Adjusting the significance of daily climate responses in tree-ring proxies

Correlating tree-ring parameters with daily resolved climate data is becoming increasingly common for understanding the complex relationships between tree growth and the surrounding environment. However, with an increased number of calculated correlations, there is an inherent risk of spurious significance. In this study, we present an analysis using synthetic weather and tree-ring data mimicking the statistical properties of ten real-world sites across Europe to quantify the extent to which numerous comparisons may inflate maximum correlations. Comparisons of different tree-ring proxies, considering varying overlapping period lengths and seasons, revealed 95th percentile correlation differences reaching 0.25 by chance. Using synthetic tree-ring chronologies with an assigned non-signal (r = 0.00), spurious correlations can reach statistical significance in over 60% of tests. Correlation inflation is greater when: (1) the climate-proxy relationship is weaker; (2) comparison periods are shorter; and (3) the length of seasonal windows is longer. Autocorrelation in the proxy records does not appear to have a major effect. These findings indicate that caution should be exercised when computing high numbers of correlations with limited observations. We provide tables listing correlation inflations for precipitation- and temperature-sensitive tree-ring chronologies that can inform interpretations of significance.