combining MAXENT output (probability) maps

[magz]

Hi,
We have started using MAXENT to model the distribution of habitat
types each of which has distinct characteristic species.

MAXENT is showing promising results with the relative probability of
distribution for each habitat type. However, we need to combine the
individual probability maps from MAXENT to produce a composite map
showing the overall distribution of habitat types.

For any given location we want to select the habitat type with the
highest probability of occurrence. Since MAXENT produces only relative
probability maps for each habitat type, whichever output option is
selected, I am not sure how to go about this correctly. If I
understand correctly 25% probability in one map will not be equivalent
to 25% probability in another, is this correct? If so how can we
really know which is the highest probability and therefore which map
layer (habitat) to select when combining the maps?

All our maps are based on the same set of environmental layers, does
this have any impact on the relative probabilities we will get output?

Any advice or suggestions gratefully received, thanks!
Margaret

[Marnin Wolfe]

Margaret,

I'd love for others to weigh in on this, but I have combined maps simply by adding them to each other. The result of that is, I think, still equivalent except you might get an area with a value of >1 that represents greater probability because multiple habitat-types support the suitability of that site for your focal species. You could standardize the values to between 0 and 1 for ease of interpretation. 

Cheers,

Marnin

--
You received this message because you are subscribed to the Google Groups "Maxent" group.
To post to this group, send email to max...@googlegroups.com.
To unsubscribe from this group, send email to maxent+un...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/maxent?hl=en.

--
Marnin Wolfe
University of Pittsburgh
Department of Biological Sciences
Ecology & Evolution Program
wol...@gmail.com (or)
md...@pitt.edu
239-595-5081

[Marnin Wolfe]

Alternatively, since we think of the output as a probability we should consider the product rule of probability. If you multiply the maps together then you are asking if the probability the species will be predicted present by, for example, Model 1 AND Model 2 AND Model 3, etc. 

[Kevin Buffington]

Margaret,

I'm looking at doing the same thing and am running into the same
questions. One thing I've considered is standardizing the different
outputs so they are all on the same scale, than choosing the layer
with the highest probability for that pixel. I haven't come up with a
good method to do this however, if I figure something out, I'll be
sure to share. I don't think simply adding or multiplying layers
together will work since you won't know which habitat to assign.

Cheers,
Kevin

[Elizabeth Jones Sbrocco]

Can you decide on a common threshold to create binary suitability maps (1 = suitable, 0 = unsuitable) and sum over those? I believe the program has the capability to create an output file based on your choice of threshold, though I don't have it open in front of me to tell you how to do it.

[magz]

Hi again, good to see there are others with interest in this topic and
thanks for the suggestions.

Combining the maps to pick the one with highest probability is not the
problem - if you use ArcGIS with Spatial Analyst there is a highest
position tool that does this nicely to produce a composite map.
Probably similar things exist in other GIS.

My concern is that the probability values output by MAXENT are not
comparable between maps so how will I really know which is the highest
probability, and therefore get correct boundaries between my habitat
types. I think the product rule suggested above may be worth a look,
but any further comments welcome, thanks.

Margaret

[magz]

Thanks for all the interest in this topic :)
From reading the MAXENT help it looks to me that both the Cumulative
and Raw options also output a relative probability for each output map
- it is just more evenly scaled than the Logistic output so I don't
think this helps with the problem that the probability values don't
mean the same in each map.
Please correct me if I am wrong.

On Feb 5, 10:21 am, Aidin <nia...@gmail.com> wrote:
> Hi
>
> I would say go for "Cumulative" or "Raw" outputs then comparision will
> make sense.
>
> Cheers
>
> Aidin
> Faculty of  Geo-information Science and Earth Observation (ITC)
> University of Twente
>
> On Feb 4, 11:52 pm, Elizabeth Jones Sbrocco <ladybluede...@gmail.com>

> > > For more options, visit this group athttp://groups.google.com/group/maxent?hl=en.-Hide quoted text -
>
> > - Show quoted text -