Drought legacy effects in radial tree growth are rarely significant under heightened statistical scrutiny

Drought legacy effects in radial tree growth have been extensively studied over the last decade and are found to critically influence carbon sequestration in woody biomass. Typically quantified as a deviation from “normal” growth, drought legacy magnitude and statistical significance depend on the definition of expected vs. unexpected growth variability under average conditions – a definition that has received insufficient theoretical validation.

Here, we revisit popular legacy effect analyses using the International Tree-Ring Data Bank (ITRDB) and employ a synthetic data simulation to disentangle four key variables influencing the magnitude of legacy effects. We show that legacy effects i) are mainly influenced by the overall auto-correlation of radial growth time series, ii) depend on climate-growth cross-correlation, iii) are directly proportional to the year-to-year variability of the growth time series, and iv) scale with the chosen extreme event threshold. Our analysis revealed that legacy effects are a direct outcome of the omnipresent biological memory.

We further found that the interpretation of legacy effects following individual drought events at specific sites is challenged by high stochasticity, and show that the commonly perceived stronger legacy effects for conifers are the result of higher auto-correlation compared to deciduous broadleaves. Given that the existing literature has not sufficiently addressed biological memory, we present two pathways to improve future assessment and interpretation of legacy effects. First, we provide a simulation algorithm to a posteriori account for auto-correlated residuals of the initial regression model between growth and climate, i.e. a corrected Null model to determine statistical significance, thereby retrospectively adjusting expectations for “normal” growth variability. The second pathway is to a priori include lagged climate parameters in the regression model. This substantially reduces the magnitude of observed legacy effects and thus challenges us to revisit estimates of drought-induced growth deviations by considering the full spectrum of expected growth behavior.