Evaluating a Simple Algorithm for an Evapotranspiration Retrieval Energy Balance Model in Mediterranean Citrus Orchards

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study evaluates the performance of the Simple Algorithm for Evapotranspiration Retrieving (SAFER) in a 0.7 ha citrus orchard located in Sicily, Italy, using remote sensing and eddy covariance data. The study employed satellite imagery without thermal bands, including Landsat 8, Sentinel-2, and MODIS. Both mass and surface energy balance (SEB) fluxes were processed using the SAFER algorithm.

The spatial resolution used in the study was 10 m/pixel, and remote sensing data were modeled through the R package proposed by Teixeira et al. (2021). Model calibration was conducted using regression analysis to account for local conditions. While the coefficients c and d provided in Equation 7 were adopted from Teixeira et al. (2021), the coefficients a and b in Equation 9 were derived through calibration with local data in the present study.

The model was applied for the years 2021–2022, and its performance was assessed using the coefficient of determination (R²). The model demonstrated satisfactory performance during the rainy season (September–October), with R² values ranging between 0.85 and 0.89. However, performance declined during the dry season (March–August), with substantially lower R² values.

The study provides valuable insights by applying remote sensing data in the Mediterranean Basin, a region highly sensitive to climate change. Nevertheless, there are several aspects where the study could be further strengthened:

1. As noted, while the model performs well during the rainy season, its reduced performance in the dry season could be addressed. The dry period is critical for sustainable water management, and improving model performance during this period would enhance its practical applicability.
2. The Conclusions section emphasizes the importance of local calibration for enhancing model accuracy under varying environmental conditions. Further elaboration on the calibration process would help clarify the approach taken and its effectiveness.
3. In the current calibration, some coefficients (Equation 7) were adopted from studies conducted in regions with different climatic conditions, while others were derived from local data. Providing a rationale for this choice and considering local calibration for all model parameters would strengthen the study.
4. The calibration process relied primarily on R² as the performance metric. Including additional metrics such as RMSE, MBE, MAE, or NSE could provide a more comprehensive evaluation of model performance.
5. The study applied a simple regression approach for calibration. Exploring alternative calibration techniques and conducting a comparative analysis may further improve model accuracy and robustness.
6. Addressing the lower performance during the dry season is important for enhancing the model’s operational reliability. If improvements remain limited after implementing the suggested adjustments, discussing the potential causes in the Discussion section would be beneficial.
7. The authors may benefit from examining recent studies where NDVI and LST values were successfully integrated with local parameters. Incorporating insights from such studies could support methodological and calibration enhancements in the revised manuscript.
8. A careful language revision is also recommended to improve clarity and readability.

Author Response

Revised Cover Letter for Reviewers - Water (Manuscript ID: water-3885759)

Manuscript ID: water-3885759

Title: Evaluating a Simple Algorithm for Evapotranspiration Retrieval Energy Balance Model in Mediterranean Citrus Orchards

Dear Editor and Reviewers,

We would like to express our sincere gratitude to the Editor for the opportunity to revise and resubmit our manuscript. We also greatly appreciate the constructive and insightful comments provided by the anonymous reviewers, which have helped us to substantially improve the quality, clarity, and robustness of the paper.

For transparency, we have reproduced the editor's and reviewers’ comments in black, followed by our detailed point-by-point responses in blue.

Editor:

Comment: Several figures are not clear enough.

Response: Thank you. Figures  5, 6, and 7 have been updated to improve their readability.

Reviewer 1: Comments and Suggestions for Authors

This study evaluates the performance of the Simple Algorithm for Evapotranspiration Retrieving (SAFER) in a 0.7-ha citrus orchard in Sicily, Italy, using remote sensing and eddy covariance data. The study employed satellite imagery without thermal bands, including Landsat 8, Sentinel-2, and MODIS. Both mass and surface energy balance (SEB) fluxes were processed using the SAFER algorithm.

The spatial resolution used in the study was 10 m/pixel, and remote sensing data were modeled through the R package proposed by Teixeira et al. (2021). Model calibration was conducted using regression analysis to account for local conditions. While the coefficients c and d provided in Equation 7 were adopted from Teixeira et al. (2021), the coefficients a and b in Equation 9 were derived through calibration with local data in the present study.

