Limits...
Dead or gone? Bayesian inference on mortality for the dispersing sex

View Article: PubMed Central - PubMed

ABSTRACT

Estimates of age‐specific mortality are regularly used in ecology, evolution, and conservation research. However, estimating mortality of the dispersing sex, in species where one sex undergoes natal dispersal, is difficult. This is because it is often unclear whether members of the dispersing sex that disappear from monitored areas have died or dispersed. Here, we develop an extension of a multievent model that imputes dispersal state (i.e., died or dispersed) for uncertain records of the dispersing sex as a latent state and estimates age‐specific mortality and dispersal parameters in a Bayesian hierarchical framework. To check the performance of our model, we first conduct a simulation study. We then apply our model to a long‐term data set of African lions. Using these data, we further study how well our model estimates mortality of the dispersing sex by incrementally reducing the level of uncertainty in the records of male lions. We achieve this by taking advantage of an expert's indication on the likely fate of each missing male (i.e., likely died or dispersed). We find that our model produces accurate mortality estimates for simulated data of varying sample sizes and proportions of uncertain male records. From the empirical study, we learned that our model provides similar mortality estimates for different levels of uncertainty in records. However, a sensitivity of the mortality estimates to varying uncertainty is, as can be expected, detectable. We conclude that our model provides a solution to the challenge of estimating mortality of the dispersing sex in species with data deficiency due to natal dispersal. Given the utility of sex‐specific mortality estimates in biological and conservation research, and the virtual ubiquity of sex‐biased dispersal, our model will be useful to a wide variety of applications.

No MeSH data available.


Age‐specific mortality estimates for male and female African lions of the Serengeti population. Polygons represent 95% credible intervals of age‐specific mortality rates. Mortality rates are plotted until the ages when 95% of a synthetic same‐sex cohort would be dead. The blue polygons represent male mortality rates obtained from the default model that accounts for natal dispersal. The green polygons represent male mortality rates obtained from an extended model, where secondary dispersal was accounted for additionally to natal dispersal by entering last detection ages of likely secondary dispersers as age of right‐censoring.
© Copyright Policy - creativeCommonsBy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4979716&req=5

ece32247-fig-0006: Age‐specific mortality estimates for male and female African lions of the Serengeti population. Polygons represent 95% credible intervals of age‐specific mortality rates. Mortality rates are plotted until the ages when 95% of a synthetic same‐sex cohort would be dead. The blue polygons represent male mortality rates obtained from the default model that accounts for natal dispersal. The green polygons represent male mortality rates obtained from an extended model, where secondary dispersal was accounted for additionally to natal dispersal by entering last detection ages of likely secondary dispersers as age of right‐censoring.

Mentions: Relaxing the assumption and accounting for higher‐order dispersal necessitates a customized extension of the mortality model we present here. The effectiveness of fitting this more complex model depends on the availability of information on both known deaths and dispersal events among immigrants. In the case of the Serengeti population, we took advantage of the expert's indication on likely dispersal state of disappearing immigrants and extended the default model (Model A) to treat all immigrants that were indicated to be likely dispersers as censored at last seen ages. The difference between the male mortality estimates from the default model and the extended model provides an indication of the amount by which male mortality is overestimated if secondary dispersal is not accounted for (Fig. 6). To improve mortality estimates, in future extensions of the model secondary dispersal can be imputed as a further latent state, similarly to what we have showcased here for natal dispersal.


Dead or gone? Bayesian inference on mortality for the dispersing sex
Age‐specific mortality estimates for male and female African lions of the Serengeti population. Polygons represent 95% credible intervals of age‐specific mortality rates. Mortality rates are plotted until the ages when 95% of a synthetic same‐sex cohort would be dead. The blue polygons represent male mortality rates obtained from the default model that accounts for natal dispersal. The green polygons represent male mortality rates obtained from an extended model, where secondary dispersal was accounted for additionally to natal dispersal by entering last detection ages of likely secondary dispersers as age of right‐censoring.
© Copyright Policy - creativeCommonsBy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4979716&req=5

ece32247-fig-0006: Age‐specific mortality estimates for male and female African lions of the Serengeti population. Polygons represent 95% credible intervals of age‐specific mortality rates. Mortality rates are plotted until the ages when 95% of a synthetic same‐sex cohort would be dead. The blue polygons represent male mortality rates obtained from the default model that accounts for natal dispersal. The green polygons represent male mortality rates obtained from an extended model, where secondary dispersal was accounted for additionally to natal dispersal by entering last detection ages of likely secondary dispersers as age of right‐censoring.
Mentions: Relaxing the assumption and accounting for higher‐order dispersal necessitates a customized extension of the mortality model we present here. The effectiveness of fitting this more complex model depends on the availability of information on both known deaths and dispersal events among immigrants. In the case of the Serengeti population, we took advantage of the expert's indication on likely dispersal state of disappearing immigrants and extended the default model (Model A) to treat all immigrants that were indicated to be likely dispersers as censored at last seen ages. The difference between the male mortality estimates from the default model and the extended model provides an indication of the amount by which male mortality is overestimated if secondary dispersal is not accounted for (Fig. 6). To improve mortality estimates, in future extensions of the model secondary dispersal can be imputed as a further latent state, similarly to what we have showcased here for natal dispersal.

View Article: PubMed Central - PubMed

ABSTRACT

Estimates of age‐specific mortality are regularly used in ecology, evolution, and conservation research. However, estimating mortality of the dispersing sex, in species where one sex undergoes natal dispersal, is difficult. This is because it is often unclear whether members of the dispersing sex that disappear from monitored areas have died or dispersed. Here, we develop an extension of a multievent model that imputes dispersal state (i.e., died or dispersed) for uncertain records of the dispersing sex as a latent state and estimates age‐specific mortality and dispersal parameters in a Bayesian hierarchical framework. To check the performance of our model, we first conduct a simulation study. We then apply our model to a long‐term data set of African lions. Using these data, we further study how well our model estimates mortality of the dispersing sex by incrementally reducing the level of uncertainty in the records of male lions. We achieve this by taking advantage of an expert's indication on the likely fate of each missing male (i.e., likely died or dispersed). We find that our model produces accurate mortality estimates for simulated data of varying sample sizes and proportions of uncertain male records. From the empirical study, we learned that our model provides similar mortality estimates for different levels of uncertainty in records. However, a sensitivity of the mortality estimates to varying uncertainty is, as can be expected, detectable. We conclude that our model provides a solution to the challenge of estimating mortality of the dispersing sex in species with data deficiency due to natal dispersal. Given the utility of sex‐specific mortality estimates in biological and conservation research, and the virtual ubiquity of sex‐biased dispersal, our model will be useful to a wide variety of applications.

No MeSH data available.