the extent of the
study region (i.e. much beyond the species distribution range), but of
course that may not make biological sense... That's one of the reasons
why AUC has been severely criticized.

Hope this helps,

Paco


El 11/12/2013 23:18, Jenna Hamlin escribi�:

> --
> You received this message because you are subscribed to the Google
> Groups "Maxent" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to maxent+un...@googlegroups.com.
> To post to this group, send email to max...@googlegroups.com.
> Visit this group at http://groups.google.com/group/maxent.
> For more options, visit https://groups.google.com/groups/opt_out.

--
Dr Francisco Rodriguez-Sanchez
Forest Ecology and Conservation Group
Department of Plant Sciences
University of Cambridge
Downing Street
Cambridge CB2 3EA
United Kingdom
http://sites.google.com/site/rodriguezsanchezf

[Jenna Hamlin]

Mike,

Yes, I have looked at scatter plots, though a bit ago. So I will re-examine those.

Paco,

Thanks for attaching the paper and I do realize that the AUC has been criticized. I think I need to do more reading for how to evaluate the model.

Jenna

On Thu, Dec 12, 2013 at 4:52 AM, Francisco Rodriguez Sanchez <f.rodrig...@gmail.com> wrote:

Hi Jenna,

The low AUC for a widely distributed species is expected, e.g. see http://onlinelibrary.wiley.com/doi/10.1111/j.1466-8238.2007.00358.x/abstract. That will happen regardless the environmental predictors you use. If the species happens nearly everywhere, models can hardly discriminate between absences and presences (which AUC tries to convey). It would be easy to increase AUC for this species just increasing the extent of the study region (i.e. much beyond the species distribution range), but of course that may not make biological sense... That's one of the reasons why AUC has been severely criticized.

Hope this helps,

Paco

El 11/12/2013 23:18, Jenna Hamlin escribió:

Hi all,

I am trying to develop SDMs for two sister species. One of those species has a pretty widespread distribution while the other species distribution is found within the distribution of the wider ranging species.

I have successfully followed the Hijmans and Elith turtorial using R with the dismo package.
For the species with the wide distribution, the auc value does not go beyond .6 however; a quick run following the same procedure for the other species, with the narrow distribution, generates an auc score of about .8

I have removed highly correlated environmental variables, which leaves 6 BioClim variables that I am working with, which are: 2, 4, 7, 8, 15, & 18.

Am stuck on what is causing such a low auc score for the widely distributed species and how I might go about improving it.

Thanks!

Jenna
--

You received this message because you are subscribed to the Google Groups "Maxent" group.

To unsubscribe from this group and stop receiving emails from it, send an email to maxent+unsubscribe@googlegroups.com.

To post to this group, send email to max...@googlegroups.com.
Visit this group at http://groups.google.com/group/maxent.
For more options, visit https://groups.google.com/groups/opt_out.

--
Dr Francisco Rodriguez-Sanchez
Forest Ecology and Conservation Group
Department of Plant Sciences
University of Cambridge
Downing Street
Cambridge CB2 3EA
United Kingdom
http://sites.google.com/site/rodriguezsanchezf

--
You received this message because you are subscribed to a topic in the Google Groups "Maxent" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/maxent/LfAJjkc_2Ts/unsubscribe.
To unsubscribe from this group and all its topics, send an email to maxent+unsubscribe@googlegroups.com.

[Jose Maria Costa]

Usually, if your species occur in a small area inside your whole area the auc is good. For species which are widespread along your study use to be low. Check if your  projections are more or less  adjusted to your presence points. If not, maybe your independent variables not are the best one.'s. Check if the widespread specie could have too many false positives

--

You received this message because you are subscribed to the Google Groups "Maxent" group.

To unsubscribe from this group and stop receiving emails from it, send an email to maxent+un...@googlegroups.com.