the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 05 Feb 2026
Tropical forest responses to climate extremes: an analysis using an individual-based demographic vegetation model
Abstract. Tropical forests play a crucial role in the global carbon and water cycles, yet their response to the climate extremes remains uncertain. Here, an individual-based demographic vegetation model is used to investigate the effects of warming and drought on ecosystem dynamics across three neotropical sites that span a precipitation gradient. By explicitly resolving plant hydraulic constraints and demographic processes, the study provides a mechanistic understanding of forest responses to climate stressors. The results reveal that warming had the strongest impact on carbon assimilation in the wettest sites (Paracou and Barro Colorado Island). This reduction was primarily driven by a rising vapor pressure deficit, which induced hydraulic failure even in the absence of soil moisture depletion. In contrast, the driest site (Tapajos National Forest) exhibited the highest sensitivity to drought, driven by severe soil moisture depletion. The analysis also shows that the timescale of imposed stress matters: short daily hot-dry events led to weaker impacts due to partial recovery between pulses, whereas yearly-scale warming and drought produced much stronger, persistent reductions in productivity. These findings highlight the site-specific vulnerabilities of tropical forests to climate extremes, where VPD-induced hydraulic stress limits carbon assimilation under warming in moist sites, while soil moisture constraints dominate in drier ecosystems.
- Preprint
(3567 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-145', Hisashi Sato, 02 Apr 2026
-
RC2: 'Comment on egusphere-2026-145', Anonymous Referee #2, 13 May 2026
This paper reports a modelling experiments that aims at disentangling the effects of warming to that of rainfall changes, that are occurring under climate change. It also incorporates different times scales at which warming and rainfall changes occur. I found accurate and novel an approach that investigates warming and drought patterns across different time scales, from daily events to multi-year climate oscillations. I also found sound using a process-based model that incorporates both physiology and demography (BiomeE), as it has been showed previously that both are important in tropical wet forests. The choice of well documented sites is sound as well.
However I found too many major shortcomings to this study. May be they are mainly a matter of presentation, or the work itself is questionable. From the current state of the article I cannot say, as too much information is lacking.
There is a massive lack of information on model parameters. All parameters used need to be clearly stated, along with their sources. From what I could gather, I suspect that maximum leaf area index (LAI), and soil water holding capacity were unrealistic. Too low for LAI, and too high for soil water. Both are paramount for correctly simulating the water cycle.
Some modelling choices were too simplistic. Using one plant functional type seems oversimplifying, given the predicted shifts in plant functional types under climate change in other studies (e.g. Levine et al. 2016 doi: 10.1073/pnas.1511344112; Longo et al. 2018 doi: 10.1111/nph.15185). Plus it seems contradictory with a model that is designed to handle multiple plant types. Another, probably more concerning choice, was to maintain relative humidity equal to observed climate under all scenarios. Variations in rainfall and temperature impact on relative humidity, which impacts on water vapor pressure deficit, which is one of the main drivers of forest response in this study.
The paper does not present any convincing model test, nor does it refer to previous tests. The only one, that is presented in fig A1, is against gross primary production obtained from eddy covariance measurements, and predictions do not match observations over multiple years. This does not lend confidence in the model’s ability to handle forest response to inter-annual climate variability. Eddy covariance data also include net ecosystem exchange, and evapotranspiration, which could also have been used for model testing. Why not? The chosen sites are well documented and present other datasets that could be used for further testing. Why not?
Then I found the simulation design was not well presented. It should state all scenarios, the number of simulations. Particularly I found the setup of the warming scenarios confusing. It apparently mixes periods of warming and cooling, without being very explicit about the latter. Obviously, introducing cooling severely hampers the realism of the predictions. Also, it is important to be able to compare those warming and drought treatments to the standard future climate projections under varying emission scenarios. For instance, the figures don’t show us if warming is close to 2, 4, or more °C.
The result section only focuses on forest scale fluxes such as GPP and transpiration. It does not show any prediction of some key variables that influence the forest response to warming and drought, such as LAI, or plant water potentials. This prevents a thorough understanding of the underlying mechanisms controlling the flux patterns. Has there been mortality, has there been shifts in LAI? Without going into further details, the analysis remains superficial.
