Tree rings are often analyzed to reconstruct past climate or disturbances, but there is much to be learned about tree growth responses from tree rings. Traditional approaches to inferring climate from tree rings involve a series of steps to remove age effects, long-term trends that could mask the climate signal, and autocorrelation among ring widths. These detrended ring widths are typically averaged across trees to produce a stand-level chronology, which is used in follow-up analyses. Such dendrochronological methods remove the high- and low-frequency trends that are central to understanding the factors governing growth, and they do not propagate uncertainty across each detrending step. Our goal is to apply the stochastic antecedent modeling (SAM) framework to learn about the factors governing tree growth. SAM can be fit to raw ring widths within a Bayesian framework to evaluate the influence of climate on growth, while simultaneously accounting for age and autoregressive effects, important sources of uncertainty, and the hierarchical nature of the sampling (e.g., cores from trees from different species or sites). We apply SAM to tree-ring datasets to elucidate the importance of climate variables and the time-scales over which they influence growth, for multiple species across the Southwest and Spain.
Results/Conclusions
For multiple conifers in the Southwest and Spain, the SAM approach generally predicts reliance on less recent temperature compared to precipitation. Often, temperature conditions 1-4 years preceding ring formation are most important, while precipitation effects are dominated by current winter or growing season conditions. Furthermore, temperature and precipitation interact, suggesting antecedent temperature governs growth responses to recent precipitation. Spatial and temporal variation emerged in both time-scales of influence and effects of temperature, precipitation, or drought indices on tree growth, depicted by three examples. (1) Ponderosa pine in dry-winter sites (northern Utah) is more sensitive to precipitation, over shorter time-scales, compared to wet-winter sites (southern Arizona). (2) During dry (La Nina) periods, growth of three Southwest conifers is more sensitive to current drought conditions, whereas during wetter periods, growth is affected by drought conditions up to 4 years prior. (3) Compared to drought tolerant trees, trees defoliated by drought in Spain were more strongly coupled to, and more sensitive to, variation in more recent temperature and precipitation (over past 2 years). The SAM approach, which can simultaneously account for age and autoregressive effects, reveals that trees in multiple arid or semiarid systems respond to climate occurring multiple years before growth.