Development of demographic models to analyze populations with multi-year data—Using Agassiz’s Desert Tortoise (Gopherus agassizii) as a case study

Abstract

We developed a model for analyzing multi-year demographic data for long-lived animals and used data from a population of Agassiz’s desert tortoise ( Gopherus agassizii ) at the Desert Tortoise Research Natural Area in the western Mojave Desert of California as a case study. The study area was 7.77 square kilometers and included two locations: inside and outside the fenced boundary. The wildlife-permeable, protective fence was designed to prevent entry from vehicle users and sheep grazing. We collected mark-recapture data from 1,123 tortoises during seven annual surveys consisting of two censuses each over a 34-year period. Additional data were collected when marked tortoises were recovered dead and removed between survey years. We used a Bayesian modeling framework to develop a multistate Jolly-Seber model because of its ability to handle unobserved (latent) states and modified this model to incorporate the additional data from non-survey years. Three size-age states (juvenile, immature, adult), sex (female, male), two location states (inside and outside the fenced boundary), and three survival states (not-yet-entered, entered/alive, and dead/removed) were incorporated into the model. We calculated population densities and estimated probabilities of growth of the tortoises from one size-age state to a larger size-age state, survival after 1 year and 5 years, and detection. Our results show a declining population with low estimates for survival after 1 year and 5 years. The probability for tortoises to move from outside to inside the boundary fence was greater than for tortoises to move from inside the fence to outside. The probability for detecting tortoises differed by size-age state and was lowest for the smallest tortoises and highest for the adult tortoises. The framework for the model can be used to analyze other animal populations where vital rates are expected to vary depending on multiple individual states.