the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 16 Apr 2024
Developing the DO3SE-crop model for Xiaoji, China
Abstract. A substantial body of empirical evidence exists to suggest that elevated O3 levels are causing significant impacts on wheat yields at sites representative of highly productive arable regions of China. Here, we extend the DO3SE model (designed to estimate total- and stomatal- O3 deposition for risk assessment) to incorporate a coupled Anet-gsto model to estimate O3 uptake, an O3 damage module (that impacts instantaneous Anet and the timing and rate of senescence), and a crop phenology, carbon allocation and growth model based on the JULES-Crop model. The model structure allows scaling from the leaf to the canopy to allow for multiple leaf populations and canopy layers. The DO3SE-crop model is calibrated and parametrised using O3 fumigation data from Xiaoji, China for the year 2008 and for an O3 tolerant and sensitive cultivar. The calibrated model can simulate key physiological variables, crop development, and yield with a good level of accuracy compared to experimental observations. DO3SE-crop accurately depicted the difference in yield reductions under ambient and elevated O3 treatments for wheat cultivars Y16 (tolerant) and Y2 (sensitive) with regressions of modelled and observed absolute yields resulting in an R² of 0.99 and an RMSE of 9.27 g/m². Further, when evaluated for 2007 and 2009 for all cultivars, the DO3SE-crop model simulated O3-induced yield losses of 4–25 % compared to observed yield losses of 12–34 %, with an R² of 0.73 and an RMSE of 58.41 g/m². Additionally, our results indicate that the variance in yield reduction is primarily attributed to the premature decrease in carbon assimilation to the grains under elevated O3 exposure. This is linked to accelerated leaf senescence, which brings leaf senescence forward by 7–9 days under elevated O3 treatments.
- Preprint
(1602 KB) - Metadata XML
-
Supplement
(824 KB) - BibTeX
- EndNote
Status: closed
-
CC1: 'Comment on egusphere-2024-694', Owen Cooper, 24 May 2024
My comments can be found in the attached pdf.
- AC1: 'Reply on CC1', Pritha Pande, 24 May 2024
- AC4: 'Reply on CC1', Pritha Pande, 29 Aug 2024
-
RC1: 'Comment on egusphere-2024-694', Anonymous Referee #1, 27 May 2024
This paper presents a modeling study to evaluate the impact of ozone damage on Chinese wheat yields using the developed DO3SE-Crop model and field experimental data. Overall, this is a comprehensive study that provides valuable insights into the effects of ozone on food security. However, there are areas that could be improved.
The title of the paper could be modified to reflect a broader scope of the model’s purposes and applications, rather than focusing solely on a specific site in China.
A primary concern is that only data from 2008 was used to train the model. This raises questions about the representativeness and robustness of the model when applied to other years and regions.
The extremely high correlation between the modeled and observed yields (R² of 0.99) suggests potential overfitting, which could lead to biased prediction results. The authors should address this issue to be more convincing.
It appears that the ozone-induced yield losses derived from field observations are significantly larger than the corresponding simulation results. The reasons behind this discrepancy should be investigated and discussed.
In the methods section, the authors cite numerous related studies for the formulas of different modules. While the sensitivity of these empirical parameters has been evaluated, the uncertainties arising from the process need to be elaborated upon.
Uncertainties may arise from assumptions, such as “We then assume that these values are consistent across cultivars and years” (Line 580). These need to be addressed in more detail.
The figures in the paper could be better presented to improve visualization and readability. For instance, the colors of different legends in Figure 3 are difficult to distinguish.
The large differences between the modeled and observed Anet and gO3 in Figure 4 need to be explained, as the model tends to underestimate these values. Additionally, correct the figure caption for panels c and d, as they appear to be swapped. Ensure consistency in font sizes, such as in Figure 5c and d.
Minor Comments
Correct typographical errors in the abstract, such as "...regions of ufor risk assessment."
In Line 50, O3 should be in subscript form.
