closed population models - problem with N

[A.Wittich]

Hi everybody!
I am trying to estimate abundance of Rissos dolphins. I have to adimit I am new at MARK and learned how the programme works by reading the book. My goal was to use the robust design, but I thought it cannot hurt to know how to fit the closed models first as they are part of the robust design. I did the exercises from chapter 14 and everything worked out nicely. Now when I use my dataset (4 month with 8 sampling occasions) I run into problems. I started out with Model mt (and yes I know that N is not definable in that model) but by eliminating the interaction terms (encounter*time-like they did in the heterogeneity example) I do not get a plausible output, but apparently there are over 600,000.00 animals where it is reasonable that there are between 300-700. Also all the p terms in that model are all zero which is probably the underlying cause for the ridicules N. Now I do not know why that model runs into such problems (by the way by introducing heterogeneity it is the same picture for the full model bht when elimating the interaction terms of encountergrp and time). Other less parameterized models (Mb, Mt, etc.) work fine. I was just wondering why I run into this problem. I thought it might come from the encounter histories. I have a lot of animals (158) only encountered once from 312 encounter histories. So more than half were only encountered once. Could that be the problem? If yes what can I do about it? If I only use the encounter histories (seen at least 2) then I still run into the same problem. I appreciate all the help and suggestions. Thank you in advance and there are probably more postings coming when I arrive at the stage for the robust design:-)

Anja

P.S. I attached the output to this post



closed with out het

Real Function Parameters of {model bt no inter} DM}
95% Confidence Interval
Parameter Estimate Standard Error Lower Upper
------------------------- -------------- -------------- -------------- --------------
1:p 0.1095321E-003 0.0014352 0.7677755E-015 0.9999999
2:p 0.9351043E-004 0.0012254 0.6539159E-015 0.9999999
3:p 0.1026297E-003 0.0013450 0.7162447E-015 0.9999999
4:p 0.5374920E-004 0.7044229E-003 0.3749446E-015 0.9999999
5:p 0.1795567E-004 0.2353444E-003 0.1250639E-015 0.9999996
6:p 0.2989521E-004 0.3918486E-003 0.2079849E-015 0.9999998
7:p 0.4085541E-004 0.5354934E-003 0.2843658E-015 0.9999998
8:p 0.4559658E-004 0.5976136E-003 0.3176346E-015 0.9999998
9:c 0.4507208 0.0339383 0.3854580 0.5177245
10:c 0.4738501 0.0305379 0.4146515 0.5337931
11:c 0.3204853 0.0278042 0.2685907 0.3772351
12:c 0.1361078 0.0198829 0.1016127 0.1799672
13:c 0.2078067 0.0225937 0.1669781 0.2555559
14:c 0.2638905 0.0239284 0.2197230 0.3133706
15:c 0.2857641 0.0234834 0.2420344 0.3339139
16:N 629988.11 8253974.8 7672.5527 53860315.

[Eric Janney]

Anja,

Estimating animal population abundance using capture-recapture methods is notoriously tricky and fraught with difficulties. I think you should step back and think about the design of you capture-recapture study and decide if it is appropriate to estimate N. Two primary assumptions must be met in order for an estimate of N to be useful (i.e. not horribly biased like your N=600,000 estimate) These assumptions are 1) population closure and 2) all animals, marked and unmarked, have the same probability of being encountered. Given what little I know about Risso's dolphins from a quick Google search: 1) they live in temperate and tropical oceans worldwide, and 2) they travel in pods of 3-30, I think it is very unlikely that your study design met either of these two assumptions. The fact that over 1/2 of your marked animals were only encountered on one occasion is definitely a red flag that one or both of these assumptions were badly violated. In the future, you may want to explore distance sampling methods to get a density estimate. I know distance sampling methods (e.g. line-transect sampling) are frequently used to estimate Cetacean densities and would probably be a more robust approach than capture-recapture.