Penguin survival estimates from tag data using a multinomial likelihood

Abstract

A simple population model for Robben Island penguins is considered which incorporates fitting to both moult counts and tag data. The latter are now fit using a multinomial likelihood which is the method used in program MARK. Probability intervals on survival rates are more reliable than those obtained previously using an over-dispersed Poisson likelihood. Furthermore, the incorporation of a prior for relative undercounts of juveniles in the moult counts generally moves penalized likelihood estimates for adult survival rate away from an upper constraint boundary.