The model was applied for the years 2021–2022, and its performance was assessed using the coefficient of determination (R²). The model demonstrated satisfactory performance during the rainy season (September–October), with R² values ranging between 0.85 and 0.89. However, performance declined during the dry season (March–August), with substantially lower R² values.

The study provides valuable insights by applying remote sensing data in the Mediterranean Basin, a region susceptible to climate change. Nevertheless, there are several aspects where the study could be further strengthened:

We thank the reviewer for recognizing the value of our study. In line with the suggestions, we have implemented the following revisions:

1. As noted, while the model performs well during the rainy season, its reduced performance in the dry season could be addressed. The dry period is critical for sustainable water management, and improving model performance during this period would enhance its practical applicability.

We agree that improving model reliability during the dry season is crucial. A new paragraph has been added to the Discussion section, where we address the causes of reduced performance, including a brief explanation of the Bowen ratio. We also outline potential strategies for integrating additional parameters in future applications (lines 490-499). 

1. The Conclusions section highlights the significance of local calibration in improving model accuracy across diverse environmental conditions. Further elaboration on the calibration process would help clarify the approach taken and its effectiveness.

Section 2.4 has been expanded to provide further details on the calibration procedure, including data sources, preprocessing steps for EC and satellite data, spatiotemporal co-location, and the log-linear regression used to estimate coefficients a and b in Eq. 9 (lines 270-290).

1.  In the current calibration, some coefficients (Equation 7) were adopted from studies conducted in regions with different climatic conditions, while others were derived from local data. Providing a rationale for this choice and considering local calibration for all model parameters would strengthen the study.

As noted, coefficients c and d from Eq. 7 were adopted from Teixeira et al. (2021), while a and b were calibrated locally. This choice was dictated by the Agriwater 1.0.2 R package, where only a and b are externally modifiable. 

“Agriwater 1.0.2 R package, specifically the radiation_s2 function, which has as inputs the values of (doy= Day of the year, RG= Global Radiation, Ta= Air temperature, ET0=Reference Evapotranspiration, regression parameters of a and b) and the satellite images Sentinel 2 in the R, G, B, IR bands. This is why we can only calibrate parameters a and b, as the others are internal to the package and cannot be modified. The calibration of satellite images in their respective bands enables the calculation of albedo, NDVI, and LST”.

Additional explanatory sentences have been added (lines 241-250). 

1. The calibration process relied primarily on R² as the performance metric. Including additional metrics such as RMSE, MBE, MAE, or NSE could provide a more comprehensive evaluation of model performance.

 

We understand the importance of including these metrics, but we would need more time to cover all points, as the results obtained using the parameters proposed for the conditions by the original author were not optimal because we calibrate.Please refer to lines 302-314.The calibration was performed using standard statistical indicators, as detailed in various articles. We have included an additional description of these results in the revised manuscript 559-570. While we acknowledge the importance of incorporating further metrics, a comprehensive inclusion would require more time. This is due to the fact that the initial parameters proposed by the original author were not optimal for our conditions, necessitating calibration.

1. The study applied a simple regression approach for calibration. Exploring alternative calibration techniques and conducting comparative analysis may further improve model accuracy and robustness.

As mentioned above, we used the Agriwater package in this study, which has a pre-fixed number of input parameters. Among them, we can adapt only coefficients a and b to the local conditions.  Due to package constraints, only a and b can be adapted to local conditions. Future work will focus on modifying the model code to allow the calibration of additional parameters, as discussed in the revised manuscript (lines 505-514).

1. Addressing the lower performance during the dry season is essential for enhancing the model’s operational reliability. If improvements remain limited after implementing the suggested adjustments, discussing the potential causes in the Discussion section would be beneficial.

The Discussion now explicitly analyzes the physical causes of underperformance during dry periods (NDVI saturation/insensitivity to stress, canopy-level shifts in energy partitioning diagnosed via the Bowen ratio, and extreme atmospheric demand). Practical implications and avenues for improvement are also discussed (lines 606-615).

1. The authors may benefit from examining recent studies where NDVI and LST values were successfully integrated with local parameters. Incorporating insights from such studies could support methodological and calibration enhancements in the revised manuscript.

