Add species column to occurrence data

[AugustT]

This will allow multi species models.

Requires changes to:

• [ ] Proccess/Model/Output modules - change 6:ncol(x) to dropping name columns
• [ ] Testing of Occurrence modules
• [ ] Insert a test after occurrence module to ensure strict column naming convention
• [ ] Also changes the module i/o vignettes

[goldingn]

Also changes the module i/o vignettes

[AugustT]

Contrary to my comment in #286 I think species should probably be a required column

[goldingn]

agreed, would be handy for plotting map legends, stacked SDMs etc

[AugustT]

These changes cannot be made until after the covCol change has been sent to CRAN. Making changes before then would cause the CRAN version to break.

[Rekyt]

Hey =) Thank you for building zoon! It seems to be a very flexible and a nice tool in the wild variety of SDM packages. I'm currently playing with it to build stacked SDMs and was wondering if this feature has been implemented or was planned to be implemented? Thanks!

[timcdlucas]

If by stacked you mean "run multiple species separately, then combine" it hasn't been implemented in any sense other than doing a list of occurrence modules.

There was discussions about more generally moving to multiple species (stacked, but also joint species distribution models). I think we decided to shelf it at least for now. It adds a load of complexity.

If you mean stacked more like stacking machine learning models (i.e. fit multiple models then model average them), that is on my to do list, but languishing near the bottom of it I'm afraid.

Sorry to not be more help!

[Rekyt]

Thanks for answering quickly. I meant "run multiple species separately then combine", I may develop a use case to show how to work with this, if it would be of interest for zoon. And it would be nice to get ensemble/model averaging techniques readily available, but I understand you have limited development power at the moment.

[timcdlucas]

"I may develop a use case to show how to work with this, if it would be of interest for zoon." It'd be great to see. Depending on how you would need it the basics is just something like

  w <- workflow(list(
                  occurrence = SpOcc('Dasypus novemcinctus',
                              extent = c(-130, -20, -60, 60)),
                  occurrence = SpOcc('Dasypus septemcinctus',
                              extent = c(-130, -20, -60, 60)),
                  occurrence = SpOcc('Dasypus hybridus',
                              extent = c(-130, -20, -60, 60)),
           covariate = Bioclim(extent = c(-130, -20, -60, 60),
                               layers = c(1:4, 6, 9, 10, 12, 15)),
           process =  Background(n = 10000), 
           model = MaxEnt,
           output = PrintMap)

stack(w$output)

I'm not exactly sure how the w$output would look, might need some list mangling.

We have discussed "metamodules" that combine modules created by lists. eg #307 I can imagine a SpeciesStack metamodule being useful. But this is not being worked on at the moment.

"And it would be nice to get ensemble/model averaging techniques readily available." This will happen. Eventually. I have grand plans.

Thanks again for your input. Knowing what people want/would use is really useful.