Very high UCL values - point count analysis of densities

[Vijay Ramesh]

Hello,

I am assuming this question has been asked before but I am having a hard time finding the answer to what I am struggling with. 

We carried out point counts across 43 locations. Each location was visited 6 times. I am trying to estimate species-specific densities for each location rather than for the entire region. The reason to do this is to compare the location specific density estimates of a species with acoustic detections of that species at that location. 

I have attached a screenshot of my dataframe and a screenshot of the density estimates produced. I am struggling with the very weird lcl and ucl values...making me think that I am not parameterizing something accurately. 

Looking forward to any help/suggestions,

Vijay

[Eric Rexstad]

Vijay

I agree that the precision of those point-specific estimates are suspicious. These estimates at the point level do not include variability in encounter rate between points, so if that were taken into account the uncertainty would be even greater.

Point-specific densities are not really the forte of these distance sampling methods. Do the point estimates seem sensible?

Can you send (off-list if you wish) the code and model object on which these estimates are based? Perhaps the detection function model is not as it should be.

---

From: distance...@googlegroups.com <distance...@googlegroups.com> on behalf of Vijay Ramesh <vr...@cornell.edu>
Sent: 25 February 2025 16:01
To: distance-sampling <distance...@googlegroups.com>
Subject: [distance-sampling] Very high UCL values - point count analysis of densities

 

--
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/e0934b02-d7d6-4229-b7fa-13185fc5293an%40googlegroups.com.

[Vijay Ramesh]

Hello,

A huge thanks to Eric for responding to my questions. I am appending the solution that we agreed on. I shared a reproducible example along with test data for Eric to take a look at.

From Eric: 

There are a couple of changes I have made to your data/code:

• rather than having each point as a stratum, I've reverted to having a single stratum sampled by 43 stations (change Region.Label)
• you noted you visited each station 6 times; hence Effort should be 6 rather than 1
• sometimes pairs of birds were detected rather than single individuals; ds() does not recognise the field "number" to indicate group size, so I change field name to "size" which is recognised by ds()

Fitting the hazard rate detection function model results in a single estimate for the stratum, rather than the station-specific estimates you desire.  The point-specific estimates are easily obtained recognising that station-specific detections divided by detection probability is the estimated number of birds at the station.

Those station-specific estimates are converted to densities as a result of dividing the estimated birds per station by the total area sampled (at the stratum level). Area sampled is

• area of a circle for a station: pi * truncation distance^2 (you used 50 for truncation)
• multiply this area by the number of visits (6)
• multiply again by the number of stations (43)
• pi*50^2*6*43 = 2,026,327 m^2 or 202.6327 ha

If you examine the component of the fitted model object trial$dht$individuals$bysample$Dhat - you will have those station-specific density estimates.

Thanks again Eric,

Vijay