A camera trap data study was developed, installing cameras for 21 days, but we completed all camera station points in a long period (24 months). Individuals of target species were identified.
This long period may include the inclusion or elimination of individuals of the study area, violating the assumption of closed population for capture-recapture methodology. There is another methodology to estimate abundance and density?
Thanks!
There is a nice review here:
Article Estimating abundance of unmarked animal populations: Account...
Maybe you have to use a open model, as Kurt said. It's probably in 24 months you can violate the assumption of closed population.
You can find a very useful course here:
http://www.mbr-pwrc.usgs.gov/workshops/scrFeb2014/
In fact every thing depends on how variable are your observation rates. If there are no big changes during the period of observations, then the population could be treated as demographically closed, which means that birth rate is appr. equal to mortality and emigration is appr. equal to immigration. Otherwise you may divide your study period to several periods whithin which your population could be treated as demographically closed basing on the observation rates
Clearly, you can't use closed CR models for the entire 24 months. The problem with using Jolly-Seber-type open models (for example with monthly sampling occasions) is that the population may not be open enough from one month to the next, so your estimates of phi (apparent survival) will be close to the boundary of 1 (and unreliable).
The other option is to use a robust design approach, either spatial (e.g. the Gardner et al. paper mentioned above) or non-spatial. However, to use the robust design, you will have to chop a lot of data so that a) you can define very short closed periods during which births/deaths (and in the non-spatial models, even movement) is unlikely, which are b) separated by long periods (during which data are discarded) when the population is open to gains and losses. While the spatial CR models are certainly better, the open model SCR versions have serious computational issues that may make them infeasible.
The other class of models for unmarked animals (e.g. Royle & Nichols 2003, Chandler & Royle partially marked, Royle N-mixture, etc.) mentioned above all require that you are repeatedly sampling the same population multiple times, to allow you to separate non-detection/capture from absence.
Finally, the closure assumtion does not state that gains (from births/immigration)=losses(from deaths/emigration), it states that there are NO gains or losses.
Hope this helps!
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India