Population matrix models play an essential role in in modern ecology as they help project population abundance over different time intervals. A population can be modeled as a stage-structured or age-structured population depending on the most important predictor of demographic properties. In recent years, the use of stage-structured models has been favored for both animal and plant populations. This is partly because of convenience in interpreting model outputs. Although it is known that the conversion of age-structured vital rates into stage-structured vital rates can introduce some bias in population growth rate estimates, the sign and magnitude of the bias have not been investigated. Furthermore, different methods exist for converting age-structured vital rates into stage structured vital rates. The objective of this study was to compare the different approaches for converting age-structured models into stage-structured models. Our study involved converting the life table of common bottlenose dolphin into a stage-structured population model with three stages. We calculated the transition rates (proportion of survived individuals making transition into the next stage) using two methods that are commonly used in the literature: one matches the mean duration in each stage, and the other matches the proportion of individuals making the transition within each time unit. We also calculated fertility rate using two common methods. One assumed all births occurred in the mid-point in a time step, and the other assumed births occurred continuously over time. The combinations of two transition rate calculations and two fertility calculations resulted in four different stage-structured models. These models were compared with age-structured population models consisting of 30 age classes.
Our results suggest stage-structured population models grossly underestimate an annual population growth rate. The population growth rate was also very sensitive to fluctuation in fecundity over age. We also found that stable stage distribution was different under four stage-structured models. The results suggest it is important to consider these potential biases in selecting the type of population matrix.