Input file for Poison log-normal mark-resight

[The other DHewitt]

We placed GPS collars on mule deer in western Texas and flew aerial surveys, recording marked and unmarked deer. We flew parallel transects over an area of about 2,000 ha, and therefore marked deer could be seen multiple times and resighting was done with replacement. Because all deer were collared with GPS collars, we knew which deer were available to be seen on any given occasion.

My question is how to handle deer that were available to be seen on one occasion, moved off the area for 1 or more occasions, then returned and were available to be seen again. Can the “..” convention be used in the capture history for such deer? Below is one of our input files and deer 151.230 is an example of a deer that temporarily left the study site. If writing their capture history in this manner is not appropriate, do we use the “-0” convention and therefore put a zero for the number of known marks?

I appreciate any guidance provided.

Dave


* Individual Marks 2 Groups */
/* 8 Occasions */
/* Begin Input File */

/*150.600*/ 01+00101+0+0+0+0 1 0;
/*150.850*/ 01+0+002+001+0+0 1 0;
/*150.895*/ 0101+0+0+0+0+0+0 1 0;
/*150.950*/ +0+0+0+0........ 1 0;
/*150.960*/ 0101+00201+002+0 1 0;
/*151.020*/ ......+00101+0+0 1 0;
/*151.060*/ +001+0+0+0+0+0+0 1 0;
/*151.080*/ 0101+0+0+0+00101 1 0;
/*151.180*/ +0+00101+003+0+0 1 0;
/*151.190*/ +001+00301+0+001 1 0;
/*151.230*/ +001+001....+0+0 1 0;
/*151.260*/ 0101+0+0+0+0+001 0 1;
/*151.280*/ 01+0+0+0+0+0+0+0 0 1;
/*151.310*/ 01+0+001+0+0+0+0 0 1;
/*151.350*/ ............01+0 0 1;
/*151.380*/ 010101+0+00101+0 0 1;
/*151.410*/ 0101010101010101 0 1;
/*151.440*/ +0+0+0+0+0+0+0+0 0 1;
/*151.530*/ 0101+001+0+0..+0 0 1;
/*151.540*/ 0201+00101+0+0+0 0 1;
/*151.560*/ 0101..........+0 0 1;
/*151.570*/ 0102020301+0+0+0 1 0;
/*151.600*/ 01+002020101+0+0 0 1;
/*151.610*/ 01+00101+001+0+0 0 1;
/*151.720*/ 010101010101..+0 0 1;
/*151.740*/ 01+0+0+0+0+0+0+0 1 0;
/*151.750*/ 0201+0+0+0+001+0 0 1;
/*151.780*/ 01+0+00101+0+0+0 0 1;
/*151.790*/ 0102+0+0+0+0+0+0 0 1;
/*151.896*/ 0201+0+001+0+0+0 0 1;
/*151.940*/ 0101+00101+0+0+0 0 1;
/*151.970*/ +001+002+0+0+0+0 0 1;
/*151.980*/ 0101010101+00101 0 1;

Unmarked Seen Group=1;
73 71 73 91 64 62 55 40;

Unmarked Seen Group=2;
195 143 133 152 101 71 95 48;

Marked Unidentified Group=1;
0 0 0 0 0 0 0 0;

Marked Unidentified Group=2;
0 0 0 0 0 0 0 0;

Known Marks Group=1;
12 12 12 13 11 11 12 12;

Known Marks Group=2;
19 19 18 18 18 18 17 20;

/* End Input File */

[dhewitt]

And here I've been thinking, ever since my first days at Virginia Tech, that it was ME that was the "other DHewitt".

[The other DHewitt]

Andrea Litt said she met you at a MARK workshop, so it wasn't a surprise when my first choice of a username for this forum was taken. I got the second best username!

[bmcclintock]

Hi Dave,

Thanks for posting to the forum. I think this depends on your study design and what you're trying to estimate. Are you trying to estimate a single abundance for each group over the entire study period (one N for each group across all 8 surveys), or estimate abundance for each group separately for all 8 occasions?

Based on the encounter histories (and the fact that the number of marks is known and temporary emigration is observed), I'm guessing it's the former and the input files are not formatted correctly -- because the encounter histories contain >1 occasion, they are currently in the format of the robust design (Z)PNE, which will estimate a separate N for the observable population on each occasion for each group, as well as apparent survival and transition probabilities between observable and unobservable states (on or off the study area in your case). To estimate a single N for each group across the duration of the study, the data need to be collapsed into a single occasion. For example, encounter history +001+001....+0+0 becomes 02, encounter history 0102020301+0+0+0 becomes 09, Unmarked Seen Group=1 73 71 73 91 64 62 55 40 becomes 529, Known Marks Group=2 19 19 18 18 18 18 17 20 becomes 20, and so forth. This approach ignores the fact that we know some individuals were unobservable on some surveys, but assuming the marked population is representative of the unmarked population in terms of detection probability and individual heterogeneity, then this should be ok. However, the estimate of N will be a "superpopulation" (i.e., the population using the study area over the period of interest) due to a lack of geographic closure; this may not be appropriate if density is ultimately what you're after. If sampling had been conducted without replacement (or the data could be subsampled to satisfy sampling without replacement), then the immigration-emigration LNE model in MARK could be used for this situation.

If you are indeed wishing to estimate N of the observable population for each group for each occasion, then using the ".." (hereafter dots) is doing exactly what we want in terms of abundance estimation (i.e., p=0 for this individual because it was unobservable for given occasion). In your case the number of marks is known and temporary emigration is observed (i.e., there's no reason to be estimating survival and temporary emigration using PNE), so the regular (i.e., no robust design) PNE should be used to estimate N for each group and occasion (i.e., 16 "groups" where each encounter history contains a single occasion). You could also "trick" MARK into doing this with your current input file by fixing phi=1 and gamma"=gamma'=0.

Cheers,
Brett

[The other DHewitt]

Hi Brett,

Because our objective in using MARK is to generate population estimates to compare to results of a sightability model, we are only interested in population size. Your description makes perfect sense and we will proceed with the second scenario you describe below. Thanks for the prompt and helpful reply.

Dave