Standardising for Effort

[emilylester]

Hello,

I have a photographic database of whale sharks between 1992-2011 and have run both POPAN and MSORD models on this data. But I have noticed that search effort has doubled in the last years of my dataset and am concerned that this may be artificially increasing my abundance estimates.

I would like to either standardize my models for search effort or to add an effort covariate into my models but am having a bit of trouble working out how to do this, since I am relatively new to MARK.

I have number of hours spent searching for whale sharks every year and whale shark sightings per unit effort.

Any help would be greatly appreciated.

Emily

[cooch]

Hello,

I have a photographic database of whale sharks between 1992-2011 and have run both POPAN and MSORD models on this data. But I have noticed that search effort has doubled in the last years of my dataset and am concerned that this may be artificially increasing my abundance estimates.

I would like to either standardize my models for search effort or to add an effort covariate into my models but am having a bit of trouble working out how to do this, since I am relatively new to MARK.

I have number of hours spent searching for whale sharks every year and whale shark sightings per unit effort.

Any help would be greatly appreciated.

Emily

In broad terms, building a model that 'accounts for effort' is relatively straightforward. If your 'effort' covariate is continuous, then you'll need to use the design matrix (Chapter 6 of the MARK book -- specifically, section 6.8). If your effort measure is strictly binary (or, discrete in some fashion), then you could 'do it all' by modifying PIMs. Look at the dipper flood example (Chapter 6) for an example discussing this approach.

[jlaake]

From my experience, while effort can partially explain sighting/capture probabilities it is rarely the best model and you are typically better off using p(t) because time incorporates both the effort effect and any other variation. So if you are going to include an effort model, also include a time model for comparison.

--jeff

[emilylester]

Thanks for the quick replies!

I will have another look at Chapter 6, I think I get it now.

That is a great tip about the time model too.

Thank you!
Emily

[Bryan Hamilton]

I can understand how a time dependent model would account for effort but won't the abundance estimates still be inflated for time periods with more effort?

If you're interested in population trends, a time dependent model doesn't seem to solve the problem. It would seem more relevant with a omodel when survival or occupancy are the parameters of interest.

[darryl]

The number of animals counted would likely be higher in times with more effort, but in allowing capture/detection probabilities to vary, the estimated quantities of interest shouldn't be 'inflated'; the estimated number of animal's not sighted should be lower in times with more effort. CV's are also likely to be lower in times with more effort.

Doesn't matter if you're looking at a trend in abundance, survival or occupancy, it's a similar underlying idea.

I'm also assuming that the search area has been the same across all times and that in years of 'more effort' the search area wasn't larger than in other years. Otherwise the population at risk of capture could be quite different in different years.

[Bryan Hamilton]

the estimated number of animal's not sighted should be lower in times with more effort. CV's are also likely to be lower in times with more effort.

This makes sense. Thanks for clarifying. I'm still trying to really understand some of these concepts.