Hello all,
I am trying to perform an occupancy analysis on a camera trap dataset that did not consider occupancy modeling when designing the sampling protocol.
We are investigating how 5 meso-predator species differentially use road cells over forest cells and the variation of this space use over 3 continuous seasons/ "bioperiods".
I have data from May- September for 2 years. May- September has been split into the 3 seasons, or "bioperiods" based off of life history traits of a target prey species. A stratified random sampling scheme was used to select road and forest locations for a survey area broken into a 250^2 m grid. Randomly chosen locations were revisited in both years for both road and forest stratum. My bigger issue is that, road cells (and some forest cells in 2014) were also revisited within the SAME year in different bioperiods/seasons.
This results in a data set that looks something like this:
yr1.s1 yr1.s2 yr1.s3 yr2.s1 yr2.s2 yr2.s3
Site 1 roadcam1 roadcam2 roadcam3
Site 2 forest1 forest2 forest3
Site 3 forest4
Site 4 roadcam4 roadcam5
Site 5 roadcam6 roadcam7 roadcam8 roadcam9 roadcam10 roadcam11
and so on...
If I analyze all sites in a multi-season model, I have many missing observations. If I analyze sites in a single season model with bioperiod and year as a covariate, I break the assumption of independence.
I would like to consolidate my data for the two years since I have low capture rates of all species. I am considering splitting up the models by strata and analyzing road cameras in a multiseason model and off road cameras in a single season model with bioperiod as a covariate. I can then compare occupancy and detection estimates from the separate models. However, this method will force me to exclude some cameras and further reduce my captures.
I know that this forum is meant for specific software questions but I am struggling to come up with a better solution. I would very much appreciate any other ideas. Thank you!