INCLUDING TRANSITION PROBABILITIES IN NEST SURVIVAL ESTIMATION: A MAYFIELD MARKOV CHAIN

Mayfield estimates of daily survival are the most commonly available nest survival estimates in the ecological and ornithological literature. However, application of the Mayfield technique requires assumptions about the distributions of hatching and fledging events that are violated in most empirical data sets. We unified several variants of the Mayfield model using a discrete Markov chain formulation and used this formulation to study the effects of assumptions about transition probabilities under varying distributions of discovery probability. In general, bias in estimated survival was greatest when nests were most likely to be discovered late in the nesting cycle. Assuming that hatching and fledging events occur at the midpoint of an observation, or with uniform probability during an observation, also resulted in large biases. Estimates of standard error of daily survival and overall success were much more accurate and consistent. Knowing the nest age can greatly reduce bias but increases the invasiveness of monitoring and requires incorporation of transition probabilities in the Mayfield likelihood function. Based on our simulation results, we suggest that traditional Mayfield models are likely to provide adequate estimates for most applications if nests are monitored at intervals of no longer than three days.