Life-stage simulation analysis: Estimating vital-rate effects on population growth for conservation.

Abstract

We developed a simulation method, known as life-stage simulation analysis (LSA) to measure potential effects of uncertainty and variation in vital rates on population growth (λ) for purposes of species conservation planning. Under LSA, we specify plausible or hypothesized levels of uncertainty, variation, and covariation in vital rates for a given population. We use these data under resampling simulations to establish random combinations of vital rates for a large number of matrix replicates and finally summarize results from the matrix replicates to estimate potential effects of each vital rate on λ in a probability-based context. Estimates of potential effects are based on a variety of summary statistics, such as frequency of replicates having the same vital rate of highest elasticity, difference in elasticity values calculated under simulated conditions vs. elasticities calculated using mean invariant vital rates, percentage of replicates having positive population growth, and variation in λ explained by variation in each vital rate. To illustrate, we applied LSA to vital rates for two vertebrates: desert tortoise (Gopherus agassizii) and Greater Prairie Chicken (Tympanuchus cupido). Results for the prairie chicken indicated that a single vital rate consistently had greatest effect on population growth. Results for desert tortoise, however, suggested that a variety of life stages could have strong effects on population growth. Additional simulations for the Greater Prairie Chicken under a hypothetical conservation plan also demonstrated that a variety of vital rates could be manipulated to achieve desired population growth. To improve the reliability of inference, we recommend that potential effects of vital rates on λ be evaluated using a probability-based approach like LSA. LSA is an important complement to other methods that evaluate vital-rate effects on λ, including classical elasticity analysis, retrospective methods of variance decomposition, and simulation of the effects of environmental stochasticity.