Distance sampling replicate

247 views
Skip to first unread message

arthur jacquemin

unread,
May 6, 2022, 3:58:17 AM5/6/22
to distance-sampling
Hello,

To briefly describe my situation: Several times a year we carried out counts by distance sampling to estimate the size of a population.
In order to have enough points we realised for each of these sessions 3 replicates. I entered the data on R following the format of this table:

Capture d’écran 2022-05-06 093052.png

For the data analysis I first fitted the best detection function on all the data (I will not go into the details of the choice of the function, my problem is not here):

hn.cos.df<- ds(dataframe, formula = ~1,key="hn", adjustment="cos", order = 0,
                 truncation=list(left=20, right=200))

To obtain a result per session and precised observers only counts on one side of the transect I then entered:

dht2(ddf = hn.cos.df, flatfile = dataframe, strat_formula = ~session , sample_fraction = 0.5,  stratification = 'replicate')

I obtain one value per session, but that value seems more like the sum of the 3 replicates, whereas I'm more looking to get an average value. I can divide the obtained value by 3, but I'm not sure if it's mathematically valid. 
What I need to change to get the result I'm looking for? 

Thanks you for your help,
Arthur

Eric Rexstad

unread,
May 6, 2022, 6:04:31 AM5/6/22
to arthur jacquemin, distance-sampling
Arthur

I hope I understand the estimates you want to produce, correct me if I am wrong.

If you want monthly estimates, derived from three surveys during each month, the simplest solution is merely to pool the data from the three within-month surveys.  That single estimate will be an average abundance over the combined survey period.  See Section 7.8.3 of Buckland et al. (2001).  Using this approach, you are not estimating abundance for each day, but rather for the period over which the three surveys occurred.

Effort is computed as the sum of the effort over the three days.  If the transects are the same for each day, then the number of spatial replicates is the number of transects surveyed (temporal replication does not add to the number of spatial replicates).

Accommodating the one sided survey effort is either performed using dht2​ (with the sample_fraction​ argument) or by doubling the point estimates, standard errors and confidence interval bounds.

Here is a simple example of a line transect survey where the observer walked the transect twice, notice how effort has been adjusted to account for the temporal replication.  For simplicity, I've combined data across species as well.

birds <- read.csv("https://synergy.st-andrews.ac.uk/ds-manda/files/2016/11/montrave-line.csv")
cu <- convert_units("meter", "kilometer", "hectare")
birds$Effort <- birds$Effort * 2
ans <- ds(birds,  convert_units = cu)
summary(ans)
adjusted <- dht2(ans, flatfile=birds, strat_formula = ~1, convert_units = cu, sample_fraction = 0.5)
print(adjusted, report="both")

No need to invoke the idea of stratification unless estimates are required at the level of strata, which would then lead to stratum-specific detection function modelling, which you have noted to be problematic because of number of detections per day.

From: distance...@googlegroups.com <distance...@googlegroups.com> on behalf of arthur jacquemin <ajacq....@gmail.com>
Sent: 06 May 2022 08:58
To: distance-sampling <distance...@googlegroups.com>
Subject: [distance-sampling] Distance sampling replicate
 
--
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/a07b87a2-d4d6-42c2-9cca-bc7552594da6n%40googlegroups.com.

arthur jacquemin

unread,
May 6, 2022, 8:35:13 AM5/6/22
to distance-sampling

Thank you Eric for this clear answer , I followed your advice and obtained the results I was looking for!

Eric Rexstad

unread,
May 6, 2022, 9:12:55 AM5/6/22
to arthur jacquemin, distance-sampling
Pleased to hear it Arthur.

Sent: 06 May 2022 13:35
To: distance-sampling <distance...@googlegroups.com>
Subject: Re: [distance-sampling] Distance sampling replicate
 
Reply all
Reply to author
Forward
0 new messages