From what is presented in the current paper, I don’t know if the outcomes would have been different with a big leaf model using prescribed LAI, than with a more refined model such as BiomeE with more physiology and demographics.
Detailed comments :
line 43: please explain what ENSO means.
Lines 120-125: spin up runs need more details, and ideally should be in a separate section than the model description. Were they followed by warm up runs?
Line 123: the use of the term “parameter” can lead to confusion, here “climate forcing” seems more appropriate.
Line 127: using only one PFT seems contradictory with the model choice. What is the justification?
Line 127-129: please provide all details of parameters. As it is, how the reader knows what value comes from what study?
Line 131: is it realistic to use the same hydraulic parameters for all sites? It seems contradictory for three sites covering a rainfall gradient. What data is there to back up this choice? A separate section on model parameter values and sources is needed.
Lines 132-135: Exactly how these parameters were “tuned” to match observational data?
Lines 137-139: previous studies, though, have emphasized the importance of shifts in PFT composition under climate shifts.
Line 146-148: source of dataset?
Lines 148-151: source(s) of data? Max LAI seems low for a tropical wet forest.
Lines 154-158: source(s) of data? Max LAI is low compared to measured LAI at Guyaflux (Bonal et al. 2008 report values well above 6).
Lines 163-166: source(s) of data? Max LAI seems low.
Lines 167-172: what is the point of providing a test, if it then discarded as not fully relevant? I would argue that it is relevant, not sufficient, and rather strongly suggests that the model is not suitable to “examine the relative ecosystem responses to imposed climate stressors”.
Fig. 1: figure seems identical to a previous publication, please ensure proper reference. There is no mention of senescence/turnover of plant tissues in this figure, nor in the text. Surely mortality is not the only way tree C and N are transferred to the soil organic matter? Also full names of all acronyms and symbols should be provided in the legend. Not clear what the spatial representation is. I assume it is 1D (only vertical layers) but the figure suggests otherwise, please be more precise.
Line 175: in this section about the climate scenarios, it is not obvious that temperature and rainfall treatments are mixed and how. It should state how many scenarios there are, and also how they compare with current climate model prediction scenarios.
Line 200: I consider using a constant relative humidity an issue in this study, and in contradiction with the level of details in the changes in temperature and precipitations in the different climate scenarios.
Lines 208-210: I do not figure out how drought can be lengthened at the daily scale. Is it only by reducing the amount of rain on rainy days?
Lines 243-261: Water stress arises when the water resource does not meet the water demand. Trees can to a certain extent adapt their water demand by modifying their leaf area. Mortality can also reduce the water demand. Hence rainfall, soil water content, and transpiration are not direct indicators of tree water stress. I find strange and sorely lacking that biological water stress indicators, such as plant water potentials, which the model computes, are not provided in the results.
The differences between sites largely depend on model parameters, which were not provided: stomatal conductance, soil water holding capacity, etc. An isohydric or anysohydric behavior is defined by plant hydraulic parameters. Therefore the sentence “suggesting a more anisohydric strategy” (l 252), is strange, especially since it is stated that the same hydraulic parameters were used at all sites (l 131).
Fig 2: Soil water and transpiration should have the same unit (e.g. mm) for better readability. Why displaying GPP in one case and NPP in another? What is the unit of WUE: g C of GPP or of NPP? Please make units more consistent. Soil water contents are very high, especially if they are multi-decadal averages. Can those soils actually hold that much water? Bonal et al. 2008 report max soil humidity of 840mm at Guyaflux.
Fig 3: I do not understand what “temperature range” means. Surely we are not talking about ambient temperatures increased 7 fold! But then, what T/Ta means? How do we compare those predictions to future climate changes of +2, +4, or +6°C? A more tractable index (e.g. increase in °C) would be easier to interpret.
Fig 4: what is the maximum VPD? Daily, yearly, over the whole simulation?
Lines 284-285: details on simulation design should all be presented in the methods section.
