Combination of point count and line transect methods

[Colin Guilfoyle]

Hi all,

As part of my PhD I've carried out breeding bird surveys on blanket bog habitat. For a couple of reasons I ended up doing both line transect and point count surveys (each of our sites was surveyed with both point counts and line transects, with approximately the same effort). 

I've used Distance to calculate bird abundances at each site separately for point count and transect surveys and now I'm wondering is there any way to combine the abundance estimates from the two survey types into a single density value for each species (other than just taking an average value). 

Any ideas or papers related to this topic would be appreciated!

Thanks!

[Eric Rexstad]

Colin

I don't know of a method to combine the data from both data sources in using design-based inference. However, the topic is touched upon in a model-based context in a paper by Miller et al. (2021):

https://peerj.com/articles/12113/

Extending density surface models to include multiple and double-observer survey data peerj.com

Perhaps others on the list might offer some other ideas.

---

From: distance...@googlegroups.com <distance...@googlegroups.com> on behalf of Colin Guilfoyle <cguil...@gmail.com>
Sent: 02 November 2023 11:11
To: distance-sampling <distance...@googlegroups.com>
Subject: [distance-sampling] Combination of point count and line transect methods

 

--
You received this message because you are subscribed to the Google Groups "distance-sampling" group.
To unsubscribe from this group and stop receiving emails from it, send an email to distance-sampl...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/distance-sampling/2ee1e674-41f6-45df-aaff-30c9147d7f7en%40googlegroups.com.

[Mark Wilson]

Hi Colin,

I think that combining densities derived from point counts and transects should be quite straightforward. I think you simply need to weight the contribution of each estimate to the combined density according to the effective area sampled. This weighting should obviously take into account sample size (number of point counts and total length of transects) but the question of how number of point counts translates (in effort terms) to length of transect depends, I think, on detectability. For a species with very low detectability (e.g. assuming unrealistically that detection probability is 100% up to10m and 0% beyond this) then 10 point counts might only effectively sample an area of 10*10*10*pi = 3142m2, which would be equivalent to the area sampled by a transect length of 157m. On the other hand, if effective detection distance was 100m, the area sampled by 10 point counts would be 10*100*100*pi, which is 314159m2, and would be equivalent to 15.7km of transects! 

I think you can approximate effective detection distance from a given density by working out the total area that would contain the number of animals you encountered on your survey if they were at this density, and then working out what distance from your sampling points or lines would give you this area.

Cheers,

Mark

--

[Tiago Marques]

Hi Colin et al.,

I think Mark's idea is heuristically appealing but what might be the right weighting is not a simple question to answer. In a way, one might argue that the weights should be, if one can assume both points and lines to be unbiased, inversely proportional to the variance associated with each estimate. But other options are certainly possible, like the effective area covered... Not sure how to define what is optimal though.

I think ideally you would set up a hierarchical model where you observe the same reality, density, with two observation modes, lines and points. Then there are parameters associated with each detection mode but they both allow inferences over a common process. I can only assume this has been done before, but off the top of my head I can't find a reference for it. Maybe some of the hierarchical models in say Andy royle or Marc Kery papers/books already do this? Dave Miller's paper suggested by Eric might be a good starting point, it does kind of that in a way, by correcting each component of the survey, lines and points, for detectability first, then modelling density over space as a function of covariates.

Cheers,

Tiago

To view this discussion on the web visit https://groups.google.com/d/msgid/distance-sampling/CAAXwF%2BNRJYp0vbtxqoBmUdskE%3D2ovFbpjLf2p6R5fd1ep8kvqQ%40mail.gmail.com.

[Len Thomas]

I'd go for either a simple average or (probably better) an inverse
variance-weighted average. Cheers, Len

> --
> You received this message because you are subscribed to the Google
> Groups "distance-sampling" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to distance-sampl...@googlegroups.com

> <mailto:distance-sampl...@googlegroups.com>.

> To view this discussion on the web visit

> https://groups.google.com/d/msgid/distance-sampling/2ee1e674-41f6-45df-aaff-30c9147d7f7en%40googlegroups.com <https://groups.google.com/d/msgid/distance-sampling/2ee1e674-41f6-45df-aaff-30c9147d7f7en%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
I am a member of the University and College Union and am currently
participating in industrial action to improve UK higher education
staff pension, pay, equality and working conditions. For more
information, please see
https://www.ucu.org.uk/article/12469/FAQs#USS_pensions_dispute
and
https://www.ucu.org.uk/article/12469/FAQs#Pay_and_working_conditions_dispute

--
Len Thomas (he/him) len.t...@st-andrews.ac.uk lenthomas.org @len_thom
Centre for Research into Ecological and Environmental Modelling
and School of Mathematics and Statistics
The Observatory, University of St Andrews, Scotland KY16 9LZ
Office: UK+1334-461801 Admin: UK+1334-461842

While I may be sending this email outside of my normal office hours,
I have no expectation to receive a reply outside of yours.

The University of St Andrews is a charity
registered in Scotland, No SC013532.

[Colin Guilfoyle]

Hi all,

Thank you all for the suggestions - really appreciated. I have plenty to go off there.  A weighted average is something we had talked about internally as a potential option and based on the suggestions I'd say we are leaning towards that now. 

Thanks again,

Colin