Question about overlapping segments in DSM analysis

[marianne...@gmail.com]

Dear Distance Sampling Group,

I am currently using the dsm package to estimate the abundance of narwhals from aerial photographs, treating each photo as a segment in the dsm analysis. The photos overlap by approximately 30% with adjacent images, which results in some narwhal detections being duplicated across segments.

I would appreciate clarification on whether this overlap needs to be explicitly corrected for in the analysis, or if it is adequately accounted for by specifying segment.area in the model formula.

Thank you for your time,

Marianne

[Tiago Marques]

Hi Marianne,

Narwhals in photos... I seem to have heard about this before :)

As far as I can tell, the segment area will adequately account for the effort. So that is OK. But what it does NOT account for is for the lack of independence of the counts per segment.

One could argue that if the overlap was 1% this would be negligible. On the other hand, if the overlap was 99% you would probably introduce severe overfitting, as each detected narwhal would appear about 100 times in the data, with 100 times or so the same covariate values, with unpredictable consequences in terms of the precision of the estimates. Whether 30% overlap, where each narwhal could appear in about 2 to 4 segments, is something that could cause problems in practice, is a little bit of a grey area... so harder to tell.

One could possibly check by simulation what is the effect of this overlap on the outcome of results. But possibly the cleaner option would be to actually locate the animals along the transect line and use segments which are not the photos themselves. But maybe you will tell me why that's not possible from your data collection 

 perspective?

Hope this helps,

T

Enviado de Outlook para Android

---

From: distance...@googlegroups.com <distance...@googlegroups.com> on behalf of marianne...@gmail.com <marianne...@gmail.com>
Sent: Monday, October 6, 2025 7:16:52 PM
To: distance-sampling <distance...@googlegroups.com>
Subject: [distance-sampling] Question about overlapping segments in DSM analysis

 

--
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 visit https://groups.google.com/d/msgid/distance-sampling/69d95f22-031e-457b-a784-bf0a31ae27e1n%40googlegroups.com.

[marianne...@gmail.com]

Hi Tiago,

Thanks for the quick reply! This is very helpful.

In previous analyses, I cropped the photos to eliminate overlap and only used detections from the non-overlapping areas to avoid counting duplicates. I was hoping to simplify the data pre-processing this time by skipping that step, but as you suggested, it’s best to remove duplicates and use properly segmented data for the analysis.

This brings me to a second question: for some of the strata, I’m planning to use a strip transect approach. My plan was to calculate narwhal density across all photos in the transect—including the overlapping areas—and then multiply that density by the total area covered by the transect to get the total number of narwhals detected on each transect. Assuming the photos are read independently (i.e., the probability of detecting a narwhal in one photo doesn’t affect the probability of detecting narwhals in adjacent/overlapping photos), do you see any issues with this approach?

Thanks again for your help,
Marianne

[Tiago Marques]

Hi Marianne, 

Apart from the slight non independence issue that arises from you covering the same areas multiple times with multiple images I don't think there's any major problem in doing so. 

On the other hand if from multiple pictures in the same area, you actually get different numbers of animals that will tell you something about availability for detection.

Hope this helps, feel free to contact me off the list if you want to discuss further.

Cheers 

T

Enviado de Outlook para Android

---

From: distance...@googlegroups.com <distance...@googlegroups.com> on behalf of marianne...@gmail.com <marianne...@gmail.com>

Sent: Tuesday, October 7, 2025 4:10:44 PM
To: distance-sampling <distance...@googlegroups.com>
Subject: Re: [distance-sampling] Question about overlapping segments in DSM analysis

 

To view this discussion visit https://groups.google.com/d/msgid/distance-sampling/6f7a53d1-d94c-4117-be8c-913f07e53867n%40googlegroups.com.