Fig 5: this figure shows simulation results at higher VPD than in fig 4 for BCI and GYF, why? I do not understand what the cycles mean, hence I do not understand the right panel at all. What was the rationale for selecting the simulations to be displayed in the right panel?
Fig 6: same comments as fig 5.
Lines 313-318: what mechanisms, within the modelling framework, explain those patterns?
Line 325: what means “still retains more water”?
Line 327: what about variations in LAI? or mortality?
Lines 322-344: predictions of hydraulic failure occurrence, LAI variations, and mortality, are necessary to discuss the differences between sites.
Lines 334: in simulations in which only rainfall was manipulated, then of course, soil moisture – fed by rainfall- will be the limiting factor...
Lines 334-335: plants hydraulic strategies and community structure actually define how soil water decreases.
Line 341: again, what about LAI variations and mortality?
Line 344: in the figures, I do not know where the “4◦ C above ambient” is.
Lines 355: a figure to illustrate “partial hydraulic and carbon recovery” would be useful.
Lines 394-395: so LAI was fixed? Actually reduction in LAI also reduces plant water use, relieves water stress, and represents a major management mean to reduce water stress, so this statement about additional water loss when LAI decreases is more than questionable.
Lines 399-405: this positioning is hardly justifiable, given than those variables are key to understand the results presented in this paper.
Lines 407-417: a lot of factors can have an influence in the model response to climate change (fire, pest outbreak, species composition shifts, etc). One that is known to directly influence plant physiology, and thus possibly more relevant here, is air CO2, and the ongoing debate on the compensating effect of increasing CO2 in a warmer and drier climate. What about that?
Line 437: Please be more accurate when citing sources, e.g. Bonal et al. 2008 does not provide data for the 2004-2014 period!
Lines 438-443: I do not agree, see comment for Lines 167-172 above. Model testing also need statistics (R2, RMSE, bias…) to be interpreted.
Lines 495-499: If I understand correctly (although I do not understand the concept of the replicates described here), simulations for GYF, TNF, and BCI covered 900, 900, and 540 years, respectively. From those years, the 450-500, 450-500, and 270-300 years were used for presenting the results, respectively. So my question is: what was the point of simulating the remaining, 400, 400, and 240 years?
Fig C2: From the text I found hard to understand what the warming scenarios actually implied. From this figure it seems that warming is mixed with cooling. I beg to understand what the cooling is doing in a study on drought and warming, and I can only think it compromises the realism of the predictions.
Citation: https://doi.org/10.5194/egusphere-2026-145-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,256 | 417 | 103 | 1,776 | 288 | 244 |
- HTML: 1,256
- PDF: 417
- XML: 103
- Total: 1,776
- BibTeX: 288
- EndNote: 244
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Dear authors,
This study investigates how tropical forests respond to drought, warming, and their compound effects across multiple timescales using a demographic vegetation model (BiomeE). By applying controlled perturbations to precipitation and temperature, the authors aim to disentangle the relative roles of soil water supply and atmospheric demand in shaping ecosystem stability. The results highlight the importance of vapor pressure deficit (VPD), baseline hydroclimate, and stress persistence in determining forest resilience and collapse thresholds across contrasting tropical sites.
I find that the topic is relevant and timely, and the modeling framework provides useful mechanistic insights into the interactions between drought and warming in tropical ecosystems. The manuscript is well-structured and addresses an important question in the context of future climate extremes. However, I have several concerns about the consistency between the figures and their interpretations, as well as the clarity of some key definitions and the conceptual framing. These issues should be addressed to improve the robustness and readability of the manuscript.
_____ Major Issue _____
(1) Figure 2
I find this figure somewhat difficult to interpret due to inconsistent axes and response variables across panels (a-c). While each panel individually provides useful information, their combined presentation obscures the underlying logic and makes cross-panel comparison challenging.
Specifically, panel (a) uses rainfall as the x-axis and GPP as the response, panel (b) uses soil water content and WUE, and panel (c) uses transpiration and NPP. These shifts in both explanatory and response variables make it difficult for the reader to follow the causal chain from water availability to ecosystem response.
