the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Greater variability in environmental stress favours trees that prioritise storage of carbohydrate reserves over growth: a modelling analysis
Abstract. Trees use reserves of non-structural carbohydrates (NSC) to help them survive and recover from stress periods. However, accumulation of reserves is at the expense of growth, resulting in a growth-storage trade-off. Tree species may pursue different storage strategies to optimise fitness in environments with differing degrees of stress, but it is not clear which storage strategies provide a competitive advantage in which environments.
We use a forest gap model to explore competitive outcomes among idealised tree species with different combinations of two carbon storage-related traits: carbon utilisation rate (fast-slow spectrum) and switch time from growth to storage (risky-safe spectrum). We investigate the competitive success of alternative growth vs storage strategies in simplified environments which have a non-specific annual stress period. We vary stress intensity (the mean stress duration) and stress stochasticity (the variance of stress duration) to determine the effect of increased stress on composition outcomes.
Community composition shifted from growth-prioritising strategies to storage-prioritising strategies with increasing stress intensity and stochasticity. The major driver of this shift in community composition was increased mortality, due to depletion of carbon reserves, in species with growth-prioritising strategies.
Our results demonstrate that considering carbon storage strategies can provide new insights into tree survival and adaptation of tree communities to increasing stress caused by climate change.
- Preprint
(1732 KB) - Metadata XML
-
Supplement
(2711 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-1474', Anonymous Referee #1, 17 Jul 2026
-
AC1: 'Reply on RC1', Elisa Stefaniak, 04 Nov 2026
Dear Reviewer 1,
Many thanks for your insightful comments. Please find the original comments and responses to the comments below.
The experiment crosses only two mean stress intensities (54.75 and 91.25 days; t_crit=0.85 and 0.75 yr) with four levels of stochasticity. The Supplementary Information shows that these same two stress durations were also used in the single-plant sensitivity analyses from which the four allocation strategies were selected, but I could not find an ecological or empirical justification for choosing these particular values. With only two mean-stress levels, it is difficult to characterise a stress-intensity gradient or to assess whether responses are linear, nonlinear, or threshold-like. This also makes it difficult to evaluate how robust the conclusion is that the effect of mean stress intensity is inconsistent across stochasticity treatments (Lines 316–325). Would an additional sensitivity analysis across a broader gradient of mean stress duration help distinguish a genuinely weak or inconsistent intensity effect?
Thank you for this comment. The limited number of stress values used was a consequence of computational limitations, which remain at this stage, but we recognise that further explanation of the choices for this value would be beneficial. The original choice of stress durations was made as a reflection of a potential seasonally dry environment in which 2-3 months of drought with potential variation may be biologically meaningful. We present the intraannual dynamics of tree growth for single tree to present the different strategies but we recognise that expanding this may be beneficial.
To address this comment, we will extend the single tree analysis dynamics to produce the single species sensitivity analysis around a range of stress means and stochasticities and include this in the supplementary document as well as where relevant in the main text. We will also clarify the decision of environmental stress values in the main text of the manuscript.
The key mortality coefficients (𝛼_𝑑P1 and 𝛼_𝑑P2) in the productivity-dependent mortality function and 𝛼_𝑑S1 and 𝛼_𝑑S2 in the storage-dependent mortality function are listed as being derived from simulations (Supplementary Table 2) rather than constrained by empirical data. This is important because differential mortality is presented as the primary mechanism driving the competitive outcomes (Line 373) and storage-dependent mortality in particular determines mortality during the imposed stress period (Lines 343-344). The qualitative shift from Slow-Risky to Slow-Safe may therefore depend on the selected baseline and risk coefficients. I consider a sensitivity analysis across plausible ranges of these mortality parameters to be important for establishing the robustness of the central conclusion.
