the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 31 Aug 2026
Limited modelled legacy effects of drought on carbon and water fluxes in a temperate oak forest despite hydraulic impairment
Abstract. The projected increase in the intensity, frequency and duration of summer drought extremes under climate change poses a growing risk to forest ecosystems, even in mesic regions such as the UK. To accurately predict the impact of future summer droughts, it is necessary to account for both their immediate and legacy effects, as well as their influence on ecosystem sensitivity to drought over time. We introduce a stomatal optimisation scheme, Profit-Max, within the Joint UK Land Environment Simulator (JULES) land surface model (LSM) to improve predictions of plant response to drought induced water stress. We further extend the model to represent drought legacy effects through two approaches to capture hydraulic impairment of xylem. We evaluate these developments by simulating an old growth oak forest at the Alice Holt site in southern England during the 2022 UK summer (July to September), when rainfall was substantially below average (77 mm of rainfall compared with a 1999–2018 median of 140 mm) and national temperatures exceeded 40 °C for the first time. During the 2022 drought year, the stomatal optimisation scheme predicted reductions in annual GPP and LE, −26 % and −33 % respectively, relative to an average year (2011–2017). This was larger than the measured reduction in LE of −6 %. The hydraulic legacy schemes simulated percentage loss of conductance of 9 % to 40 % ,with a carry-over effect on LE flux (−1 % to −11 %). We further explored ecosystem responses under more extreme future summers, including zero rainfall in spring (similar to the meteorology conditions of 2025) and a 50 % reduction in rainfall over the remainder of the year. Despite substantial increases in the percentage loss of conductance (10 % to 55 %), the simulated reductions in GPP and LE remained modest in 2023, consistent with observations. The magnitude of these legacy effects, however, was sensitive to the representation of hydraulic impairment and recovery. Overall, our results demonstrate the feasibility of representing drought legacy effects within a LSM and highlight the importance of hydraulic impairment assumptions in shaping projected ecosystem responses to extreme drought. While the simulations suggest limited legacy impacts on carbon uptake and latent heat fluxes at Alice Holt following the 2022 drought, evaluation of the underlying plant water status and hydraulic function will require independent observations of water potential and hydraulic function across the soil-plant-atmosphere continuum, as flux tower data alone may not fully constrain longer-term impacts on growth and resilience.
- Preprint
(2881 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-4729', Anonymous Referee #1, 06 Oct 2026
-
RC2: 'Comment on egusphere-2026-4729', Anonymous Referee #2, 26 Oct 2026
Baguley and colleagues implemented the stomatal optimisation model Profit-Max into the land surface model JULES. They then used two different ways to change maximum hydraulic conductivity over time. They compared the results of the different model formulations with eddy covariance measurements at an oak stand.
Unfortunately, I could not find anything that I have learned from the manuscript. The authors modelled an oak stand. Oak trees have large early-season vessels that are responsible for much of the water transport. These mostly cavitate, either during the summer or during the winter. Firstly, therefore, growth of new xylem is essential for oaks, but this is not included in the study. Secondly, there is no discussion of water transport, oak anatomy, or similar, and including this would probably have led to a very different PLC and its evolution.
Most of all, I missed any physiological insights or explanation about the realism, the physiological assumptions, or similar. Currently, it is a study where you put something into a model and something changes.
The manuscript is also written very sloppily. I noted typos as I read along but stopped noting them around line 210. There are just too many. Some equations are introduced with colons (:), some with commas (,), some with semicolons (;), and even some with full stops (.). Quite some times full stops are missing at the end of sentences. The figure captions are very poor: there are lots of typos, missing full stops, colours mentioned do not always match the ones in the figures but there are no legends that might help in the figures, etc.
I won't even list the typos I noted. The whole manuscript needs some care.
My major concerns are listed below. There are by line numbers in order of appearance, not in order of significance.l. 140: You use Mauna Loa CO2 data. Why not using Mace Head data, which would be more appropriate?
Fig 1: The gray area is 1 sigma. The 5-95th percentiles would be better especially for increasing time series such as temperature.
Fig 1: SPEI should have the time span mentioned.
l. 149 and Fig 1: Month should be in the middle of the time span. Cumulative precipitation starts increasing between ticks labelled Aug and Sep in 2022. In the text is written (l. 149): "with the rainfall starting to recover after September (Fig 1b)." So the tick is the 15th of the month?
l. 149f and Fig 1: What is the "cumulative climatic water deficit"? And on the Figure, it does not seem to be cumulative: "Daily climatic water deficit"?
l. 132 and l. 151ff: Why is there the section 2.3 for year 2023? In section 2.1 was written that you model 2011-2017 and 2022?
