the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 Nov 2025
Interpreting carbon-water trade-offs in Daisy crop model using Pareto-based calibration
Abstract. Improving the simulation of carbon and water exchanges is crucial for reliable crop modelling under changing climate conditions. Although model calibration is a key step, optimising multiple outputs can be challenging and often reveals trade-offs between calibration objectives. We applied a Pareto-based multi-objective calibration with the Speed-constrained Multi-objective Particle Swarm Optimisation (SMPSO) algorithm to the Daisy soil–plant–atmosphere model, targeting dry matter (DM), net ecosystem exchange (NEE), and latent heat flux (LE) of winter wheat crops.
The optimal parameter set achieved good accuracy for all objectives (RMSE = 0.948 t ha-1 for DM, 1.49 gC m-2 day-1 for daily NEE and 30.7 W m-2 for daily LE) but revealed singular trade-offs. The strong compromise between dry matter and NEE likely suggests wrong parameterisation and measurement bias, while the trade-off between NEE and LE reflects equifinality issues from evapotranspiration partitioning. Lastly, this analysis also pointed out limitations in simulating stomatal regulation during heatwaves conditions, supporting the decoupling between transpiration and carbon assimilation. These findings show that Pareto-based calibration can also serve as a diagnostic tool, identifying structural weaknesses and guiding targeted improvements in process representation for more robust crop model evaluation.
- Preprint
(1950 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-4987', Anonymous Referee #1, 12 Jan 2026
-
RC2: 'Comment on egusphere-2025-4987', Anonymous Referee #2, 21 Jan 2026
General comments
This manuscript presents a technically strong and carefully executed modelling study. The optimisation framework is clearly described, the methods are sound. The discussion of model behaviour and limitations is insightful. Overall, I find the work to be of good quality and relevant to the modelling community. To further strengthen the manuscript, I recommend minor revisions focused mainly on clarifying the scientific questions motivating the study and better positioning the model improvements in terms of real-world process understanding, rather than methodological advancement alone.
Specific comments
1. Scientific framing and justification of tool choice.
The introduction would benefit from a clearer articulation of the scientific questions motivating this work. Parts of the current manuscript read as if improving DAISY model performance is the primary objective, rather than a means to address broader bio-ecological questions. Explicitly clarifying what real-world processes or uncertainties this model improvement aims to resolve, and why DAISY is an appropriate tool for this purpose, would strengthen the scientific contribution and better justify the modelling effort.
In addition, the choice of output variables lacks clear justification. The importance and relevance of NEE and LE should be explained more explicitly. These concepts may be straightforward for the authors, but additional explanation would benefit general readers and clarify their scientific relevance.
2. Model simplification, dynamic processes, and interpretation
The authors provide a clear and well-reasoned discussion of structural limitations of the current model (e.g. constant SLA). To strengthen this section, it would be helpful to clarify whether introducing additional dynamic processes is expected to significantly improve predictions (e.g. of LE and NEE) for answering the research questions, and what will be the trade-offs by making the model more complex.
3. Linking model limitations to real-world mechanisms (LE and NEE).
The discussion of limitations in predicting LE and NEE is thoughtful and balanced. It would be particularly insightful to briefly expand on how these diagnosed limitations relate to real-world mechanisms. For example, are the relevant processes already well understood from experimental studies but not yet parameterised in models, or do the results point to areas where targeted experiments or observational designs are needed? Even a short reflection along these lines would enhance the scientific relevance of the work.
4. Clarity and presentation
Overall presentation is good, but several instances of vague wording (e.g. “good accuracy”), long sentences, and undefined acronyms reduce clarity. Addressing these issues would improve readability.
Technical corrections
Abstract
Line 12: “good accuracy” is vague. The term “good” lacks context for readers unfamiliar with the subject. Consider removing it or replacing it with quantitative information.