Thank you for your comment. The choice of mortality parameters was done by modelling the relationship between the mortality likelihood and levels of storage for different choices of parameters. This was not recorded in the manuscript and we recognise this was an oversight on our part. Moreover, we recognise that we have not done a sensitivity analysis which would indeed provide a lot of value to this manuscript and to interpret the results. To address this comment, we will recreate and document the original outputs of this analysis for both stress and productivity related mortality. We will also extend these outputs to conduct a single tree sensitivity analysis (as above) to ensure results are produced for all species and across different tree ages. Because of the stochastic nature of the model a full patch sensitivity analysis would at this stage be infeasible due to computational demands around model runs.
The definition of the productivity-dependent mortality driver X appears inconsistent between the main text and Supplementary Table S1. In the main text, X is defined as B/A_l and as net carbon uptake per unit leaf area (Lines 168–170). In Supplementary Table S1, however, the argument of the exponential appears as (dB/dt)/dA_l (page 2 under ‘Mortality- Productivity-dependent mortality rate’) while dB/dt is separately defined as net mass production (‘Storage allocation’ section). Could the authors clarify whether the intended driver is B/A_l, (dB/dt)/A_l, or a derivative ratio involving dA_l/dt? Based on the verbal definition, I suspect that the intended expression is (dB/dt)/A_l
Indeed the intended expression in the mortality function is (dB/dt)/A_l and this mistake in transcribing will be corrected.
A paragraph is duplicated in the Supplementary Information (p. 5). “From these simulations, two values for each parameter were chosen for a total of four contrasting allocation...” appears twice in immediate succession, remove the duplicate.
Thank you. The duplicate paragraph will be removed.
Again, many thanks for your comments and time.
Elisa Stefaniak
Citation: https://doi.org/10.5194/egusphere-2026-1474-AC1
-
AC1: 'Reply on RC1', Elisa Stefaniak, 04 Nov 2026
-
RC2: 'Comment on egusphere-2026-1474', Anonymous Referee #2, 06 Oct 2026
The authors present a novel extension of the plant gap-model framework by incorporating NSC storage dynamics and examining how growth-storage allocation strategies influence competitive outcomes under varying stress intensity and stochasticity. The modeling exercise addresses an important and timely question: how carbon storage strategies may influence tree survival and community composition under environmental stress. The integration of optimization theory concepts (switch time and storage utilization rate) into a trait-based gap model bridges physiological theory and community dynamics. The manuscript is generally well written.
The main conclusion is that increasing stress intensity and stochasticity favor storage-prioritizing strategies. However, the environmental stress representation is highly simplified in the modeling framework. For example, photosynthesis is reduced to zero during a generic annual stress period, independent of water relations, temperature, hydraulic status, phenology, or resource availability. In other words, the model essentially examines a carbon-starvation framework and does not consider other environmental stress scenarios commonly experienced by forests. For example, hydraulic failure mechanisms are absent despite extensive discussion of drought mortality, and the competitive outcomes might differ if hydraulic traits were considered along with NSC. As written, mortality is determined only by NSC storage levels which do not necessarily represent the more complex results observed in field studies. I do not think it is that simple and I am not sure we learn a lot by an overly simplified model.
The two-by-two trait level combinations (strategies) are fairly restrictive and don’t likely represent the actual strategies in most plants. I wonder if something else emerges from a more continuous gradient/range of traits or even just intermediate levels that are not restricted to the all or nothing levels.
The idea that stochasticity may be more important in determining the result of competitive interactions more than the duration of stress events is interesting, but the manuscript would benefit from a deeper discussion about why this may be the case (i.e., what is the mechanism).
The manuscript would benefit from more context regarding how realistic the model predictions of NSC concentrations, growth impacts, and competitive outcomes are with field or greenhouse based empirical studies.
The manuscript would also benefit form a brief analysis of parameter sensitivity.
Citation: https://doi.org/10.5194/egusphere-2026-1474-RC2 -
AC2: 'Reply on RC2', Elisa Stefaniak, 04 Nov 2026
Dear Reviewer 2,
Many thanks for your insightful comments. Please find the original comments and our responses below.