But then how do you look at legacy if you do not model the years after 2022? So if 2023 is modelled then Fig. 1 might want to show also 2023. A legacy effect might also be longer than one year, thus 2024 and 2025 might have been interesting.l. 181: What does "visually fitting" mean? Did you focus "visually" on the mean, the maximum, the spread?
l. 206: Why not use an optimisation algorithm like Newton or similar?
l. 217: psi_{s crit}: a space in subscript is "unusual" to say the least. Taking subscript "s" for a root variable is also misleading.
l. 225ff, section 2.4.3: I did not understand the two-phase water retention curves. What is it?
You say all simulations used the van Genuchten water retention curve (l226). You also say that the two-phase water retention curves used in JULES produce unrealistically low values (225f). So van Genuchten is a two-phase water retention curve? Never heard that.
"S" is defined as soil water content (with a typo in the units on line 232). Is it really volumetric soil water content or is it saturation (defined how)? If the former, why not using theta as a symbol like in the literature?
Is psi^0_s really the soil water potential at zero volumetric soil water content? That should be infinite, I guess. Or is this a new constant to which the linear model will converge?
You give Eq. (8). It would have been interesting to see the same for van Genuchten. Also psi^*_l and S^*_l are following the van Genuchten equation. Does that mean that one of the two conditions is enough (then you already have two equations for two unknowns)?l. 252ff, section 2.4.4.1: I really got flabbergasted reading section 2.4.4.1. You cannot follow this section.
For example, what is k_l and what is k_r? They are not given before.
Once w_l is 0.4 then it is 0.5 later.
Once w_k_stem is 0.01 and then again 0.04. And the value should probably be dependent on the size of the time step.
There seems to be an update of the parameters b and c of Eq. (4). But this is probably written in a roundabout way in l. 283ff but never really stated (perhaps in the first sentence of the section though). Fig. A1 is referenced but does not really help; it explains nothing.
The sensitivity analysis references appendix G but not at all Fig. A2.l. 294ff, section 2.4.4.2: I did not understand why I need to consider the variation of k with height. If I write E = k * dpsi for the current and the historic minimum, I get immediately Eq. (16).
It would have been nice to give the formula for the integral in Eq. (17), which I later found in appendix C.
If I understand right, E_historic / (psi_lmin - psi_rmin) = int_psi_rmin^psi_lmin k(psi) dpsi / (psi_lmin - psi_rmin) = k. Is it clear that k is changing if psi_lmin changes because the integral changes but also the denominator? And it also looks like that this k is fixed as long as I do not reach a new minimum psi_lmin. Where is the dependency with psi then, i.e. k(psi) as in Eq. (4)?
I could not understand Fig. 2a. The grey curves seems to be a vulnerability curve. But there is "Fractional height along trunk" on the y-axis.
And in Fig 2b, it looks like that once I have impaired (orange line), I will not be able to descend to very low leaf water potentials anymore because conductance will be low and stomata will close down. So then I will stay with the orange curve forever.l. 330ff, section 2.4.5: Now you lost me. If the whole impairment is not used in the stomatal limitation, how will Profit-Max ever give different transpiration? Profit-Max outputs A, Ci, from which you calculate g_s and the E with VPD. If Profit-Max always uses the same kmax, then it will give the same A, Ci pairs and hence g_s, E.
Perhaps a flow diagram might be useful? Or the equations in order of calculation? Or anything so that I could redo the study in my model.l. 369: did you re-scale satellite-derived LAI so that it is 0 during winter? Satellite-derived LAI is often non-zero in winter which leads to the same effect than seen here.
l. 371: If you "visually fitted" Profit-Max, did you also "visually fitted" CTRL?
What else changed? It is hard to me to see how you can have larger LE in spring for CTRL vs Profit-Max while having smaller GPP. Larger LE means normally larger stomatal conductance and hence larger Ci thus GPP.
You also have chosen a bad beta-function. But there must be more than one given all the work of Anne Verhoef and colleagues. And then I thought, JULES is using SOX now?l. 393, section 4.3: Any wood growth was not taken into account between years? This is very unrealistic for oak trees where you almost always have cavitation but only about 3-5 rings contribute to water transport. I know that it is a modelling study but what do we actually learn from the experiments? Especially, what do the more extreme drought scenarios tell us then? It eluded me also in the rest of the manuscript.
l. 367, section 4: The whole results section is not very informative and very tedious to read. I would recommend to merge it with the discussion.
l. 440ff: the beginning of discussion section is a summary. The specific summaries should rather be at the beginning of each discussion sections. For example, without a summary I do not know what "diverging hydraulic impact" means in section 5.1.
l. 478f, section 5.2: the discussion focuses mostly on the technical implementation but not on the physiological processes. It is very easy to formulate another mathematical (or algorithmic) description of impairment. But what is the idea behind the description? Are the one or the other descriptions based on different physiological assumptions (k_max depends on soil water potential or k_max depends on leaf water potential are not physiological assumptions but rather computational assumptions)? Would it be possible to bring the two assumptions of the two different descriptions into one mathematical description?
