Tests of multistate CJS models to estimate survival conditioned on a partially-observed latent state

Polar bears (Ursus maritimus) of the Southern Beaufort Sea (SBS) historically spent nearly the entire year on the sea ice but are increasingly using land habitat as sea ice becomes less available seasonally due to climate warming. The U.S. Geological Survey, Alaska Science Center maintains a research program on the SBS polar bear subpopulation and one important research question is whether habitat selection during the seasonal open-water period meaningfully influences fitness via annual survival rates. However, data collection for mark-recapture survival rate estimation takes place in the spring when all bears are on the sea ice and habitat selection during the preceding open-water period will only be known for a relatively small number of individuals. Consequently, as an exploratory exercise prior to working with the polar bear data, we evaluate the estimation of survival rates using multistate Cormack-Jolly-Seber mark-recapture models with survival dependent on a state matrix whose values are not all known (partially observed). We constructed a total of 36 combinations of (1) a model for use of land habitat in summer, (2) survival rates for bears in sea ice or land habitats, (3) capture probabilities, and (4) the proportion of habitat selection values that are known. The values of these parameters that were considered in our larger work are listed in the file "Table_Test_Cases.pdf". These 36 test cases can be viewed as two sets of 18 test cases which share an approach to modeling the proportion of individuals that choose land habitat in summer. Consequently, to avoid unnecessary replication, this repository contains R code files for one example from each set of 18 test cases, namely Model_01 and Model_19.