Climatically controlled reproduction drives interannual growth variability in a temperate tree species

Abstract Climatically controlled allocation to reproduction is a key mechanism by which climate influences tree growth and may explain lagged correlations between climate and growth. We used continent‐wide datasets of tree‐ring chronologies and annual reproductive effort in Fagus sylvatica from 1901 to 2015 to characterise relationships between climate, reproduction and growth. Results highlight that variable allocation to reproduction is a key factor for growth in this species, and that high reproductive effort (‘mast years’) is associated with stem growth reduction. Additionally, high reproductive effort is associated with previous summer temperature, creating lagged climate effects on growth. Consequently, understanding growth variability in forest ecosystems requires the incorporation of reproduction, which can be highly variable. Our results suggest that future response of growth dynamics to climate change in this species will be strongly influenced by the response of reproduction.


Appendix S2
Intra-NUTS correlation of masting series for regions used in model development and fitting and model validation, representing synchrony of independent masting records from each region. Mean spearman rank correlation of all pairs of observations from the same region with >=7 years of overlap, with Hotelling correction for n<30 (see Vacchiano et al. 2017

Appendix S4
Justification of the original theoretical model. See also Figure 1.
In our model, growth is influenced directly by climate conditions in the growing season (temperature and precipitation), which are known to influence physiological processes including leaf phenology, photosynthesis, and xylogenesis processes (e.g. Breda et al. 2006;Leuschner et al. 2001;Muller et al. 2011). Numerous studies have reported lagged correlations between tree-ring growth and previous summer climate conditions (Babst et al. 2013;Hacket-Pain et al. 2016;Piovesan et al. 2005), and so the model also includes links to represent these lagged effects. Growth is also influenced by prior-year growth, as shown by the typically high first-order autocorrelation present in tree-ring widths. This biological 'memory' may reflect the importance of stored carbohydrates on current increment, especially at the start of the growing season (Richardson et al. 2013;Skomarkova et al. 2006). Additional climatic factors may be important locally (Dittmar et al. 2003;Drobyshev et al. 2010;Piovesan and Schirone 2000, Hartl-Meier et al. 2014, 2015, but are not consistently relevant across populations (Hacket-Pain et al. 2016), and so are not included in this analysis. Growth is also influenced by annual reproductive effort.
Our model also includes linkages between reproductive effort and summer temperature in the two previous summers (Vacchiano et al. 2017), as well as a link to previous year growth, as some studies have indicated that increased investment in reproductive effort maybe associated with higher growth in the previous year (Drobyshev et al. 2010). Previous summers' temperature is the main driver of variation in seed production in Fagus sylvatica (Ascoli et al. 2017;Drobyshev et al. 2010;Vacchiano et al. 2017), so no linkage between current year climate and reproduction was included in our model. Where regions failed a test, and MV outliers were identified, the tests were re-applied on datasets with outliers removed. Visual inspection of the data using MVN tools indicates that all datasets were suitable for further analysis, and that any non-normality is minor.
p-values of various tests in MVN package. Red = highly significant, orange = marginally significant. Years in bold are highlighted as potential outliers in multiple regions. Mediation analysis is used as a formal comparison of the performance of these three difference models for each of the five chronologies analysed.
In all five cases, the model C and B are statistically indistinguishable, and including the direct links between previous summers' temperature and RWI does not increase model performance.
Additionally, in SE2 and DE2-low, the model where RWI is modelled as a function of climate and previous years RWI (and not RE) performs significantly worse than models that do include RE. In DE2-high, this improvement is marginally insignificant.

Appendix S13
Multi-group model used for the prediction of RWI in 16 independent regions ( Figure 6). Raw coefficients are plotted, as used in the predictive model.