Estimating stage-specific vital rates using a hidden Markov model for juvenile stages in a nymphalid butterfly

Background/Question/Methods

In butterflies, little is known about the dynamics of vital rates (survival and stage transitions) among juvenile stages in nature.  We aimed to better understand the progression of juvenile stages (egg, 5 instars, pupa) for a common butterfly of the southeastern U.S., Anaea aidea (Nymphalidae).   We ask if the probability of surviving and transitioning between stages depends on the current stage, and if these vital rates vary with time or density? We surveyed 510 caterpillars in their natural environment every 3 days for one month.  In the laboratory, 249 caterpillars were reared and observed daily from egg to death. We modeled the development of each individual in 3-day time steps as a Markov process defined by a stage-structured matrix model containing the probabilities of surviving and transitioning between stages. In nature, we were unable to observe each stage transition and the fate of individuals was often unknown resulting in a hidden process.  We developed this matrix model using data from the laboratory cohort with known fate, then modeled our field observations as a hidden Markov process, estimating the most likely survival and stage transition probabilities as well as detection probabilities for live and dead caterpillars given each individual encounter history. 

Results/Conclusions

Caterpillar densities increased over time to a maximum then declined as individuals either died or pupated.  The abundance of caterpillars attracted multiple predators, and survival declined as per capita predation rate increased over time. Transition rates increased over time, likely due to increased feeding and metabolism with warming spring temperatures.  Both survival and transition rates differed among stages.  Accounting for imperfect detection in the field and incorporating information from a laboratory cohort in a Bayesian framework enabled us to estimate vital rates for a natural population from data with missing observations.  Our stage-structured matrix model accurately represents the progression of individuals from egg to adult, which is the recruitment process that is absent in most butterfly demography.  These vital rate estimates parameterize the juvenile stages of a periodic stage-structured matrix projection model that characterizes the life cycle of this bivoltine butterfly in 3-day time steps over one year. The properties of this matrix reveal life history traits (e.g. mean life expectancy) and the sensitivity of vital rates through time for Anaea aidea.  Our next step will be to apply this model with limited data for a closely related endangered butterfly, A. floridalis, to better understand its life history and viability.