Pitfalls and windfalls of detecting demographic declines using population genetics in long-lived species

Abstract

Detecting recent demographic changes is a crucial component of species conservation and management, as many natural populations face declines due to anthropogenic habitat alteration and climate change. Genetic methods allow researchers to detect changes in effective population size (N_e) from sampling at a single timepoint. However, in species with long lifespans, there is a lag between the start of a decline in a population and the resulting decrease in genetic diversity. This lag slows the rate at which diversity is lost, and therefore makes it difficult to detect recent declines using genetic data. However, the genomes of old individuals can provide a window into the past, and can be compared to those of younger individuals, a contrast that may help reveal recent demographic declines. To test whether comparing the genomes of young and old individuals can help infer recent demographic bottlenecks, we use forward-time, individual-based simulations with varying mean individual lifespans and extents of generational overlap. We find that age information can be used to aid in the detection of demographic declines when the decline has been severe. When average lifespan is long, comparing young and old individuals from a single timepoint has greater power to detect a recent (within the last 50 years) bottleneck event than comparing individuals sampled at different points in time. Our results demonstrate how longevity and generational overlap can be both a hindrance and a boon to detecting recent demographic declines from population genomic data.

Keywords: age; conservation; demographic decline; effective population size; genetic diversity; simulations.

FIGURE 1

(a) In species with overlapping generations, older individuals born before the bottleneck will have genomes representative of a larger population size. Figure shows a population of tortoises that have undergone a population decline. At sampling points (open circles) before the decline, old and young turtles have similar levels of genetic diversity. During the decline, old individuals who were born before the decline have high levels of diversity compared to younger individuals, born after the decline from a limited set of possible parents. At some point after the decline, the population reaches a new equilibrium level of diversity, and old and young individuals have similar levels of genetic diversity. Dotted lines represent comparisons presented in this study: comparisons between adjacent timepoints (temporal) and comparisons between different age bins (age). (b) Schematics depicting life cycles in the annual and perennial models. In the annual model, each individual lives for one tick of the simulation and reproduction occurs once. In the perennial model, each individual lives for multiple ticks, with a probability of mortality (p) in each tick and has opportunities to reproduce in each tick.

FIGURE 2

Plots show ${\theta }_W$ and $\pi$ from all sampled individuals across time (ticks) for the perennial (average age, A = 2, 5, 10, and 20) and annual (A = 1) simulations with three bottleneck severities (R, bottleneck severity increases with R). Colored lines represent different simulation replicates and black lines are the mean values across replicates. The red dashed line indicates when the bottleneck occurred.

FIGURE 3

Plots show percent detection using temporal sampling for annual model (A = 1) and perennial models (A = 2, 5, 10, and 20) for three bottleneck severities (R, bottleneck severity increases with R). Percent detection is defined as the percent of replicates where we found a significant difference between individuals subsampled from the present and previous timepoints. The red dashed line indicates when the bottleneck event occurred. The gray‐shaded portions of the plot indicate when samples were taken every 50 ticks of the simulation, whereas the unshaded portion of the plot indicates when samples were taking every five ticks of the simulation.

FIGURE 4

Plots show percent detection in ${\theta }_W$ and $\pi$ between old and young individuals for perennial models (A = 2, 5, 10, and 20). The old age bin is made up of individuals whose age at a given timepoint fell above the 90th quantile. The young age bin is made up of the youngest individuals at a given timepoint. Sample size is the same between bins. Percent detection is defined as the percent of replicates where we found a significant difference between the old and young age bins. The red dashed line indicates when the bottleneck event occurred. The gray‐shaded portions of the plot indicate when samples were taken every 50 ticks of the simulation, where the unshaded portion of the plot indicates when samples were taking every five ticks of the simulation.

FIGURE 5

Plots show the difference in detection power between temporal and age bin sampling methods. Positive values indicate that age bin methods had higher percent detection and negative values indicate that temporal methods had higher percent detection. The red dashed line indicates when the bottleneck event occurred. The gray‐shaded portions of the plot indicate when samples were taken every 50 ticks of the simulation, where the unshaded portion of the plot indicates when samples were taking every five ticks of the simulation.