1. Non-structural carbohydrates (NSCs) mediate plant survival when the plant’s carbon (C) balance is negative, suggesting that NSCs could predict plant survival under C stress. To examine this possibility, we exposed saplings of six temperate tree species to diverse levels of C stress created by the combination of two light conditions (full light availability and deep shade) and two defoliation levels (severe defoliation and non-defoliation). We then measured survival, biomass, and total NSCs and soluble sugar (SSs) concentrations in different organs of both dead and live saplings.

2. We estimated mean NSCs and SSs contents and concentrations per sapling and fitted logistic generalized mixed-effects models to determine if NSCs and SSs predict survival. Using inverse prediction modelling, we also determined whether there is a common NSCs and SS threshold across species at the time of sapling’s death.

3. Defoliation and shade reduced the mean sapling’s NSCs and SSs contents, indicating C stress. Mean sapling NSCs and SSs contents and concentrations predicted survival and the robustness of the models improved with the inclusion of species. At death, saplings of the exotic deciduous tree species Acer pseudoplatanus exhibited significantly lower mean NSCs and SSs contents than saplings of the evergreen conifer species Podocarpus nubigenus and lower stem NSCs and SSs concentrations than the broadleaf evergreen species Drimys winteri.

4. The energetic role that NSCs and SSs play in plants under C stress was evidenced by the capacity of these compounds to predict sapling survival under C stress. No common threshold of NSCs and SSs contents or concentrations for sapling survival amongst species was found, indicating that the level of these compounds may not be good proxies for interspecific comparisons of tolerance to C stress. Presumably, there are species-specific limits for the mobilization and use of NSCs and SSs in metabolism.

5. Our results anticipate that the inclusion of NSCs and SSs in modelling will improve predictions regarding tree responses to ongoing climate change. Nonetheless, a better understanding of the many roles that carbohydrates play in plant survival under C stress is required to scale predictions up to the community level.

We conducted an outdoor pot experiment over two growing seasons at the El Mallín nursery (Corporación Nacional Forestal, Conaf), located in Puerto Aysén, southern Chile (45° 59′′ S, 71° 52′′ W, 30 m above sea level [asl]). We studied two winter deciduous species (Acer pseudoplatanus and Nothofagus pumilio) and four evergreen species (N. betuloides, N. nitida, Podocarpus nubigenus and Drimys winteri) of contrasting successional status and light requirements. In October 2013 (early spring), two-year-old individuals of each species were transplanted to 2 L pots, except A. pseudoplatanus whichas was transplanted into 4 L pots. All species’ saplings were randomly assigned to four blocks with two light treatments (severe shade and open conditions). The open light condition was represented by full light availability, whereas the severe shade condition was achieved by placing the pots on the ground beneath the sparse shade of Alnus glutinosa trees, and by completely covering the lateral sides and the top with frames of a black Raschel mesh with 80% shading. In order to provoke further C limitation in saplings, by late spring (December 2014), a complete manual defoliation was applied to half of the saplings of each light condition. By late March 2015 (early autumn), all saplings were assessed for survival and subsequently harvested. Saplings were considered dead when they had shed all their leaves or when their leaves were all brown, in addition, the stem was dead (i.e. dry and brown). A total of 130 saplings, including 75 live and 55 dead, were harvested for biomass and non-structural carbohydrate (NSC) and soluble sugar (SS) determination. Organ-specific NSC concentrations were scaled up to the whole sapling using the weighted mean concentrations across all sapling organs, which consider the organ specific concentration (% dry matter) of NSCs and the organ-specific biomass. Additionally, organ-specific NSCs and SS contents (i.e. pool size, or mass) were calculated per sapling as the product of NSCs or SS concentration and the biomass of each organ. Organ-specific contents were then added to obtain whole-plant NSCs and SS contents. Finally, the NSCs and SS fraction represented by each organ was estimated for each sapling as the ratio between the organ-specific content and the whole-plant content. 

All statistical analyses were conducted in R version 4.2.0 (R Development Core Team, 2022). to determine if the light and defoliation treatments provoked C stress across species, we fitted linear mixed-effects models (LMMs) using the R/NLME package. we evaluated treatment and species effects on the NSCs and SS contents and concentrations of each organ and of the whole-sapling, as well as for NSCs and SS fractions. The fixed effects of the model were light condition, defoliation level, species, and all of their interactions. The random effect included the blocks. Similar models were used to compare biomass. Sapling survival was further analyzed using generalized linear mixed-effects models (GLMMs), with a binomial family and a logit link function because the sapling was considered either live or dead, with random effects for blocks and fixed effects for light condition, defoliation level, species, and the interaction amongst them. Similar to the previous analysis, we tested the significance of fixed effects using likelihood-ratio tests (LRTs). 

NSCs and SS contents and concentrations were compared between live and dead saplings using LMMs, where the block was the random factor and the survival, the species, and the interaction between them were the fixed factors.To determine whether the levels of C reserves predict sapling survival across species, we fitted mixed-effects binary logistic regression models, i.e. logistic GLMM using the R/LME4 package v1.1-27 (Bates et al. 2015), considering the NSCs or SS contents or concentrations for each organ or for the whole-sapling, with species as fixed factors, the block as a random factor, and the survival response as the binary response variable (Barker Plotkin et al. 2021). To determine if the means were significantly different from one another, we performed post-hoc comparisons with a Tukey's honest significance test, and a α of 0.05. For all models, we assessed model fit by examining standardized residuals using the R/DHARMA package (Hartig 2020). Finally, to determine if the NSCs and SS levels predicting survival under C stress were similar across species, we estimated the NSC contents and concentrations and their respective confidence intervals for 1%, 25%, 50%, 75% and 100% survival by inverse prediction, using the R/GGEFFECTS package (Lüdecke 2018). For NSCs, the model structure was as follows:

Survival ~ 1+species+NSCs+(1|block)

where Survival is the response variable expressed as either 1, live sapling, or 0, dead sapling; species and NSCs (or SSs) correspond to the fixed factors, whereas block is the random effect. These analyses were performed at both the organ- and whole-sapling level.