Line 14: “wrong parameterisation” is a strong claim. The discussion suggests that simplification may contribute to bias, but no quantitative evidence is provided to demonstrate that the parameterisation is incorrect. Model simplification is a design choice; the issue is appropriateness rather than right or wrong. Rephrasing would improve accuracy.
Introduction
Line 35: The sentence introducing g₁ is abrupt. It moves directly from physiology to a specific parameter without transition. A brief link from physiological concepts to their model formulation would improve readability.
Line 51: The term equifinality is introduced without explanation. A brief definition here, similar to the one later in the discussion, would improve accessibility.
Line 57: NEE and LE are introduced without justification. Please briefly clarify why these outputs were selected and what ecological or biophysical significance they represent.
Lines 58–59: The flow may improve if concepts are introduced in a consistent order (e.g. DM, then NEE, then LE), rather than switching between topics.
Methods
Line 69: TOB, SKY, SAH, and SMA are unclear. Please clarify whether these refer to varieties or growing seasons. If they represent seasons, it may be clearer to use digits.Line 102: Unsure whether g₁ here refers to the same parameter introduced in the introduction.
Line 104: It would be beneficial to clarify why model modifications were required.
Line 136: The description of “individual exploration” and “social learning” is unclear. Does this refer to local versus global optima?
Line 154: The description of “once for each growing season” seems to suggest all seasons were used.
Line 157: Please clarify how swarm size and total simulation number were chosen. Were these based on empirical experience or hyperparameter tuning? Showing convergence behaviour (e.g. in an appendix) would increase confidence.
Line 165: Please clarify what is meant by EF being “easily interpreted.” Is this relative to other metrics, or related to its physical meaning for the output variables?
Results
Line 176: “Most solutions” is vague. Please consider reporting this as a percentage.
Line 202: “Good overall agreement” lacks quantitative support. Please provide numerical values or remove the qualitative descriptor.
Line 204: Figure 3 also shows underestimation toward the end of the season; this could be mentioned explicitly.
Lines 245 and 253: Table numbering appears inconsistent. Table 1a and 1b may need to be Table 3a and 3b?
Discussion
Line 301: The statement that constant values may lead to significant bias is plausible, but quantitative evidence is not presented. Please clarify whether incorporating dynamic SLA (e.g. dependent on leaf age or water status) is expected to substantially improve predictions, or whether this remains a hypothesis.
Lines 303–305: The sentence describing EC data filtering and uncertainty estimation could be rewritten for clarity.
Line 335: CLM is introduced without prior explanation. Please define it when first mentioned.
Line 374: Please rewrite for clarity. It is unclear what is meant by “This decoupling behaviour.” Does this refer to increased transpiration while assimilation remains unchanged? Also, please clarify whether “April 19” refers to a date April 19th or the year 2019.
Citation: https://doi.org/10.5194/egusphere-2025-4987-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 133 | 86 | 16 | 235 | 11 | 13 |
- HTML: 133
- PDF: 86
- XML: 16
- Total: 235
- BibTeX: 11
- EndNote: 13
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General comments
The paper clearly demonstrates the challenges of simulating the links between stomatal opening, dry matter production, and the fluxes of water and CO₂ using an advanced soil–water–plant–atmosphere model with a fully coupled photosynthesis and energy balance scheme for simulating plant transpiration. The work is extensive and builds upon a comprehensive sensitivity analysis of 200 parameters to identify the most influential ones. In the present study, these parameters are calibrated using a Pareto-based calibration strategy and high-quality eddy covariance data from four growing seasons. Overall, the work is well structured and appears to be of high quality. In general, I have some comments, mostly suggestions related to the discussion on specific parts of the manuscript, as outlined below.
In my opinion, the main limitation of the study is the lack of data on soil water dynamics in different soil layers. It would have been very informative to evaluate how well the model simulated soil drying, both due to root water uptake and soil evaporation. Although different combinations of soil physical properties were tested in the sensitivity analysis and some root parameters were calibrated, the selected parameter set may not have been optimal. Such data could also have helped to identify the causes of the discrepancies between simulated and observed evapotranspiration. This aspect is missing from the discussion. A similar issue applies to the lack of leaf area data, which is only represented indirectly through biomass measurements of the different plant organs.