Integration of NDVI and LST with local parameters:

Recent investigations on the NDVI-LST relationship have primarily focused on local and regional scales, whereas global analyses remain outdated and fail to account for recent environmental changes induced by climate change and human activity (see, for example, https://www.mdpi.com/2076-3298/12/2/67). In this study, NDVI (Eq. 8) and LST(Eq. 3) were derived from satellite imagery and local parameters and subsequently incorporated into Equation 9:

 ETfr =expa+b(LSTa0NDVI)  

Where the coefficients a and b were estimated using meteorological data from the study area. This approach follows methodologies applied in previous works, such as  Safre (2022) (https://doi.org/10.1007/s00271-022-00810-1), where different values of a and b were calibrated for each study site. The use of locally calibrated parameters enhances the accuracy of ET estimates and provides a more robust representation of the role of agroecosystems in regulating land surface temperature. This, in turn, yields valuable insights to support conservation and adaptation strategies under climate change. A clarifying sentence to this effect has been added at lines 253-250 of the revised manuscript.

1. A careful language revision is also recommended to improve clarity and readability.

Finally, the entire manuscript underwent a careful language review to improve clarity and readability.

 

We believe these revisions have significantly improved the manuscript, addressing all concerns raised by the editor and reviewers. We once again thank the reviewers and editor for their valuable feedback and trust that the revised version meets the standards of Water.

Sincerely Author.

Reviewer 2 Report

Comments and Suggestions for Authors

Evaluating a Simple Algorithm for Evapotranspiration Retrieval Energy Balance Model in Mediterranean Citrus Orchards

The authors are commended for their excellent effort in documenting the previous literature, as well as their research methodology and findings.  The paper is well written.  However, they are promoting the adoption of the method despite several severe limitations.  For example, lack of agreement with EC data during dry periods, transferability issues (e.g., empirical coefficients a and b, which may require calibration), and the complexity of applying the technique. 

They suggest that the method can be used to provide irrigation recommendations; however, it does not seem practical to use this method for scheduling irrigation.  Other methods are available which are much easier to use and probably provide equally accurate results. This tool is perhaps better used as a research tool. 

The future work suggested at the end of the paper is good.

Figure 5.  Shows seasonal LE and H results obtained with the EC technique.  Net radiation should be included in the graph so that the reader can evaluate the energy balance. If the SAFER model produces LE and H, it would be good to include those results in Figure 5 for comparison. 

Figure 6.  The regression results, referred to in the text, should be shown in a figure.  It does not appear that the wet season is significantly more correlated than the dry season. Showing the regression results in a figure would help to confirm the statement in the text. 

"Integrating indicators of plant water status, such as stem water potential or stomatal conductance, as suggested by Anderson et al. (2007), could improve model responsiveness." Ortega-Farias et al. (2008) provide a method for increasing surface resistance under dry soil conditions. FAO 56 (Allen et al., 1998) provides a crop stress factor, which decreases linearly after soil water deficits are greater than the Readily Available Water (RAW), to reduce the Crop Evapotranspiration under well water conditions.

Author Response

Revised Cover Letter for Reviewers - Water (Manuscript ID: water-3885759)

Manuscript ID: water-3885759

Title: Evaluating a Simple Algorithm for Evapotranspiration Retrieval Energy Balance Model in Mediterranean Citrus Orchards

Dear Editor and Reviewers,

We would like to express our sincere gratitude to the Editor for the opportunity to revise and resubmit our manuscript. We also greatly appreciate the constructive and insightful comments provided by the anonymous reviewers, which have helped us to substantially improve the quality, clarity, and robustness of the paper.

For transparency, we have reproduced the editor's and reviewers’ comments in black, followed by our detailed point-by-point responses in blue.

Editor:

Comment: Several figures are not clear enough.

Response: Thank you. Figures  5, 6, and 7 have been updated to improve their readability.

 

Reviewer 2: Comments and Suggestions for Authors

The authors are commended for their excellent effort in documenting the previous literature, as well as their research methodology and findings.  The paper is well written.  However, they are promoting the adoption of the method despite several severe limitations.  For example, lack of agreement with EC data during dry periods, transferability issues (e.g., empirical coefficients a and b, which may require calibration), and the complexity of applying the technique. 