You talk about other literature that apparently used different impairment schemes. Yao et al. (2025) used even three different ones? What were they? Why did you not use them?
When you discuss soil water potential vs. leaf water potential, it sounds like that leaf water potential is the right thing to use. Why? The important one is xylem potential. So how much is the difference between xylem potential at the top of canopy vs. the bottom of the trunk? There should be an argument why one assumption is better than the other; and because you use a model, the argument should be quantitative.l. 535, section 5.2.1: I find this more fundamental than one might think. The last sentence states: "This assumption is consistent with the view that stomata respond primarily to local hydraulic status rather than directly to whole-plant hydraulic dysfunction." Why do we have Profit-Max then? We won't need any optimisation then and are back to Tuzet et al. (PCE 2003) or other mechanistic models like Dewar et al. (New Photo 2018) or Buckely et al. (GCB 2026).
l. 545: I am not aware of literature that says that stomatal conductance depends on leaf water content. Quite a few argue rather with water potentials (Leuning, Mott, Buckley, Dewar, ...). This is the second time in the paper that it is written that psi_l is a proxy for leaf water content.
l. 679, section 5.3: I would say that these are future directions for the model JULES in general, and not so much for the work presented in the current manuscript. I would have thought to see that, for example, xylem growth coupled to the impairment might be a future direction; perhaps even with a reference to a growth model that you find suitable. Or the transfer to sites with diffuse-porous species, where the presented impairment schemes might be better suited. Etc.
l. 590ff, section 6: Only the two last sentences are conclusions; the rest is a summary.
l. 654: How do you solve for c'?
Citation: https://doi.org/10.5194/egusphere-2026-4729-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 365 | 165 | 57 | 587 | 72 | 63 |
- HTML: 365
- PDF: 165
- XML: 57
- Total: 587
- BibTeX: 72
- EndNote: 63
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This study investigates how stomatal scheme and hydraulic legacy effect representation affects model behavior in JULES. The authors implemented a stomatal optimization scheme based on whole plant hydraulics (Profit-Max). Compared with existing Medlyn-style model in JULES, the Profit-Max version can better match flux tower observations at a deciduous forest site. Then, the study implemented and compared two different hydraulic legacy effect schemes: one with gradual accumulation (and recovery) of maximum hydraulic conductance loss following Mackay et al. 2015; the other one assumes instantaneous damage and no repairment. Model simulations for an extreme drought year and other drought scenarios predicted modest hydraulic legacy effects on GPP and ET.
While it is always good to see modeling explorations to better represent drought legacy, I have some major concerns of the robustness and usefulness of the study.
First, the two hydraulic impairment schemes and their parameterization are not well justified. I understand it is difficult to directly benchmark simulated PLC but Fig. A2 show large sensitivity of PLC to the three model parameters and it remains unclear to me why the particular set is chosen for interpretation and future simulations in Fig. 5-9. It also appears to me that PLC does not recover in the second scheme but wouldn't growth of new sapwood increase conductance and reduce apparent PLC? Overall, as the key module of the paper, I expect more details and discussions of impairment models. I have to say that the model description (section 2.4.4) is really hard to read due to numerous variable/parameters/subscripts. Reorganization is needed to clarify the model structure and a variable table would be helpful.
Second, there is inconsistency between Table 1 and Figure 3. Table 1 report CTRL predicts almost doubled GPP compared with Profit-Max while Figure 3 shows CTRL greatly underestimate GPP. In fact, I am wondering whether this is a fair comparison because it seems Profit-Max model is calibrated with the tower observation (L180-182) but CTRL does not. I worry this delivers an incorrect message that Profit-Max has substantial advantage over the Medlyn model.
Minor and editorial comments:
L68: mechanist -> mechanistic
L85: related -> similar
L105: capture -> investigate or explore? There are no observations of legacy effects to capture in this study
L115-118: This is the key question and hypothesis of the study, but I feel it does not read novel/useful to the modeling community (cumulative reduction in conductance -> more legacy effects)
L130: To to -> To
L 194: max(An(psi_l)) is confusing to me. Is it equivalent to An when psi_l is 0 (fully saturated)
L 225-235: The whole section 2.4.3 reads confusing to me. Why extremely dry soil water conditions need special treatment? Wouldn't it be easier to simply set a lower threshold for psi_s?
L316: I don't understand eqn 14 to eqn 15, why k(psi) cancel out when dividing two transpiration terms. In addition, is E_extreme the same as E_historic? If not, what's their difference?
L517-520: This statement highly relies on the parameters of each impairment model, which lacks benchmarking and calibration.
Figure 2. Panel (a) is really confusing because y axis represents both height and PLC. I suggest reorganizing panel (a)