Diagnosis of Atmospheric Drivers of High-Latitude Evapotranspiration Using Structural Equation Modeling
Round 1
Reviewer 1 Report
MS:
Diagnosis of drivers of high-latitude evapotranspiration using structural equation modeling
-----------------------------------------------------------------------------
The article is interesting in its content and may attract the attention of hydrologists and specialists in applied climatology. However, it can be improved since some inconsistencies in its construction, repetitive phrases, and unnecessary sentences are detected. [...] See the attached pdf file.
Observations and suggestions. –
Page 1
Abstract. -
It says:
Abstract: Evapotranspiration (ET) is a major component of the surface moisture budget, and variations over daily to interannual timescales have inter-related drivers. This study uses structural equation modeling to diagnose the drivers over an ensemble of 45 high-latitude sites, each providing at least several years of in situ measurements, including ET fluxes derived from flux tower sites.
Suggestion:
Abstract: Evapotranspiration (ET) is a relevant component of the surface moisture budget associated with different drivers. Their interrelation causes variations from daily to interannual timescales. This study uses structural equation modeling to diagnose the drivers over an ensemble of 45 high-latitude sites, providing each at least several years of in situ measurements, including ET fluxes derived from flux tower sites.
Lines 42-46. -
It says:
The present paper is an evaluation of the drivers of evapotranspiration in the Arctic 42 and subarctic. The selection of variables for inclusion in the assessment of drivers builds 43 on several other previous studies of the surface moisture budget in northern terrestrial 44 regions. It is well known that ET increases as relative humidity decreases, and this relationship is captured by the bulk transfer formulations used in many climate models….
Suggestion:
The present paper is an evaluation of the drivers of evapotranspiration in the Arctic 42 and subarctic. The selection of variables for inclusion in the assessment of drivers builds 43 on several other previous studies of the surface moisture budget in northern terrestrial 44 regions. As it is known, ET increases as relative humidity decreases, and this relationship is integrated into the bulk transfer formulations used in many climate models….
Page 2
Lines 62-67. –
It says:
Eddy covariance measurements of the latent heat flux were used by Thunberg et al. (2021) to evaluate variations of ET at individual locations, with a focus on Alaska. The mean seasonal cycle and interannual variations were the foci of this previous study, which showed that the annual net P-ET at the high-latitude sites is generally positive but can vary interannually by an order of magnitude. The surface moisture budget at boreal forest sites underlain by permafrost showed a moisture deficit in the early summer, then a moisture surplus for the remainder of summer through late autumn.
Suggestion:
Eddy covariance measurements of the latent heat flux were used by Thunberg et al. (2021) to evaluate variations of ET at individual locations, with a focus on Alaska; in this study, the mean seasonal cycle and the interannual variations were the focus. The authors stated that annual net P-ET at the high-latitude sites is generally positive but varying interannually within an order of magnitude. The surface moisture budget at boreal forest sites underlain by permafrost showed a moisture deficit in the early summer, then a moisture surplus for the remainder of summer through late autumn.
Lines 76-80. -
It says:
The present paper is a more comprehensive assessment of the drivers of variations of ET over daily, weekly and monthly timescales. Its main objectives are to (1) document the interrelationships among the variables affecting ET over these timescales and (2) distinguish the ET regimes and dependencies based on classification of the sites according to vegetation type and the presence or absence of permafrost.
Suggestion:
The present paper is a comprehensive assessment of the drivers of variations of ET over daily, weekly and monthly timescales. Its main objectives are: (1) to document the interrelationships among the variables affecting ET over these timescales and (2) to distinguish the ET regimes and dependencies based on classification of the sites according to vegetation type and the presence or absence of permafrost.
Lines 89-93. –
It says:
The present analysis focuses on determining meteorological drivers of ET variations, comparing these relationships between ranges of vegetation and permafrost classification. As described in Section 1, several key variables have been well-documented to have impacts on ET. This study focuses on comparing the differences of importance between these previously identified variables. Table 1 shows the full list of variables used in the present analysis.
Suggestion:
The analysis focuses on determining meteorological drivers of ET variations, comparing these relationships between ranges of vegetation and permafrost classification. As described in Section 1, several key variables have been well-documented to have impacts on ET. This study focuses on comparing the differences of importance between these previously identified variables. Table 1 shows the list of variables used in the present analysis.
Page 3
Table 1. – see the pdf file.
...on Line 29 …. you defined evapotranspiration as ET (not precisely Latent heat flux)
Comment:
For readers who are specialists in hydrology, this is understandable, but for other readers, it may cause confusion... it is suggested to explain how the concept 'Latent heat flux' is interpreted or translated into units of evapotranspiration.
You can use: the following formula for converting watt ( energy ) in to depth in mm of water or ET
1 Watt /m2 = 0.0864 MJ /m2/day
1 MJ /m2/day = 0.408 mm /day .
Please follow the above calculation procedure. Refer FAO Irrigation and Drainage 56 , page no. 212…. Or consult other sources.
