Trouble with resistance values

[martinsc93v]

Good afternoon, greetings from Argentina. First of all I would like to thank you very much for the creation of this tool that I am using extensively in my doctoral thesis project.

I'been stuck the last few days in the analysis of the resistance surfaces obtained by using the "all.comb" function. I am analyzing 5 categorical rasters layers, taking advantage of this function to analyze the interactions between them. At the end of the "single surface" part of the analysis, I obtain that 3 of them have a resistance value of 1000 and two of 1 (as I can see in the surface obtained with .asc format).

Analyzing this recent work using the package ResistanceGA: https://link.springer.com/article/10.1007/s11252-023-01434-9

I find a table in which resistance values are specified for each variable, but these are not interactions but values for each landscape layer (Table 4).

Is there a way to obtain a resistance surface with values for all the landscape categories analyzed as obtained in this work? I am stuck with values of 1000 and 1 (depending on where the variable is present or not).

Thank you very much for any help

[martinsc93v]

This is the categorical results I obtained

[wpeterman]

I'm not exactly sure what you're describing. I think the results you're looking for will be in the combination of all five surfaces. Note, the default of ResistanceGA is to scale resistance surfaces to have a minimum of 1 and maximum of 1,000 (or whatever you specify). Because you only have two levels (habitat present = 1 and habitat absent = 0), the outputs and interpretations are kind of wonky.

[martinsc93v]

Dear Bill, thank you very much for your quick response. I admire your work very much. I have two more questions:

1- When using the function "all.comb", with 5 my surfaces, it does not perform the analysis of the joint surface of the 5 layers, but ends up in the analysis of all the combinations of 4 surfaces. Is there a way to make the surface that includes all of them?

2- When I finish the bootstrap analysis and the summaries of the model with the function "all.comb", I find that the best model to explain my data is the distance model (as expected in my case). Being that a model of 4 or 5 surfaces has a larger AICc than the distance model, does it make sense to consider the resistance values of that joint surface even though it is not the best model?

I hope I have explained myself well and thank you very much for your answer.

[wpeterman]

1- When using the function "all.comb", with 5 my surfaces, it does not perform the analysis of the joint surface of the 5 layers, but ends up in the analysis of all the combinations of 4 surfaces. Is there a way to make the surface that includes all of them?

• The default is to run all combinations up to 4 layers. Set max.combination = 5 in the all_comb function.

2- When I finish the bootstrap analysis and the summaries of the model with the function "all.comb", I find that the best model to explain my data is the distance model (as expected in my case). Being that a model of 4 or 5 surfaces has a larger AICc than the distance model, does it make sense to consider the resistance values of that joint surface even though it is not the best model?

• There isn't a clear cut answer here. It sounds like you might be analyzing a data set with relatively few sample locations/individuals, such that each additional parameter in the model is incurring a notable penalty. If the resistance values from a 4–5 surface combination make biological sense, I might make a more holistic assessment that includes marginal R² to determine if there is something more than distance to discuss. However, clearly temper interpretations of such surfaces since support is equivocal.

[jamesryan1896]

Hello,

I am curious about this becasue I have also recently updated ResistanceGA and have been optimising using Circuitscape in Julia, and since that occured my binary categorical RS have also been giving me resistance values of 1 and 1000 for my two feature classes (also presence/absence layers). Previously my optimised binary categroical surfaces had been giving me resistances of 1 (as always) and another value which was usually somewhere between 2 and 150 (for example), but each layer ended up with unique values. Is there any reason why now when I perform single surface optimisation that I always get resistances of 1 and 1000 for each of my 9 categorical layers? Is the value unimportant and the real take away is wether the feature is impeding or facilitating gene flow?

Thank you!

[wpeterman]

Thanks for bringing this to my attention. There were adjustments made to the rescaling of categorical resistance values a while back (4.2-4), but this was not the intended result. This issue should be corrected in the latest GitHub version.