Ecosystem Evapotranspiration Partitioning and Its Spatial–Temporal Variation Based on Eddy Covariance Observation and Machine Learning Method
, Sha Zhang 1, Yun Bai 1 and Shanshan Yang 1,*Round 1
Reviewer 1 Report
The manuscript “Ecosystem evapotranspiration partitioning and its spatial-temporal variation based on eddy covariance observation and machine learning method” by Lu et al. devoted to development of an objective and data driven method for estimation of vegetation transpiration and soil evaporation.
The paper introduction is informative and contains a review of the existing methods their limitations and aims of the paper. The methods described in details, with listing and explanation of used data and approaches. The obtained results are clear and supported by colorful figures. Discussion and conclusions are sufficient and reasonable.
Specific comments
Authors present a processing of ecosystem evapotranspiration data from FLUXNET sites. According to definition, evaporation or transpiration is a flux of water vapor. Actually the manuscript analyses a latent heat fluxes related with evapotranspiration, because the dimensions of studied variables is [W/m2] but not [mm or mg of water]. It is clear that these variables are functionally related. Please add a comment about the actual studied variable and relation with evapotranspiration. Use [W/m2] instead of [w/m2].
Figure 3 show the importance of model features. What is the meaning of “importance” parameter, gow it was calculated? Add an explanation.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Lu et al. present a machine learning model which can improve the accuracy of soil evaporation estimation. This is interesting and valuable. However, the English writing should be improved. Precise and clear writing is important and sufficient to report the new findings to our science community. Specially, for the introduction, the novelty of this study should be stronger thorough summarize some main pointes of their achievements and undoes, not only just using several sentences to state the same question. I recommend this manuscript needs between moderate and major revision before it can be published. My main comments are as follows:
General comments:
(1) In the Introduction section, there are some irrelevant sentences for the background. And the innovation of this study should be stated better.
(2) There are many sites in Chinaflux, why don’t you choses these sites?
(3) The writing of many sentences are not native, for example, figure x illustrates, shows, etc.
Minor comments:
L29-30: This is confusing that ‘the ET partitioning method’ or ‘less assumption or prior knowledge’ make it easy to apply across different ecosystems and spatial scales?
L44-46: this sentence should state the importance of Traditional measurement techniques and then present their shortcomings
L51: what is the meaning of ‘Another limitation of the ET partitioning’?
L60: revised as: applications.
L189: this subtitle should be ‘soil moisture data’.
L317-318: please rewrite it.
the English writing should be improved. Precise and clear writing is important and sufficient to report the new findings to our science community.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Evapotranspiration decomposition has been a major challenge in the research field for a long time. Process-based models have been used to simulate it in the past, however this manuscript presents a novel approach which is deemed to be highly beneficial. However, there are several issues that need to be addressed before publishing this manuscript.
(1) Please provide information on each flux tower in the appendix, including latitude and longitude, coverage time, name, and IGBP type, in order to reproduce the results.
(2) The analysis in section 3.4 is problematic. Using a simple single factor and T/ET for regression is not sufficient and does not indicate the problem. The variance factor decomposition method needs to be used to quantify the contribution of LAI and VPD to T/ET. Additionally, a significance test needs to be provided.
(3) Please provide information on the T/ET of each site for future research to compare with your study.
(4) Please test different machine learning algorithms in the discussion to see if they bring significant uncertainty to T/ET.
(5) How does the vegetation index change from a daily scale to a half hour scale.
(6) Has the issue of energy closure of ET data been considered and corrected?
(7) Could you please discuss the possible application of this method in global scale T/ET mapping.
(8) Can this method be combined with the process model (e.g. PMLv2, PM) to further improve the accuracy of ET and T/ET? Please discuss it.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The manuscript has been revised point by point. And it can be published after the English language improvement.
I suggest that the authors shold spend much time for language improvement.
Reviewer 3 Report
This work is good, but how to integrate it with process models in future work still requires more careful consideration.