In Line 104, "Evaluation" should be in lower case.
Ensure references are complete and correctly formatted, including publication year, journal name, etc. For example, the reference to Yang, L. et al. should include all pertinent information.
Add missing parentheses in Figure 5b.
Citation: https://doi.org/10.5194/egusphere-2024-694-RC1 - AC2: 'Reply on RC1', Pritha Pande, 29 Aug 2024
-
RC2: 'Comment on egusphere-2024-694', Anonymous Referee #2, 27 May 2024
The manuscript “Developing the DO3Se-crop model for Xiaoji, China” written by Pande et al. used the existing measured data of winter wheat under elevated and ambient free-air chamber condition from Xiaoji from 2007 to 2009 to parametrize and develop/extend their current ozone deposition model -DO3SE.
The MS contains many error of spacing. Line justification and larger spacing would be required to support better reading and evaluation of the MS. The MS was not in balance between introduction, material and method and results sections. Introduction is still minimal which needs further elaboration and literature review the recent developments of crop models (which includes ozone) and it needs an explanation why DO3SE needs to include crop modeling routine. Materials and methods were heavy with modeling description and equations which at some points one do not know which processes need to be focused. Also, which field data supports the parameterization? The calibration and validation and sensitivity analysis were mixed up and not well described which give an impression that the model was over fitted of grain yield simulation based on many parameters. The results section did not follow the objectives mentioned in the introduction.
While the MS emphasized the different cultivars responses. The simulation steps and results for these are not clear.
Many detailed comments are listed below.
Introduction
Line 55-65 needs to be more elaborated about state of art in crop modeling with ozone routine.
Line 94: is there any modeling comparison in the MS e.g. JULES and DO3SE-crop?
Materials and methods
Fig1. There is no connection of input climate variables to photosynthesis and crop growth. Do ozone and thermal time drives the crop biomass growth? There is no legend or further caption to describe the information inside the diagram.
Table 1. The table was rather simple and less informative since one does not know the real, absolute values of climatic and ozone variables. How much are different among treatments, year and growing season in term of input variables. Fumigation time? Lack of summary on climatic input and ozone that is really hard to understand the crop responses and modeling performance.
Line 512. Did the trial measure the gO3?
Section 2.2:
There is a mixed and unclear description how models were tested. The sensitivity analysis should be done firstly to screen the most important parameters, then coming to calibration, then validation. Where is the validation description, which treatment, modeling performance metrics, which outputs were validated?....
Line 533-543: It means that sensitivity analysis was not done for phenology? There are 11 parameters and three outputs, how the calibration is done in sequence to simulate the different developmental stages. It is not clear.
Section 2.1. Line 504-513, it is not clear with the measurement data, was there significant impacts of ozone among ozone treatment and interactions with cultivars? Why 2008 was selected for calibration? If there is no considerable impact of ozone on crop physiological processes, the modeling parameterization is very weak. At least, this point is hardly seen in the description of data.
Line 544-562: It is really unclear for the calibration process? How the calibration was done manually because there are a lot of parameters and outputs were involved. Line 549, in which steps? How parameter were changed? What are key processes should be considered for which treatment (ambient and elevated ozone). If the JULES model was already used and well calibrated why the newly crop model (DO3SE-Crop) needs to consider some other processes here (e.g. phenology, carbon allocation…). Did observed data really prove the change of carbon allocation to different organ due to ozone? (line 646-649). The description seems the many processes have been touched and kind of overfitting with using many parameters to get correctly simulated final biomass and yield?
There is measurements which has been described from 504-518, but this was not clear how these measurements that were used for the calibration. It might be better to list out the key measured variables, which one is used for calibration. Xhu et al., 2011 is not in the literature?
Line 551: Why the vcmax and jmax need to be calibrated again? Why the measured vcmax or jmax could not be used directly for the models to simulate Anet and gs?
