Relationship between forest edge habitat and abundance - thoughts or suggestions ?

[JulianeStruve1]

Dear ArcGIS people,

I am trying to figure out how to examine the effect of forest edge habitat (independent variable) onto the abundance of certain species (dependent variable) associated with those edges. Abundance data consist of spatially distributed counts in a study area that includes the edge habitat and large spaces without it.

It is believed that the habitat effect may include not only presence of the edge habitat, but also length of the edge and possibly width. Complexity of the shape of the edge may also play a role.

I am looking at different ways to describe the habitat effect and have so far come up with:
1. Distance from the habitat edge
2. Presence of certain tree species within the immediate neighborhood of the dependent variable, defined as a round buffer.

I am thinking of intersecting the buffers with the edge habitat feature, and to make the neighborhood succcessively larger in order to test if the effect of the edge habitat onto abundance may be dependent on the scale of observation.

I was wondering if anyone has experience with such analysis, and what problems may arise. There are a lot of zeros in the data where no animals were found.  I have not figured out yet how to describe the complexite of the edge itself. It appears that the Patch Analyst extension was written for some related questions, but I have not been able to download it for ArcGIS 10.

Thank you very muh for any suggestions and best wishes,

Juliane

[SebastianSantibanez]

I think that the concept of "edge" is rather hard to define, specially when you are analyzing its influence in the presence of certain species.  Instead of trying the hard-boundary "buffer" I would consider using fuzzy logic. 
Take a look at the fuzzy tools in spatial analyst, they can give you a few ideas.

[JeffreyEvans]

I would take a look at Fragstats. There is a moving window option that will allow you to calculate metrics at multiple scales. You can then assign these surface values to your sample locations. Some common edge metrics are edge length and fractal dimension. Edge characteristic is an important component in this type of analysis that a mere distance measures will not get at. Although, you may see an interaction with distance. I should note that a quick Google search for "edge landscape metrics" provided a number of hits relevant to your question. Some basic research on your part is in order! 

Since it sounds like you have a zero inflation issue, you will certainly need to account for this during statistical analysis. You may consider a ZIP (Zero Inflated Poisson) regression model. If you are really meaning species richness when referring to abundance please note that there are many problems associated with richness and you may want to consider a measure of diversity such as Shanon's H'. This index will represent alpha and beta diversity with an option for an evenness correction and will also scale across your system. If you are addressing single specie counts, then abundance is an appropriate measure although, correction for observer bias is likely in order.

[JulianeStruve1]

Thank you for your suggestion, the fuzzy tools are a promising tool !

I think that the concept of "edge" is rather hard to define, specially when you are analyzing its influence in the presence of certain species.  Instead of trying the hard-boundary "buffer" I would consider using fuzzy logic. 
Take a look at the fuzzy tools in spatial analyst, they can give you a few ideas.