effect of map size on prediction?

[Florian Bodner]

Hello!

first of all apologies if this is somehow a double post - I tried posting this two days ago, but somehow it does not show up.

I'm relatively new to MaxEnt, trying to do potential range modelling based on bio-climatic background layers and GPS locations of sample sites.
When trying different geographic scales I noticed pronounced differences in predictions, even though the data has not changed other than that the environmental layers are larger (resolution also did not change).
Areas that are all blue on the scale of a few provinces are turning green and yellow on multiple country scale.

Is this normal? How should I deal with this?
I find it rather worrying that the predictions would be influenced by map size.

Any advice would be most welcome.
Thanks!
Florian

[Derek Corcoran Barrios]

Hey Florian:

The change It's normal, even thought maxent is supossed to be presence only, it uses psrudoabsences, and in order to do that it asigns random points from the backround as absences. If your backround area is larger, absences will be on average further away from your "core" area, and thus your model will be more permisive. You could try different area sizes and check on values such as kappa or AUC for quick desissions.

Cheers 

--
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 https://groups.google.com/group/maxent.
For more options, visit https://groups.google.com/d/optout.

[Florian Bodner]

Hi Derek,

thanks for your answer!
Is there some way to avoid or at least standardise this behaviour?
I'm struggling to see how predictions can be meaningfully compared if changing the scale will effectively automatically change the prediction.
Is there a guideline to what scale to choose? Is this covered in some paper I should read?
Would increasing the maximum number of background points with area size help?

Sorry for the many follow up questions, and thanks again!
Florian

[Derek Corcoran Barrios]

Hey Florian:

Here's a good start http://onlinelibrary.wiley.com/doi/10.1111/ddi.12249/full, If you need help discussing this or figuring out How to do I can help

--

[Florian Bodner]

Hi Derek!

Thanks a lot for the link and the offer!

[Jamie M. Kass]

This paper will also be very useful as you think about the size and shape of your study region:

http://media.wix.com/ugd/e41566_6110b6e938a94ae29c0d856de89e9c0f.pdf

Jamie Kass
PhD Candidate
City College, NYC

[Florian Bodner]

Many thanks to you as well!