Line 558: why the LAI needs to be between 4-7 m2 m-2. Did it mean that there is no measurement of LAI? Recent work which compared three models including from DO3SE-Crop model in Nguyen et al., (2024), it seems that the DO3SE-crop model did not perform well for the green leaf area simulation. LAI is the most important crop growth metric which influences the simulating assimilation, stomatal conductance (thus ozone fluxes, see diagram fig 1) at leaf and canopy scale which in consequence affects to leaf senescence due to ozone, then thus biomass. It is not clear how DO3SE-Crop model improves the LAI simulation.
Results
The result section did not follow the mentioned objectives or at least the headings were not clearly shown these. The objective 2 was not touched with regards of tolerance and sensitive cultivars. How the simulation/parameters for tolerance and sensitive cultivars were done?
Line 580-585: This is important. It is not clear how to simulate onset of leaf senescence? Did the observed result show both the early onset of senescence and increase of senescence rate due to ozone?
Figure 3: It is not clear. The duration of flag leaf period means that from flag leaf appearance to dead leaf? The line 604: this Fig3c was wrong. Line 614 with Fig. 4 has the same issue. Data for which day for which cultivar? It is weak to show the simulated value and where are the measured data? Fig 4c &4d for the same AA treatment…Why the modelled and simulated were so much deviated? Where is error bar of measurement?
Section iv) did the trial measure the dynamic change of dry matter in different organs or only just final biomass and yield? Figure 5a, what is harvest? 5b? Result from which models? Which treatment and cultivar? Really confusing and unclear.
Figure 6. Relative to what? Was the CCI absolute? It is really hard to understand why the slope of fLS in ambient was already very steep as similar in elevated treatment? Model overestimated the drop of leaf senescence?
Fig7. The section emphasized the seasonal variability but the factor of year was not shown. Why the data was not displayed for each year? Where is error bar of measurement?
Line 688-699 One could not see from the figure? What is reason why ozone fumigation was too short? Why the measured data show bigger yield reduction in ozone in 2007 compared to other years? Problem of measured data?
Figure 8: for both ambient and elevated?
Discussion
L697: it is not clear to see the distinction of simulated tolerance and sensitive? How model has configured and related results? The range was almost similar for two groups of cultivar? Is this significant?
Line 702-703 it is not clear. Line 706: how about the role of phenology? Because model overestimated phenology?
Line 710 and elsewhere, check unit
Line 721: This is not true. Model overestimated for 2008-2009 at least 15-20 days?
Line 737-747. It is confusing and need clarification also for the MM and result e.g. Fig 1 & Fig2. The model simulated the ozone uptake for only flag leaf (top canopy) and divide these to different layers (Gaussian integration over flag leaf depth) or the whole canopy and divide these to different layers (Gaussian integration over LAI depth)?
Line 750: this is not proved. I did not see the observation of Anet and gO3
Line 757: value of this vmax is really high while the yield was not super high. Is there any explanation for this?
Line 765: There is no data of measured carbon allocation. Why one need to change the JULES configuration? Line 769: It is not clear. Did measurement include respiration, LAI, stem biomass data? Reply only the final biomass but calibrating many parameters means that this is an overfitting of the model.
Line 770-771. And mean what?
Line 801, it is not true based on the slope in Figure 6a.
Line 802-803. I could not see for other validated years and cultivar. Data was only in Fig 6 which is weak.
Supplementary materials
Are the appendix S1a similar to Figure S1…?
Figure S1.
While the calibration results looked very well, the validation of maturity was not good, mostly overestimated by model, almost which are too much. This inaccuracy of maturity might cause the simulated leaf area and biomass even more than the ozone impacts. The model must be improved phenology simulation before taking consideration of ozone impacts.
Even here is supplementary material, legends were very confusing. What is all models phenology? Use consistent the term calibration for training, evaluation for testing. I did not see the emergence symbols.