The main conclusion is that increasing stress intensity and stochasticity favor storage-prioritizing strategies. However, the environmental stress representation is highly simplified in the modeling framework. For example, photosynthesis is reduced to zero during a generic annual stress period, independent of water relations, temperature, hydraulic status, phenology, or resource availability. In other words, the model essentially examines a carbon-starvation framework and does not consider other environmental stress scenarios commonly experienced by forests. For example, hydraulic failure mechanisms are absent despite extensive discussion of drought mortality, and the competitive outcomes might differ if hydraulic traits were considered along with NSC. As written, mortality is determined only by NSC storage levels which do not necessarily represent the more complex results observed in field studies. I do not think it is that simple and I am not sure we learn a lot by an overly simplified model.
Indeed, the model we develop simplifies mechanics of stress and does not explicitly represent hydraulics and other dynamics. Our aim in this work was to specifically focus on the controls of NSC dynamics in a forest community and comparing different strategies in a competitive environment. To this end we believe our model was appropriate for the scope of this study.
All models are simplifications of reality; the choice of degree of simplification depends on the model purpose. Here we have very deliberately chosen to represent stress impacts in as simple a fashion as possible, because we are interested in examining the underlying dynamics of interactions between conservative and risky behaviours in competition.
More detailed models that include representations of plant hydraulics, soil physics etc are also valuable – but interpreting the behaviour of such models is considerably more difficult. A gradual increase in complexity of models, from simple to detailed, allows a better understanding of underlying mechanisms to emerge and provides a basis on which to disentangle dynamics in more complex exercises.
We think that the current degree of abstraction for this initial study is appropriate: it already demonstrates differences in competitive outcome between different strategies as the nature of stress changes, providing context for understanding “optimal” adaptation to stress. This work provides a solid baseline for development and interpretation of more complex models in future.
The two-by-two trait level combinations (strategies) are fairly restrictive and don’t likely represent the actual strategies in most plants. I wonder if something else emerges from a more continuous gradient/range of traits or even just intermediate levels that are not restricted to the all or nothing levels.
Thank you for your comment. While the two-by-two trait combination indeed won’t represent most strategies in real plants, we believe that the current trait combination, nevertheless, is appropriate for the scope of the study. Our reasoning is the following:
Firstly, we build on optimisation work which shows that, in simple models under deterministic conditions, varying the storage versus growth priority within a goal function results in behaviour that defaults to either maximising growth or maximising storage with clear boundaries between the two behaviours (Stefaniak et al 2024). Therefore, past work has already explored what happens when a more continuous gradient/range of traits is considered, as suggested by the reviewer, and we build upon this past work to inform our current choices. We acknowledge that a clearer reflection of this in the text would be appropriate and will expand the methodology to make sure this is clear in the trait selection section.
Secondly, the values selected in the final simulation runs were based on a sensitivity analysis described in the section Parameter Estimation for alpha_S and t_S. The final values chosen for these two parameters were close but not equal to the maximum in each level to correspond to the modelling study in Stefaniak et al 2024.
Overall, while the model formulation in our study is significantly more complex than the simple model from previous work, we argue that the selection of the specific trait combinations we employed here is supported by previous studies and our sensitivity analysis and our process allows appropriate comparisons with previous work as well as a tractable analysis of the results.
The idea that stochasticity may be more important in determining the result of competitive interactions more than the duration of stress events is interesting, but the manuscript would benefit from a deeper discussion about why this may be the case (i.e., what is the mechanism).
We are happy to extend the discussion to include more on this point. We would like to note that originally, we could not identify experimental studies exploring stochasticity and NSC and most of the research in this aspect still comes from theoretical modelling studies.
We would posit that the mechanism for the difference between the average and stochastic outcomes emerges from the fact that we are exploring traits related to resilience. As such we would assume (under the active carbon storage theory) that these traits have evolved to account for environmental stochasticity therefore providing a strong signal for increasing variance, while under a deterministic environment a single optimal value could be found.