Specific comments
It would be relevant to mention the work of Delhez et al. (2025) already in the introduction. Currently, this study is cited only in the Materials and Methods section as the source of data, management and soil information, and the internal sensitivity analysis. However, it may be useful to clarify that the present study is a follow-up to that work, focusing on the calibration of the most relevant parameters identified previously. This could be stated in the final part of the introduction.
Line 55:
It could be useful to explain why Daisy was selected for this study compared to the other models mentioned earlier (lines 25–30). It is unclear whether Daisy is unique in coupling Richards’ equation for soil water dynamics with a Farquhar-based photosynthesis model, or whether the choice was driven by other factors. For example, Daisy being open source and implemented in C++ may have facilitated the implementation of the Pareto-based modelling framework.
Lines 90–95:
Is the Daisy setup similar to that used in Delhez et al. (2025)? If so, it may be relevant to mention that a drainage system based on the Hooghoudt equation is included. Alternatively, this information could simply be referenced to Delhez et al. (2025), where it is described. Otherwise, there is a risk of soil profile flooding when an aquitard layer is present and no drainage system is implemented.
Figure 5. A latent heat flux of 600 w/m2 in late April seems extreme, since the theoretical potential clear sky radiation is 750 w/m2. Would have been relevant to see the data on air temperature, humidity, and wind speed during this event.
Section 4.2:
The issue of calibrating the model separately for each cultivar grown in only one season is not discussed. For instance, the large differences observed in some parameters may be difficult to explain purely on a genetic basis. The large range in parameters such as SOrgPhotEff and stemPhotEff could potentially be caused by seasonal stress factors—such as disease or water or nitrogen stress—not explicitly represented in the model, rather than by genetic differences. It might have been more robust to calibrate a single cultivar across all seasons, given that modern wheat cultivars generally do not differ substantially in yield potential or growth patterns.
Lines 330–340:
It could be added that models based on Richards’ equation tend to overestimate soil evaporation. One reason is the difficulty in obtaining accurate hydraulic parameters for the surface soil layers, which is further complicated by soil water hysteresis. As a result, the hydraulic conductivity curve used to calculate potential matrix exfiltration may be too high, leading to an overestimation of soil evaporation in Daisy. This issue will also be relevant for a model with a more mechanistic coupling between the surface energy and water balance, simulating microclimate effects on soil EP as a fully coupled approach.
Lines 330–340 (continued):
Daisy also includes a transfer function controlled by the EpInterchange coefficient, which allows energy transfer from a dry soil surface to the canopy (default value β = 0.6 [–]). This function could potentially explain the relatively high simulated latent heat flux (LE) during periods of low leaf area at the beginning and end of the time series. In theory, this parameter could convert some soil water into LE under dry surface conditions. The parameter was not included among the 200 parameters in the initial sensitivity analysis by Delhez et al. (2025), and its omission may have resulted in an overestimation of transpiration as a starting point for the SSOC iterations.
Again, this interpretation assumes that Daisy uses accurate hydraulic conductivity curves for the surface layers, as discussed above.
Technical corrections
Line 70:
The following sentence is confusing: “(Meza et al., 2018; 2023). As the same cultivar was sown for VAL and SAH, the VAL season was set aside for validation.” This is unclear. Instead of naming growing seasons after cultivars, it might be clearer to refer to them by year. Furthermore, as noted above, it is not clearly stated whether a specific calibration was performed for each cultivar/season or whether a single parameterised cultivar was used across all seasons in the text. This information is only apparent from Table B1 in Appendix B.
Table 2:
Not all parameters listed can be found in the Daisy documentation. For example, it is unclear what k_net refers to in the Daisy reference manual. It would be helpful to include the exact name from the setup files in this list as a separate column.