Predicting the occurrence of Portugese man o' war

[DEHewitt]

Hi @doserjef,

I hope this is the appropriate place to ask a question like this - I am currently working on a project to predict the occurrence of Physalia physalis ('Portugese man o' war' or 'Bluebottle' here in Australia). I have several detection/non-detection and detection only datasets recording when they become stranded on the beach. In addition, I have access to records of when swimmers were stung in the water.

My initial thought is that beaching and stinging can be treated as two different ways to sample the same ecological process - bluebottle occurrence. Thus, these different data types could be integrated using {spOccupancy} in a single-species model, including different covariates on the detection process (i.e., for some days I have estimates of beach attendance -> more swimmers -> higher chance of stings).

Are you aware of any applications like this? Would it be more suitable to use a multi-species approach and treat them as different processes altogether (possibly estimating the residual 'species' correlations)?

Thanks for any advice you can provide and for developing such a cool and thoroughly documented package - I'm excited to get started using it!

[doserjef]

Hi @DEHewitt

Thanks for the note and kind words! Sounds like some very cool data that are working with.

That sounds to me like you would want to fit what's called an "integrated occupancy model", where you have two different data sources that sample the same ecological process (bluebottle occurrence), but they are subject to different sorts of sampling biases (i.e., you would want different covariates on the detection process for the two types of data). This sort of model can be fit with the intPGOcc() (non-spatial) and spIntPGOcc() functions in spOccupancy, so I'd encourage you to take a look at the documentation and vignettes related to those functions to see how that could be applied to your example. This sort of integrated occupancy model has been applied in a few different contexts (e.g., bottlenose dolphins and gamebirds).

Hope that helps! GitHub issues are more so for problems related to the actual functionality of the package so I'm going to go ahead and close this issue, but if you have any other questions as you get into your analysis then feel free to email me directly (doserjef@msu.edu).

Cheers,

Jeff