Page 6
Lines 115-122. -
It says
Following the data processing procedure used by Thunberg et al. (2021), these data were processed into aggregate daily, weekly, and monthly datasets covering the warm season (May through September) for each station and adjusted for missing data. Not all stations measured every variable. For stations missing precipitation data, the ERA5 land hourly reanalysis (available at https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanal-ysis-era5-land?tab=overview) was used to fill in the missing data. To create the most complete datasets for the structural equation models described in Section 2.2, each variable required at least 75% of data present to be included in the site’s model run.
Suggestion:
We processed data into aggregate daily, weekly, and monthly datasets covering the warm season (May through September) for each station and adjusted for missing data, according to Thunberg et al. (2021). Not all stations measured every variable. For those stations missing precipitation data, we applied the ERA5 land hourly reanalyses for filling the missing data (available at https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanal-ysis-era5-land?tab=overview). To create complete datasets for the structural equation models described in Section 2.2, each variable required at least 75% of data present to be included in the site’s model run.
Page 7
Lines 160-171. -
It says:
To understand how each variable contributes unique information to variability in ET, a path analysis was used. This SEM is commonly used in comparing direct effects of meteorological variables onto ET. Zhang et al. (2015) used a path analysis in a similar application to compare the effects of net radiation, air temperature, vapor pressure deficit, and wind speed on ET. The UCLA: Statistical Consulting Group also describes the path analysis and similar models in depth (https://stats.idre.ucla.edu/r/seminars/rsem/). A path 165 analysis is a specific type of SEM which uses a set of exogenous variables (variance is independent of other variables) to predict endogenous variables (variance is dependent on other variables) while allowing the variables to predict each other in the process. This analysis used the R lavaan package SEM function to define and run the model to predict ET using precipitation, temperature, relative humidity, wind speed, sensible heat flux, ground heat flux, net shortwave, and net longwave radiation.
Suggestion:
A path analysis was used to understand how the variables contribute with unique information to the variability of ET. This SEM is used commonly in comparing the direct effects of meteorological variables onto ET. Zhang et al. (2015) used a path analysis in a similar application to compare the effects of net radiation, air temperature, vapor pressure deficit, and wind speed on ET. The UCLA: Statistical Consulting Group also describes the path analysis and similar models in depth (https://stats.idre.ucla.edu/r/seminars/rsem/). Path analysis is a specific type of SEM, which uses a set of exogenous variables (variance is independent of other variables) to predict endogenous variables (variance is dependent on other variables) while allowing the variables to predict each other in the process. This analysis used the R-Lavaan package SEM function to define and run the model to predict ET using precipitation, temperature, relative humidity, wind speed, sensible heat flux, ground heat flux, net shortwave, and net longwave radiation.
Lines 192-196. –
This paragraph is unnecessary; I suggest to delete it:
Previous studies cited in Section 1 have indicated that variations of ET are associated with variations of various atmospheric variables: air temperature, relative humidity precipitation, wind speed, and shortwave and longwave radiation. ET is also affected by the partitioning of the available surface energy into the sensible heat to/from the atmosphere, the heat flux into/from the ground, and evapotranspiration.
Pages 12-13
Lines 270-289. –
Please simplify this paragraph (3.2. Path Analysis):
The paragraph contains too many statistical descriptors generated by principal component analysis, which makes it difficult to read. It is understood that there should be numerical correlations between the components of the method, because of their own statistical profile, but it would be more useful to focus mainly on the physical correlations of the phenomenon. Please eliminate statistical parameters that are not strictly necessary for the explanation in order to facilitate reading.
3.2. Path Analysis
While a factor analysis quantifies relationships among a complex set of intercorrelated variables, correlations deduced from factor loadings do not provide measures of correlation with a variable after removal of the effects of other inter-correlated variables. Therefore, we supplement the factor analysis with a path analysis to quantify the direct dependencies of ET on each variable independent of all other variables. Following the approach used by Zhang et al. (2015), the distinction between the direct and indirect dependencies is illustrated in Figure 7, which shows the associations between the different variables in terms of physical pathways and corresponding correlations between pairs of variables. The correlations are based on the daily values (Figure 2a). Consistent with the factor analysis, Figure 7 shows that net solar radiation correlates most highly with ET. However, as shown in the left portion of each diagram, solar radiation drives ET through 281 its effect on temperature via the flux of sensible heat from the ground surface. As parameterized in many models, the sensible heat flux is proportional to the difference between 283 the ground (skin) temperature and the air temperature. Air temperature, in turn, affects ET through its effects on relative humidity and downwelling longwave radiation. ET is also correlated negatively with the net longwave radiation reaching the surface and positively with the ground heat flux. Correlations of ET with precipitation and wind speed 287 are weak, although the correlation with wind speed increases when the unique contributions of each variable are evaluated as described below.
