Photoperiod and temperature as dominant environmental drivers triggering secondary growth resumption in Northern Hemisphere conifers

Significance Forest trees can live for hundreds to thousands of years, and they play a critical role in mitigating global warming by fixing approximately 15% of anthropogenic CO2 emissions annually by wood formation. However, the environmental factors triggering wood formation onset in springtime and the cellular mechanisms underlying this onset remain poorly understood, since wood forms beneath the bark and is difficult to monitor. We report that the onset of wood formation in Northern Hemisphere conifers is driven primarily by photoperiod and mean annual temperature. Understanding the unique relationships between exogenous factors and wood formation could aid in predicting how forest ecosystems respond and adapt to climate warming, while improving the assessment of long-term and high-resolution observations of global biogeochemical cycles.


Supplementary Information Text
Methods S1 to S3 S1: Chilling and forcing Trees in temperate and cold ecosystems usually enter the dormancy state in later autumn or early winter after growth cessation and bud set in late summer or early autumn. The dormancy state involves the "rest" period (also called endodormancy phase) and the "quiescence" period (also called ecodormancy phase) (1). During the rest period, trees requires to be exposed to low chilling temperatures for a longer period (days to months) to break "rest" state. At the time of rest completion, the "quiescence" period is attained.
During quiescence, the buds attain full ontogenetic competence, which is the full ability to respond to high forcing temperatures by anatomic development towards bud burst or cambial cell division (2). The chilling requirement is thus normally defined as the length of the period (days or hours) during which temperature remains within a specific range (3). Compared to the most effective temperature range of 0~5°C usually calculated starting from November 1 st of the previous year to the onset day of spring phenology of primary meristems reported previously (3), we found the temperature range of -5~5°C is most effective for wood formation onset during our trial modeling analyses (SI Appendix   Table S6). We therefore summarized the days when daily temperature was within this range, starting from November 1 st of the previous year to the onset day of wood formation. The forcing requirement is normally defined as the length of period (days or hours) during which temperature remains above a specific threshold. The temperature threshold above 5°C for the period starting from January 1 st to the onset day of spring primary phenology was considered to be most effective and commonly used to calculate spring forcing requirement (2). We thus followed this threshold and the same reference period mentioned above for calculation of forcing requirement for wood formation onset.

S2: Trends fitted by generalized additive models (GAMs)
Generalized additive models (GAMs) were used to examine the general trend in mean annual temperature (MAT) and photoperiod of sites against latitude. Linear regression was performed to quantify the relationship between the onset date of wood formation and the photoperiod of the sites. 4 GAMs were also applied to examine the changes in the onset date of wood formation in relation to MAT and the scPDSI drought index, as well as the changes in the onset date of wood formation in relation to photoperiod, chilling, and forcing.

S3: Preliminary analysis with linear mixed models (LMM)
A linear mixed effects model (LMM), including species and site as random effects, was used to model the changes in the dates of onset of wood formation with photoperiod.
where Dijk is the date of onset of wood formation of species i at site j in year k; Pijk represents the photoperiod corresponding to Dijk; α is the intercept; β1 is the slope, and ai and bj, are, respectively, the random effects of the species i and site j; and εijk is the error term.
In addition, LMM was also used to model the changes in the dates of onset of wood formation with MAT instead of photoperiod.
where Tijk represents the MAT corresponding to Dijk, while the other parameters are the same as described above.
Texts S1 to S2 S1: General trend revealed by GAM The fitted generalized additive model (GAM) shows that the mean annual temperature (MAT) of the sites generally decreases toward higher latitudes (SI Appendix, Fig. S3), but the photoperiod of the sites increases toward higher latitudes (SI Appendix, Fig. S5). The 5 onset of wood formation was positively and linearly associated with photoperiod (R 2 =0.68, p<0.001) (SI Appendix, Fig. S5).
The relationships between the onset of wood formation and MAT and scPDSI show a downward surface, where the onset of wood formation is negatively related with MAT but shows a flat variation in the scPDSI (SI Appendix, Fig. S6). The relationships between the onset of wood formation and photoperiod vs. forcing and chilling also presented an "L" shape. The "L" shape reflects less variability in days of the year (DOY; ca. 140-160) when photoperiod continues changing at high latitudes (SI Appendix, Fig.   S7).

S2: Photoperiod or MAT alone as a significant factor affecting wood formation onset
Our LMM results show that the changes in the dates of wood formation onset in the Northern Hemisphere conifers can be modeled as a function of photoperiod alone, along with the random effects of site and species. The marginal R 2 and conditional R 2 are 0.46 and 0.98, respectively (SI Appendix, Table S2).      Dataset S1: Data that support the findings of this study are available in Data.csv.