I would suggest the authors consider the following improvements:
A. Clarify the conceptual role of each panel in the main text, explicitly stating that panels (a-c) represent different stages of the water-carbon pathway (e.g., external forcing --> internal state --> flux response).
B. Unify response variables where possible (e.g., use GPP consistently across panels, or explicitly justify the use of NPP in panel (c)).
C. Consider reorganizing the figure, for example:
Using a consistent x-axis (e.g., soil moisture or rainfall) across panels, or
Splitting the figure into multiple figures with clearer thematic grouping (e.g., thresholds vs. water-use strategies vs. flux relationships).
D. In panel (c), the linear relationship appears to mask important nonlinear behavior (as noted for TNF). It would be helpful to explicitly highlight this in the figure (e.g., by showing the raw points more clearly, adding a breakpoint, or avoiding a simple linear fit).
Overall, improving the structural consistency of Figure 2 would significantly enhance readability and clarify the key model behavior.
(2) Figure 3
The definition of the x-axis variable "T/Ta" in Figure 3 is unclear. The figure caption does not explicitly define this quantity, and it appears that different panels represent different temperature metrics (e.g., mean temperature vs. temperature range at various timescales).
This lack of clarity makes it difficult to interpret the magnitude of warming and to compare across panels. I suggest that the authors:
A. Explicitly define "T/Ta" in the figure caption, including whether it refers to mean temperature, temperature range, or another metric.
B. Use distinct labels for different temperature metrics if they are not directly comparable (e.g., mean temperature vs. diurnal or interannual range).
(3) Figure 5 & 6
The patterns shown in Figures 5 and 6 appear to be inconsistent with the interpretation provided in the text, particularly for GYF.
In Figure 5 (yearly-scale perturbations), GPP declines relatively gradually with increasing VPD, whereas in Figure 6 (daily-scale perturbations), the decline is more abrupt. This seems counter to the general expectation stated in the manuscript that longer-term stress should lead to stronger or more abrupt ecosystem responses.
Would the authors clarify this apparent discrepancy? Specifically:
A. Is the difference due to the way VPD or GPP is normalized or aggregated across timescales? If so, the normalization procedure should be clearly defined and justified, as it may affect the apparent sensitivity across figures.
B. Are the VPD ranges directly comparable between Figures 5 and 6?
(4) Conceptual inconsistency between soil moisture limitation and VPD dominance
The manuscript presents soil moisture limitation (Section 4.1) and atmospheric demand (VPD; Sections 4.2-4.3) as primary drivers of ecosystem instability, but the relationship between these two mechanisms is not clearly articulated.
In particular, Section 4.1 emphasizes cumulative soil water supply as the determinant of ecosystem decline, whereas Section 4.3 suggests that VPD dominates ecosystem response under compound stress. It remains unclear whether these are complementary processes or competing explanations.
I recommend that the authors clarify the conceptual framework linking water supply (soil moisture) and atmospheric demand (VPD), and explicitly state the conditions under which each mechanism becomes dominant.
_____ Minor Technical Issues _____
(1) Definition of normalized variables
The manuscript frequently refers to normalized variables (e.g., normalized GPP and normalized VPD), but the normalization procedures are not clearly defined. It would be helpful to explicitly state how these variables are normalized (e.g., relative to mean ambient conditions, baseline simulations, or maximum values), as this affects the interpretation of sensitivity across scenarios.
(2) Definition of "threshold."
The term “threshold” is used throughout the manuscript, but it is not formally defined. Please clarify how thresholds are identified (e.g., visually, statistically, or based on a predefined criterion), as this would improve reproducibility and interpretation.
(3) Use of GPP versus NPP
The manuscript switches between GPP and NPP across figures (e.g., Figure 2), but the rationale for this distinction is not clearly explained. A brief justification of when and why each metric is used would improve clarity.
(4) Figure captions
Some figures (e.g., Figure 3) lack sufficient information in the captions to be interpreted independently of the main text. Improving figure captions to be more self-contained (e.g., defining variables and explaining axes clearly) would enhance readability.