Modeling reproductive trajectories of roe deer females: fixed or dynamic heterogeneity?

The relative role of dynamic and fixed heterogeneity in shaping the individual heterogeneity observed in most life-history traits remains difficult to quantify. In a recent work, Tuljapurkar et al. (2009) suggested modeling individual heterogeneity in lifetime reproductive success by a null model building reproductive trajectories from a first-order Markov chain. According to this model, among-individual differences in reproductive trajectories would be generated by the stochastic transitions among reproductive states (such as breeder and non-breeder) due to dynamic heterogeneity. In this work, we analyze the individual variation in three reproductive metrics (reproductive status, fecundity, and reproductive success) in two populations of roe deer intensively monitored using Tuljapurkar et al. (2009)’s dynamic model. Moreover, we challenge the Tuljapurkar model previously used as a biological null model to test whether the observed distribution of reproductive success over the lifetime was generated by a stochastic process by modifying two steps of the previous model to build a full stochastic model. We show that a distribution generated by the full dynamic model proposed by Tuljapurkar et al. (2009) can be consistently interpreted as only generated from a stochastic biological process provided that the probabilities of transition among reproductive states used are independent of the current reproductive state and that the positive covariation that usually occurs between survival and reproduction among individuals is removed. Only the reproductive status of roe deer females could be restricted to a stochastic process described by the full stochastic model, probably because most females (>90%) were breeders in a given year. The fecundity of roe deer females could not be adequately described by the full dynamic and full stochastic model, and the observed distribution of female reproductive success differed from the one generated by a full dynamic model in which each individual reproductive trajectory was independent of the individual lifespan (second step of the full dynamic model changed). While there was clear evidence that dynamic heterogeneity occurred and accounted for a large part of the observed among-individual variation in reproductive trajectories of roe deer females, a stochastic process did not provide a suitable model for all reproductive metrics. Consequently, models including additional fixed and dynamic traits need to be built in order to separate the relative role of fixed and dynamic heterogeneities in generating reproductive trajectories.