Figure S2: the model performance was perfect which raising question of overfitting (again) is or over calibration or the variation of observation was very minor. It is not clear about the units, where is root mean square error which have mentioned in the text, what are color of symbols, data from which plots?
Citation: https://doi.org/10.5194/egusphere-2024-694-RC2 - AC3: 'Reply on RC2', Pritha Pande, 29 Aug 2024
-
AC5: 'Reply on RC2', Pritha Pande, 29 Aug 2024
Please find the further changes in the two responses regarding the Table and Figure numbers.
In the main paper, we have included earlier Tables S3 and S4, now as an appendix (as Table A1 and Table).
Also, we have combined Table S2 with the earlier Tables S3 and S4 (now Tables A1 and A2 in the appendix). I have now made the changes to the comments relating to the same.
There was an error in one of my responses regarding (Table S2a. and S2b), which is now corrected in the attached document.
Status: closed
-
CC1: 'Comment on egusphere-2024-694', Owen Cooper, 24 May 2024
My comments can be found in the attached pdf.
- AC1: 'Reply on CC1', Pritha Pande, 24 May 2024
- AC4: 'Reply on CC1', Pritha Pande, 29 Aug 2024
-
RC1: 'Comment on egusphere-2024-694', Anonymous Referee #1, 27 May 2024
This paper presents a modeling study to evaluate the impact of ozone damage on Chinese wheat yields using the developed DO3SE-Crop model and field experimental data. Overall, this is a comprehensive study that provides valuable insights into the effects of ozone on food security. However, there are areas that could be improved.
The title of the paper could be modified to reflect a broader scope of the model’s purposes and applications, rather than focusing solely on a specific site in China.
A primary concern is that only data from 2008 was used to train the model. This raises questions about the representativeness and robustness of the model when applied to other years and regions.
The extremely high correlation between the modeled and observed yields (R² of 0.99) suggests potential overfitting, which could lead to biased prediction results. The authors should address this issue to be more convincing.
It appears that the ozone-induced yield losses derived from field observations are significantly larger than the corresponding simulation results. The reasons behind this discrepancy should be investigated and discussed.
In the methods section, the authors cite numerous related studies for the formulas of different modules. While the sensitivity of these empirical parameters has been evaluated, the uncertainties arising from the process need to be elaborated upon.
Uncertainties may arise from assumptions, such as “We then assume that these values are consistent across cultivars and years” (Line 580). These need to be addressed in more detail.
The figures in the paper could be better presented to improve visualization and readability. For instance, the colors of different legends in Figure 3 are difficult to distinguish.
The large differences between the modeled and observed Anet and gO3 in Figure 4 need to be explained, as the model tends to underestimate these values. Additionally, correct the figure caption for panels c and d, as they appear to be swapped. Ensure consistency in font sizes, such as in Figure 5c and d.
Minor Comments
Correct typographical errors in the abstract, such as "...regions of ufor risk assessment."
In Line 50, O3 should be in subscript form.
In Line 104, "Evaluation" should be in lower case.
Ensure references are complete and correctly formatted, including publication year, journal name, etc. For example, the reference to Yang, L. et al. should include all pertinent information.
Add missing parentheses in Figure 5b.
Citation: https://doi.org/10.5194/egusphere-2024-694-RC1 - AC2: 'Reply on RC1', Pritha Pande, 29 Aug 2024
-
RC2: 'Comment on egusphere-2024-694', Anonymous Referee #2, 27 May 2024
The manuscript “Developing the DO3Se-crop model for Xiaoji, China” written by Pande et al. used the existing measured data of winter wheat under elevated and ambient free-air chamber condition from Xiaoji from 2007 to 2009 to parametrize and develop/extend their current ozone deposition model -DO3SE.
