Pantropical tree rings show small effects of drought on stem growth

Increasing drought pressure under anthropogenic climate change may jeopardize the potential of tropical forests to capture carbon in woody biomass and act as a long- term carbon dioxide sink. To evaluate this risk, we assessed drought impacts in 483 tree- ring chronologies from across the tropics and found an overall modest stem growth decline (2.5% with a 95% confi dence interval of 2.2 to 2.7%) during the 10% driest years since 1930. Stem growth declines exceeded 10% in 25% of cases and were larger at hotter and drier sites and for gymnosperms compared with angiosperms. Growth declines generally did not outlast drought years and were partially mitigated by growth stimulation in wet years. Thus, pantropical forest carbon sequestration through stem growth has hitherto shown drought resilience that may, however, diminish under future climate change.

Methods
We used tree-ring data from naturally regenerating tree populations in tropical and subtropical vegetation (30°N to 30°S; fig S1A-C). We compiled raw ring-width data from the ITRDB (International Tree-Ring Databank, https://www.ncdc.noaa.gov/data-access/paleoclimatology-data/datasets/tree-ring) (339 chronologies) and contributing authors (144 chronologies). Raw ring-width data for 98 out of the 339 chronologies available on the ITRDB had been uploaded as part of an earlier publication from the Tropical Tree-ring Network network (see www.tropicaltreeringnetwork.org). We selected chronologies that extended for at least 30 years beyond 1930 (our quality-based cut-off year for climate data) to obtain a minimum of three drought years (taking the 10% driest years). After selection, our dataset contained 483 chronologies, representing >10,000 trees, 163 species, 33 plant families and an average chronology length (since 1930) of 56 years.

To ensure that drought effects on radial tree growth are comparable across sites, we constructed single-species chronologies from raw ring-width series by applying a single detrending procedure. To assess effects of droughts occurring during a single year (i.e., excluding extended droughts), we chose to apply a flexible detrending of the raw ring-width series that removes low-frequency (decadal to centennial) variation and retains the high-frequency (annual) variation of interest. After testing spline detrending with various degrees of flexibility, we chose to use a spline with a 50% frequency cut-off at 20 years for the main analyses.

We evaluated drought effects on tree growth during extreme years and the following two years using Superposed Epoch Analysis (SEA). SEA is used in tree-ring analysis to quantify and test the significance of ring-width anomalies during rare events, including climate extremes and volcanic eruptions. We used the “sea()” function from the dplR package, which calculates mean RWI values during a specified range of years around events, generates 1000 bootstrapped distributions of RWI values from the chronology (with a sample size equal to the number of event years, and a minimum of 3 in this study), compares the mean of the ‘event RWI’ values with the randomly generated RWI distributions and returns a the probability that ‘event RWI’ values differ significantly from ‘normal’ years (p<0.05).