Page 23. –
This paragraph is unnecessary; I suggest to delete it:
Lines 549-553:
Remote sensing products (e.g., Suzuki et al., 2018) can serve as a bridge between in situ measurements and climate model results (e.g., Wang et al., 2015), although the finite pixel sizes of remote sensing products also present challenges to the upscaling of the types of measurements used in this study.
Comments for author File: 
Comments.pdf
Author Response
See attached file for point-by-point responses.
Author Response File: Author Response.pdf
Reviewer 2 Report
Review of "Diagnosis of drivers of high-latitude evapotranspiration using structural equation modeling", by Thunberg et al. This work is an assessment of the drivers of evapotranspiration (ET) in 45 high-latitude locations at daily, weekly and monthly timescales using two structural equation modeling (SEM) approaches: the factor and the path analyses. My overall impression is that the work is novel and it deserves to be published. However, there are many aspects that must be improved to achieve the quality of the papers published in Atmosphere.
Major comments: - The introduction has to be revised to include more work that is related to the present study. The literature review is very brief. Moreover, the novely of the work must be highlighted. In the literature review, the authors should tell the reader what has been done to analyze ET with SEM, and what is the new aspect that is incorporated in the current study. - Even when the authors refer the reader to the work of others for the theory of SEM, I think the paper requires to have a better description of the methods used. I do not think they need to incoporate a detailed description of the factor analysis and of the path analysis, but at least they should show the main equations that are used to apply theses methods. - The selection of the different groups of variables analyzed is poorly supported. Please see comments in the attached pdf file. - There are two relevant variables that were not included in the analysis, and there is consensus on their effect on ET: soil moisture and vapor pressure deficit. Since data from Fluxnet and AmeriFlux typically include these variables, I do not understand why they were not included in the analysis. I understand that soil moisture could not be available in continuous permafrost sites, but it should be in discontinuous or non-permafrost sites. These must be explained in the text. - The sample size for some of the results are very small (e.g., see Figure 10 e and f). How can we achieve good conclusions having a sample size of less than 10? (even 1 or 2). - In the discussions, the authors should explain the importance of their investigation with regards to other studies.
Minor comments: - The format used in the citations is not the one requested by Atmosphere. Hence, the format of the citations and references must be fixed. - Some links to websites are not working. - The cross-reference to section numbers is not correct in all places. Please revise the document to improve this issue. - The attached document has many comments that must be addressed by the authors. Based on the previous comments, I think the paper is not suitable in its current form, and that major revisions must be performed to achieve the standard of Atmophere. I urge the authors to address the comments of all the reviewers to improve their work.
Comments for author File: Comments.pdf
Author Response
See attached file for point-by-point responses.
Author Response File: Author Response.pdf
Reviewer 3 Report
This paper collects reliable data from 45 sites and compares the differences between those sites using a statistical analysis modeling method called SEM.
It is clear and wonderful that this analysis reveals the structure of the mutual influence of complex meteorological elements. Furthermore, we are examining their mutual effects by changing the time scale. This is a rational analysis policy based on the basis of meteorology that all meteorological phenomena change the main factors depending on the time scale.
The explanation about the analysis result is rational. There is no contradiction in the conclusions drawn from those results. In particular, the difference in the primary contributors to ET between tundra and forests is a very interesting result and provides new insights for future comparative hydrological studies in this area.
This article is one of well-planned comparative hydrological studies. And it will provide new insights for water cycles of vegetation in the Artic and Subarctic region.
Author Response
We thank the reviewer for the positive comments. In the absence of requested revisions from Reviewer 3, our revisions are responses to Reviewers 1 and 2 only.
Round 2
Reviewer 2 Report
Second review of "Diagnosis of atmospheric drivers of high-latitude evapotranspiration using structural equation modeling", by Thunberg et al.
In general, the authors have addressed in a good way the comments provided by three reviewers, which is something that this reviewer appreciate very much. This has improved the presentation of this work.
After reading the new version of this manucript, I have some minor comments, which are attached in the pdf file, that must be addressed before this manuscript can be published.
First, I think the new introduction does not follow a logical sequence and this is detrimental to the manuscript. In the pdf file, I have provided a suggestion of how to order the different ideas to improve it (e.g., right now there is no mention of SEM, which surely will confuse a reader that reads this document for the first time).
Also, I understand that the authors added equations (1) and (2) because of the comment of another reviewer. In my humble opinion this simply does not work. Adding two equations to show how you can change units between LE and ET is surely a waste of space. To improve this, I have added a suggestion that I think it will work much better. I hope this helps.
Please note that my comments have the aim of improving the presentation of the work, which is something the authors will appreciate as we all want to publish high-quality and polished manuscript, so my suggestions go on that way.
In the attached pdf file I have a few more comments, but I am sure that all of them will require a really quick revision from the authors. So my recommendation is to accept after minor revisions.
Comments for author File: Comments.pdf
Author Response
File attached.
Author Response File: Author Response.pdf