The MS contains many error of spacing. Line justification and larger spacing would be required to support better reading and evaluation of the MS. The MS was not in balance between introduction, material and method and results sections. Introduction is still minimal which needs further elaboration and literature review the recent developments of crop models (which includes ozone) and it needs an explanation why DO3SE needs to include crop modeling routine. Materials and methods were heavy with modeling description and equations which at some points one do not know which processes need to be focused. Also, which field data supports the parameterization? The calibration and validation and sensitivity analysis were mixed up and not well described which give an impression that the model was over fitted of grain yield simulation based on many parameters. The results section did not follow the objectives mentioned in the introduction.
While the MS emphasized the different cultivars responses. The simulation steps and results for these are not clear.
Many detailed comments are listed below.
Introduction
Line 55-65 needs to be more elaborated about state of art in crop modeling with ozone routine.
Line 94: is there any modeling comparison in the MS e.g. JULES and DO3SE-crop?
Materials and methods
Fig1. There is no connection of input climate variables to photosynthesis and crop growth. Do ozone and thermal time drives the crop biomass growth? There is no legend or further caption to describe the information inside the diagram.
Table 1. The table was rather simple and less informative since one does not know the real, absolute values of climatic and ozone variables. How much are different among treatments, year and growing season in term of input variables. Fumigation time? Lack of summary on climatic input and ozone that is really hard to understand the crop responses and modeling performance.
Line 512. Did the trial measure the gO3?
Section 2.2:
There is a mixed and unclear description how models were tested. The sensitivity analysis should be done firstly to screen the most important parameters, then coming to calibration, then validation. Where is the validation description, which treatment, modeling performance metrics, which outputs were validated?....
Line 533-543: It means that sensitivity analysis was not done for phenology? There are 11 parameters and three outputs, how the calibration is done in sequence to simulate the different developmental stages. It is not clear.
Section 2.1. Line 504-513, it is not clear with the measurement data, was there significant impacts of ozone among ozone treatment and interactions with cultivars? Why 2008 was selected for calibration? If there is no considerable impact of ozone on crop physiological processes, the modeling parameterization is very weak. At least, this point is hardly seen in the description of data.
Line 544-562: It is really unclear for the calibration process? How the calibration was done manually because there are a lot of parameters and outputs were involved. Line 549, in which steps? How parameter were changed? What are key processes should be considered for which treatment (ambient and elevated ozone). If the JULES model was already used and well calibrated why the newly crop model (DO3SE-Crop) needs to consider some other processes here (e.g. phenology, carbon allocation…). Did observed data really prove the change of carbon allocation to different organ due to ozone? (line 646-649). The description seems the many processes have been touched and kind of overfitting with using many parameters to get correctly simulated final biomass and yield?
There is measurements which has been described from 504-518, but this was not clear how these measurements that were used for the calibration. It might be better to list out the key measured variables, which one is used for calibration. Xhu et al., 2011 is not in the literature?
Line 551: Why the vcmax and jmax need to be calibrated again? Why the measured vcmax or jmax could not be used directly for the models to simulate Anet and gs?
Line 558: why the LAI needs to be between 4-7 m2 m-2. Did it mean that there is no measurement of LAI? Recent work which compared three models including from DO3SE-Crop model in Nguyen et al., (2024), it seems that the DO3SE-crop model did not perform well for the green leaf area simulation. LAI is the most important crop growth metric which influences the simulating assimilation, stomatal conductance (thus ozone fluxes, see diagram fig 1) at leaf and canopy scale which in consequence affects to leaf senescence due to ozone, then thus biomass. It is not clear how DO3SE-Crop model improves the LAI simulation.
Results
The result section did not follow the mentioned objectives or at least the headings were not clearly shown these. The objective 2 was not touched with regards of tolerance and sensitive cultivars. How the simulation/parameters for tolerance and sensitive cultivars were done?
Line 580-585: This is important. It is not clear how to simulate onset of leaf senescence? Did the observed result show both the early onset of senescence and increase of senescence rate due to ozone?
