The increasing societal interest in ecological forecasts requires understanding the effect of climate on populations at large spatial scales. Comparative studies are valuable to this end because they provide an expectation of how populations will respond to climate without conducting onerous species-specific studies. The buffering hypothesis predicts that the demographic rates (henceforth “vital rates”) most important to population growth rate should be the least sensitive to climate variability. We tested this expectation using survival, growth, and fecundity data from 35 plant studies contained in the COMPADRE Plant Matrix Database. After we calculated the sensitivity of population growth rate to vital rates in each study, we estimated the climate sensitivity of demographic rates by comparing models that estimate the effect of precipitation or air temperature anomalies observed in the 36 months prior to demographic observations. We fit a null model that did not include climate, models using yearly climate anomalies, and eight alternative “moving window” models that weighted the relative importance of each monthly anomaly. We fit these models in a Bayesian framework, and compared them using a leave-one-year-out cross validation. Finally, we tested the buffering hypothesis by fitting a linear model relating vital rate sensitivity to climate sensitivity.
Results/Conclusions
Surprisingly, we found that over 2/3 of selected models did not include climate. When the selected model included climate, it often (90% of cases) employed a yearly climate anomaly as predictor. We found almost no support for “moving window” models that weighted the contribution of single monthly anomalies. Therefore, we calculated climate sensitivities using the best models including a climate effect. Analyses based on these climate models supported the buffering hypothesis: we found a significantly negative relationship between vital rate sensitivity and sensitivity to climate (precipitation or air temperature).
Our results, founded upon a conservative model selection procedure, suggest that climatic variability has weak effects on population dynamics. This is true even when using relatively sophisticated “moving window” models. It is unclear whether these results were influenced by a lack of temporal or spatial replication, or whether climate variability affects populations through mechanisms too complex for our empirical approach to discern. However, the support we found for the buffering hypothesis provides hope to attain generalizations regarding the effect of climate variability on plant populations. The poor predictive ability of our climate models and the bias of our data towards perennial herbaceous species indicate two promising avenues of future research.