Comparing demography inferred from age vs. stage in a perennial plant.

Life history theories analyze and predict variation in vital rates, such as survival and reproduction, based on age. The age-from-stage method to derive age-specific vital rates from stage data was developed because age-specific data are rarely obtained for plants. Age-specific vital rates derived by this method might underestimate effects of age on vital rates, because the models assume that vital rates do not vary within stage classes. Consequently, population models and life history summaries relying on these vital rates could be biased against detecting senescence. Here, we perform a comparative study of methods to estimate age-specific vital rates using monitoring data with known age and stage. We derived age-, stage-, and age-and-stage-specific vital rates with demographic data from a long-lived perennial, Silene spaldingii. Then, we derived three age-specific population matrix models (age, age-from-stage, and age-and-stage). For each model, we derived life history summaries commonly used in ecology: population growth rate, net reproductive value, relative reproductive values, stable age distribution, generation time, and sensitivity and elasticity of population growth rate. Many vital rates depended on both age and stage in S. spaldingii. However, this species does not senesce; in fact, the number of flowers increased with age. As expected, the age-from-stage method was not able to accurately recreate the age dependence in some life history summaries, such as relative reproductive value. The age-from-stage model suggested faster reproductive dynamics in S. spaldingii than the models based on known age, i.e., plants started to reproduce earlier, and fertility remained constant thereafter, which may lead to biased predictions about evolutionary consequences of age-dependent life history traits. However, population growth rate, generation time, and net reproductive rate did not differ significantly among the models. Our study demonstrated that some metrics are robust to imprecision in model structure, while others are more sensitive. In spite of these biases, this case study provides another example of the diversity of aging patterns in plants. Age can be essential information when studying senescence in plants, but demographic metrics that were not about age per se were similar across model structures.