 

We sincerely thank the reviewer for the positive assessment of our work and for the thoughtful suggestions. We have revised the manuscript to address all concerns: 

1. They suggest that the method can be used to provide irrigation recommendations; however, it does not seem practical to use this method for scheduling irrigation.  Other methods are available, which are much easier to use and probably provide equally accurate results. This tool is perhaps better used as a research tool. 

 

We agree that SAFER has limitations—particularly in dry-season conditions and regarding coefficient transferability and operational complexity. The revised Discussion and Conclusions now explicitly state that SAFER is best suited as a research/monitoring/support tool, providing spatial ETa to complement, not replace, growers’ simpler operational scheduling methods. Please refer to the conclusion and part of the revised version of the manuscript.

 

1. Figure 5.  Shows seasonal LE and H results obtained with the EC technique.  Net radiation should be included in the graph so that the reader can evaluate the energy balance. If the SAFER model produces LE and H, it would be good to include those results in Figure 5 for comparison. 

Figure 5 was revised to include: net radiation (Rn), sensible heat flux (H), latent heat flux (LE) and ground heat flux (G) from EC measurements, allowing a complete and energy-balance-consistent comparison. 

1. Figure 6.  The regression results, referred to in the text, should be shown in a figure.  It does not appear that the wet season is significantly more correlated than the dry season. Showing the regression results in a figure would help to confirm the statement in the text. 

We also added Bowen ratio plots (Figure 6) to demonstrate seasonal differences in model performance.

1. "Integrating indicators of plant water status, such as stem water potential or stomatal conductance, as suggested by Anderson et al. (2007), could improve model responsiveness." Ortega-Farias et al. (2008) provide a method for increasing surface resistance under dry soil conditions. FAO 56 (Allen et al., 1998) provides a crop stress factor that decreases linearly after soil water deficits exceed the Readily Available Water (RAW), thereby reducing crop evapotranspiration under well-watered conditions.

To strengthen the discussion on crop stress and water deficit, we incorporated references to Ortega-Farias et al. (2008) and FAO-56 (Allen et al., 1998). These contributions emphasize the importance of accounting for surface resistance and stress coefficients under water-limited conditions.  

1. The future work suggested at the end of the paper is good.

We appreciate the reviewer’s positive remarks on the future work proposed, and we have maintained this perspective in the revised version. 

We believe these revisions have significantly improved the manuscript, addressing all concerns raised by the editor and reviewers. We once again thank the reviewers and editor for their valuable feedback and trust that the revised version meets the standards of Water.

Sincerely Author.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Editor,

The manuscript has been substantially improved by addressing most of the previous comments. I recommend adding recent references in the Discussion section on methods such as ANN and SVM, which have been shown to better capture the nonlinear nature of actual evapotranspiration compared to linear regression. This would further strengthen the paper and highlight its relevance for future research. In addition, “Table 3” should be corrected at line 492.

Author Response

We sincerely thank the reviewer for their positive assessment of the revised manuscript and for their valuable additional suggestions. We have addressed both points as described below.

Comment 1: "I recommend adding recent references in the Discussion section on methods such as ANN and SVM, which have been shown to better capture the nonlinear nature of actual evapotranspiration compared to linear regression. This would further strengthen the paper and highlight its relevance for future research."

 

Response:

We agree with the reviewer that this is an excellent point. To address this, we have added a new paragraph to the Discussion section that explores the limitations of our linear regression-based approach during the dry season and introduces advanced machine learning (ML) methods as a promising future direction. The new text cites recent, high-impact studies that demonstrate the superior performance of algorithms in capturing the non-linear dynamics of evapotranspiration, particularly under stress conditions. This addition significantly strengthens the paper's discussion and aligns it with current research trends.

 

The new paragraph has been inserted on Pages 17-18, Lines 647-669

 

Comment 2: "In addition, ‘Table 3’ should be corrected at line 492."

 

Response:

We thank the reviewer for catching this inconsistency. The reference to Table 3 at line 492 has been corrected. Furthermore, we have thoroughly reviewed the formatting and presentation of Table 3 itself to ensure clarity and consistency with the results discussed in the text.

 

The correction has been made on Page 15, Line 492

 

We believe these revisions have significantly improved the manuscript, addressing all concerns raised by the editor and reviewers. We once again thank the reviewers and editor for their valuable feedback and trust that the revised version meets the standards of Water.

Sincerely Author.

Author Response File: Author Response.pdf