Predictive ability of climate on vital rates and population dynamics depends on vital rate and model complexity

Abstract

AbstractUnderstanding mechanisms and predicting trends in vital rates and population growth responses to climate is a key goal of ecology. It is unclear whether predictive ability changes based on the response variable (e.g. survival, reproduction, population growth rate). Studies explicitly linking climate to vital rates and population dynamics remain limited, and use different statistical tools that vary in their complexity. Antecedent effect models are the most complex statistical tools, but are hypothesized to have better predictive ability because they capitalize on the evidence provided by climate and population data to select time windows correlated with a population level response. We compare the predictive performance of antecedent effect models against simpler models. We fit three antecedent effect models: (1) weighted mean models (WMM), which weigh the importance of monthly anomalies based on a Gaussian curve, (2) stochastic antecedent models (SAM), which weigh the importance of monthly anomalies using a Dirichlet process, and (3) regularized regressions using the Finnish Horseshoe prior (FHM), which estimate a separate effect size for each monthly anomaly. We compare these approaches to a linear model using a yearly climatic predictor and a null model with no predictors. We use demographic data from 77 natural populations of 3 plant species ranging between 7 and 36 years of length. We fit models to the overall asymptotic population growth rate (λ) and its underlying vital rates: survival, development (transition between stage classes), and reproduction. Contrary to our expectations, we found that simple models using yearly climate as a predictor showed, on average, slightly better predictive ability than antecedent effect models. These models had a higher ability to predict development than survival and reproduction and a higher ability to predict vital rates than population growth rate. We find that development is more predictable than other vital rates and especially population growth rates. Improving the predictive ability of survival and reproduction might be inherently more difficult, and might require larger datasets or conceptual advances. Conceptual advances could be facilitated by antecedent effect models which, despite not improving predictive ability, are well suited as exploratory tools.