Figure 3: It is not clear. The duration of flag leaf period means that from flag leaf appearance to dead leaf? The line 604: this Fig3c was wrong. Line 614 with Fig. 4 has the same issue. Data for which day for which cultivar? It is weak to show the simulated value and where are the measured data? Fig 4c &4d for the same AA treatment…Why the modelled and simulated were so much deviated? Where is error bar of measurement?
Section iv) did the trial measure the dynamic change of dry matter in different organs or only just final biomass and yield? Figure 5a, what is harvest? 5b? Result from which models? Which treatment and cultivar? Really confusing and unclear.
Figure 6. Relative to what? Was the CCI absolute? It is really hard to understand why the slope of fLS in ambient was already very steep as similar in elevated treatment? Model overestimated the drop of leaf senescence?
Fig7. The section emphasized the seasonal variability but the factor of year was not shown. Why the data was not displayed for each year? Where is error bar of measurement?
Line 688-699 One could not see from the figure? What is reason why ozone fumigation was too short? Why the measured data show bigger yield reduction in ozone in 2007 compared to other years? Problem of measured data?
Figure 8: for both ambient and elevated?
Discussion
L697: it is not clear to see the distinction of simulated tolerance and sensitive? How model has configured and related results? The range was almost similar for two groups of cultivar? Is this significant?
Line 702-703 it is not clear. Line 706: how about the role of phenology? Because model overestimated phenology?
Line 710 and elsewhere, check unit
Line 721: This is not true. Model overestimated for 2008-2009 at least 15-20 days?
Line 737-747. It is confusing and need clarification also for the MM and result e.g. Fig 1 & Fig2. The model simulated the ozone uptake for only flag leaf (top canopy) and divide these to different layers (Gaussian integration over flag leaf depth) or the whole canopy and divide these to different layers (Gaussian integration over LAI depth)?
Line 750: this is not proved. I did not see the observation of Anet and gO3
Line 757: value of this vmax is really high while the yield was not super high. Is there any explanation for this?
Line 765: There is no data of measured carbon allocation. Why one need to change the JULES configuration? Line 769: It is not clear. Did measurement include respiration, LAI, stem biomass data? Reply only the final biomass but calibrating many parameters means that this is an overfitting of the model.
Line 770-771. And mean what?
Line 801, it is not true based on the slope in Figure 6a.
Line 802-803. I could not see for other validated years and cultivar. Data was only in Fig 6 which is weak.
Supplementary materials
Are the appendix S1a similar to Figure S1…?
Figure S1.
While the calibration results looked very well, the validation of maturity was not good, mostly overestimated by model, almost which are too much. This inaccuracy of maturity might cause the simulated leaf area and biomass even more than the ozone impacts. The model must be improved phenology simulation before taking consideration of ozone impacts.
Even here is supplementary material, legends were very confusing. What is all models phenology? Use consistent the term calibration for training, evaluation for testing. I did not see the emergence symbols.
Figure S2: the model performance was perfect which raising question of overfitting (again) is or over calibration or the variation of observation was very minor. It is not clear about the units, where is root mean square error which have mentioned in the text, what are color of symbols, data from which plots?
Citation: https://doi.org/10.5194/egusphere-2024-694-RC2 - AC3: 'Reply on RC2', Pritha Pande, 29 Aug 2024
-
AC5: 'Reply on RC2', Pritha Pande, 29 Aug 2024
Please find the further changes in the two responses regarding the Table and Figure numbers.
In the main paper, we have included earlier Tables S3 and S4, now as an appendix (as Table A1 and Table).
Also, we have combined Table S2 with the earlier Tables S3 and S4 (now Tables A1 and A2 in the appendix). I have now made the changes to the comments relating to the same.
There was an error in one of my responses regarding (Table S2a. and S2b), which is now corrected in the attached document.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 446 | 147 | 50 | 643 | 52 | 29 | 24 |
- HTML: 446
- PDF: 147
- XML: 50
- Total: 643
- Supplement: 52
- BibTeX: 29
- EndNote: 24
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1