Identifying a limiting factor in the population dynamics of a threatened amphibian: The influence of extended female maturation on operational sex ratio
Description
We investigated sex‐specific population dynamics in the threatened green and golden bell frog (Litoria aurea) using intensive capture‐recapture methods in a newly created wetland complex and control sites. As hypothesised, females took longer to reach maturity compared to males. The length of female maturation was 3.9 times greater than that of males (428.68 days ± 107.6 SD and 110.16 days ± 20.59 SD, respectively). This resulted in a one‐year delay in female population size increase compared to the male population. The operational sex ratio (OSR) in the second year of monitoring in the created wetlands had the most disproportionate male bias out of any year and any site (12/1 male/female, male proportion = 0.92 ± 0.89–0.94 95% CI). In the third year, the OSR had become less male biased (2.6/1, male proportion = 0.72 ± 65–0.78 95% CI), likely attributed to the maturing of the females produced in the first year breeding events. We did not find any evidence that survival or detection probability influenced the observed OSRs in the created wetlands. Based on survival rates of each sex, we estimate that males are 77 times more likely to reach sexual maturity compared to females. We postulate that the combination of chytrid‐induced disease and sex‐biased maturation rates may be a driver of declines, especially in populations with limited recruitment. We encourage future research into investigating sex‐specific population dynamics of amphibians, especially relating to reintroduction ecology. We have included the capture-recapture data of this study and the associated code embedded in R studio of the POPAN models used to estimate the super-population size of each year at every site.
Files
Steps to reproduce
We used capture-recapture methods via standardised visual encounter surveys of a threatened frog. Each frog was marked with a unique passive integrated transponder microchip. Each site was compared by ensuring that the number of repeat surveys and the survey area was kept relatively constant. POPAN models were used to estimate the superpopulation size (N), survival probability (φ) and detection probability (P). Sex was used as a covariate for N so that estimates would be produced for males and females so compare the operational sex ratio (OSR). An overall model to compare OSRs in the created wetlands was produced from surveys occurring across all wetlands in the site from September to April each year (n = 27, 27 and 26 for the respective seasons). Here the modelled averaged estimate of N was used to estimate the population size per season. However, to compare the OSRs between the created wetlands to control site wetlands, another set of models were produced. This comparison is important to determine if sex ratios obtained at the created wetlands were different in any season compared to the control sites. Due to the reduced time period and size of surveyed areas in the control wetlands, the population size of the created wetlands was modelled with CMR data from a reduced survey area within a reduced time period that matched the size and time periods of the control sites (reduced to a ~ 1 ha area and 10 surveys each season). This was done to ensure that sex ratios between the four sites were comparable. Each N estimate was calculated by using model averaging. Annual φ and P for males and females were produced by fixing these parameters with the group variable sex and using the most parsimonious covariate combinations in the other parameter. The full dataset of the created wetlands was used for this purpose. Over‐dispersion was assessed by using a goodness of fit test with the release.gof function in RMark. This allowed an estimation of the ĉ‐hat over‐dispersion parameter. If ĉ‐hat was |<|2, the models were deemed not over‐dispersed and no adjustments were made. If ĉ‐hat was >2, the models were deemed over‐dispersed and adjustments to the models were made using the ĉ‐hat value in the function adjust.chat. If over‐dispersion adjustments were made, models were selected based on quasi‐AIC (QAIC) which is an over‐dispersion adjusted AIC value.