A Bayesian model for predicting local El Niño events using tree ring widths and cellulose δ18O

Abstract

The oxygen stable isotopic composition (δ18O) of cellulose recorded in annual tree rings reflects the climate and precipitation history experienced during tree growth and development. Here, we show proxy evidence of El Niño events over the past 30 years using juniper tree rings from southern California, United States. The relationship between tree ring δ18O in α cellulose and annual ring width was negative during most years, reflecting amount‐driven fractionation during precipitation. During El Niño years, the relationship between δ18O and ring width was positive with the largest ring widths correlated to the heaviest δ18O. Warmer sea surface temperatures during vapor formation and the strengthening of vapor transport from the eastern Pacific Ocean inland is the most likely mechanism driving heavier δ18O in precipitation during El Niño years. Based on this varying relationship between tree ring width and climate‐dependent δ18O values, we created a model to estimate the probability that a given annual tree ring was formed during an El Niño or non–El Niño year. The methods used in this analysis differ from standard dendrochronological technique because we explicitly account for the varying relationship between climate and tree ring characteristic during an El Niño or non–El Niño year. Moreover, our approach accommodates uncertainty in model parameters and predictions better than traditional classification methods. The application of this model to prehistory tree samples or samples of unknown age may allow for El Niño detection and subsequent determination of changes in El Niño frequency.