The theoretical work within the field of ecosystem resilience under stochasticity and trait selection under stochasticity is still under active development but we think that your suggestion of a deeper discussion on this topic is a very valuable addition and we would like to pursue additional literature research to compliment our work in this context.
Moreover, we would believe that more could be elucidated with additional sensitivity analysis as suggested by yourself and Reviewer 1. As such we will extend the single tree analysis dynamics to produce the single tree sensitivity analysis around a range of stress means and variances for the four species and include this in the supplementary document.
The manuscript would benefit from more context regarding how realistic the model predictions of NSC concentrations, growth impacts, and competitive outcomes are with field or greenhouse based empirical studies.
We are happy to extend the discussion to include more discussion on this point given availability of studies. A strong limitation at the time of writing has been the number of studies (both field and glasshouse based studies) that explore NSC concentrations within competetive settings. At the time of writing we have not identified studies that would explore this but we are happy to repeat this search. Moreover we will include more discussion on what kind of study and data would be useful to test our model outputs.
The manuscript would also benefit form a brief analysis of parameter sensitivity.
A single-tree parameter sensitivity has been performed for the storage parameters but indeed, single-tree sensitivity may be extended to explore additional values such as stress levels and mortality parameters. We will extend the single tree analysis dynamics to produce the single tree sensitivity analysis for:
- A range of stress means and variances for the four species (suggested by reviewer 1 and emerging from a comment by reviewer 2)
- The mortality dependent parameters 𝛼_𝑑P1, 𝛼_𝑑P2, 𝛼_𝑑S1, 𝛼_𝑑S2 (suggestion by reviewer 1)
- Additional parameters introduced in this model formulation relating to storage and stress, i.e.: portion of respiration maintained through stress, rmin; Initial storage concentration following germination, beta_S1; and Coefficient of the rate of transition between growth and no-growth periods, beta_S2
Once again, many thanks for your input and time,
Elisa Stefaniak
-
AC2: 'Reply on RC2', Elisa Stefaniak, 04 Nov 2026
Data sets
Simulation results for "Greater variability in environmental stress favours trees that prioritise storage of carbohydrate reserves over growth: a modelling analysis" Stefaniak et al. https://doi.org/10.5281/zenodo.17552724
Model code and software
plant Stefaniak and Falster https://github.com/foxeswithdata/plant
storage plant model simulation analysis Stefaniak https://github.com/foxeswithdata/storage_plant_simulation_analysis
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 376 | 178 | 57 | 611 | 110 | 71 | 84 |
- HTML: 376
- PDF: 178
- XML: 57
- Total: 611
- Supplement: 110
- BibTeX: 71
- EndNote: 84
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript addresses the ecologically important trade-off between carbon allocation to growth and to non-structural carbohydrate storage. The authors extend the plant trait-, size- and patch-structured forest model with an active non-structural-carbohydrate (NSC) storage pool. Carbon-allocation strategy is represented by two traits: the timing of the switch from growth to storage and the rate at which stored carbon is utilised for growth. Four idealised strategies (Fast-Risky, Fast-Safe, Slow-Risky, and Slow-Safe) are allowed to compete in 100-year simulations across eight environmental treatments, combining two mean annual stress durations with four levels of stress stochasticity. The extended model also incorporates storage-dependent mortality, productivity-dependent mortality, and an allometric recalibration mechanism that allows plants to restore imbalances among tissues following stress. The main results show that seasonal stress favours slower carbon-utilisation strategies, whereas risk-taking strategies are more advantageous in the absence of stress because of their longer growing periods. Overall, the study provides a useful modelling perspective on how carbon storage strategies may mediate tree survival, competitive interactions, and forest community responses under climate change, particularly as droughts, heatwaves, and other extreme events become more variable and severe. My general comments are provided below.
Specific