Review of “Measuring evapotranspiration on an eroded cropland by an automated and mobile chamber system: gap filling strategies and impact of soil type and topsoil removal” by Adrian Dahlmann et al., for HESS.

Summary

This manuscript presents a study of roughly one year of evapotranspiration measurements of a winter rye crop on a number of soil types. The ET measurements are novel, using an automated and mobile chamber system. The study reports and highlights the negative impact of eroded conditions on biomass growth and water use efficiency. Because of the novelty of the ET method, the study also investigates different gap-filling techniques. Here they find that machine learning approaches are better than simpler regression or look-up techniques. While this is an interesting study, I think there are many improvements that can be made to give it more relevance to both application users (farmers, land managers) and to the scientific community. 

Thank you for your time and effort in reviewing our manuscript. We appreciate your constructive comments and have taken all of them into consideration. A detailed reply is posted below each comment in italics for better visibility.

Major comments

1. The focus of the paper is not clear and wanders between the new ET method and its gap-filling, the erosion gradient of soils, the role of ET in the water balance, and other impacts of erosion (e.g., on biomass production). A tighter, clearer focus will help – and then the methods parts can better be written to support the key aim, which is (to me at least) to explore the impact of soil conditions on agronomic and water balance impacts.

2. Hypothesis-testing should be performed – it should be clearer both in the abstract and end of the introduction what the expectation is – how should the different soils change agronomic and ET outcomes? How should we be able to anticipate, from the literature, different performance among the gap-filling techniques?

The aim of our manuscript was indeed twofold: 1. To describe our new set up and test it with a particular focus on gap filling strategies and 2. To shed light on the impact of soil erosion and soil management on ET flux dynamics and cumulative ET. 

We fully agree that in the previous version these two aims (and corresponding hypotheses) might not have been clearly phrased. Consequently, in addition to rephrasing the aims and adding corresponding hypotheses to each aim, we will carefully revise the entire manuscript to stick strictly to the structure we define with our aims and hypotheses.

3. In general, there are many mistakes in the English language related to article use, comma positions, sentence structure, possessive apostrophe, etc. I have noted some of these below, but a careful edit should be performed with these issues in mind.

Thank you for your suggestion. The revised manuscript will be carefully read and spell corrected by a native speaking colleague.

Minor comments

The intro could be condensed, particularly the first paragraph (e.g., L36-40, 47-52, and remove “according to the European Union”).

Done.

-     Removed first sentence.

-     Removed second part of second sentence.

-     Removed “According to the European Union”

-     Connected two sentences about soil degradation

Consider moving L53-62 to the site description section in the methods.

Done.

L77-81, Remove/condense; consider “field” instead of plot for eddy covarnacie

Done.

L103-5 consider at minimum three aims (i.e., separate soil type and management from spatio-temporal variability). In general note other comments that there should be clearer hypotheses in this area about expectations related to soil/management type, seasonal development of ET and WUE, and the gap-filling methods

We fully agree that in the previous version our aims (and corresponding hypotheses) might not have been clearly phrased. Consequently, in addition to rephrasing the aims and adding corresponding hypotheses to each aim, we will carefully revise the entire manuscript to stick strictly to the structure we define with our aims and hypotheses.

L141 what kind of digestate?

We will include a table with digestate information in the appendix and refer to it in this section.

Section 2.3 misses info on measurement times, accuracy, precision, etc. of this novel system

We agree that some information is missing or mentioned in the wrong place. All information concerning the system (mainly from chapter "2.5.1 ET Flux calculation") and other missing information will be incorporated in the revised manuscript.

L157 describe the adjacent field a bit – given that the soils and management status can drastically change the soil moisture, how was the adjacent field treated? Also please add where this field is to the site map.

Sorry for the misleading description, it was not an adjacent field it was a plot on the same field. We have now marked the adjacent plot in Fig. 1 (< 25m away). We’ll add the location to the site map.

L180 some chamber info is missing – what is the headspace, how tall is it relative to the biomass)

See above comment on section 2.3. The maximum plant height was 1.35m to 1.65m depending on the treatment. At a chamber height of 1.94m there was a minimum headspace of 29 cm.

L177 consider moving all or some of this on flux calculations and gap-filling to section 2.3

Done.

L214 clarify if there is a moving window to this NLR or it’s just clumping all the data – if so – how about trying a moving window or adding a term like days after planting (though RVI may be sufficient, it’s not clear – were regressions of the residuals tested?).

Indeed initially a variable moving window approach (joining 2 to up to 21 consecutive measurement days) was implemented for the NLR. However, due to limited amount of data, this approach performed less well than the NLR for the entire crop season (including RVI as parameter to account for crop growth).

L295 “differ between” and “minor differences” – are statistics performed here? Perhaps also some ranges of values can be given in the text

This sentence is misleading: we were trying to describe the general trend of the measured fluxes. This section will be rewritten in the revised version also adding some value ranges.

L297 it’s not clear what magnitude is being described

We changed it for better understanding. It now reads “After the non-growing season, fluxes quickly return to their maximum fall level and then rapidly increased.” 

L302 start with the result as a topic sentence, rather than the exposition. The result and not “figure 4” should lead the paragraph.

We changed the formulation. It now reads: “Pronounced differences of tested gap-filling approaches in terms of respective calibration statistics could be found. Calibration model-performance differ in their scatter around the 1:1-agreement plots (Fig. 4) and associated coefficients of determination (r2). ”

L308 the allocation problems could be described / tested? (also “one can quickly see…” can be edited)

We will add a table of different performance classes in the methods section (2.7), including ranges for the statistical parameters that we used to define the quality of gap-filling approaches. We also changed the wording. It now reads: “Considering the k-fold-cross validation (Fig. 5, table 2), ANN_Br and SVM perform well according to our defined performance criteria, while LUT is not suitable due to allocation problems that arise when no class is found for the given climate conditions and the mean is used.”

L310 “a large number of negative ET fluxes” – first, how many? Second, is there evidence of dew? A negative ET isn’t so implausible

It's true that negative ET could be possible, but we never measured negative ET fluxes over the entire growing season. Negative ET occurred only for gap-filled/modelled fluxes using ANN-Br in large amounts (35.5% for an example plot).This amount of negative fluxes, coupled with a magnitude of up to -6 mm/d, cannot be explained by dew or uptake via leaves. It is much more likely that this magnitude is an extrapolation problem of ANN-Br. We clarify this in the revised version.

L318-320 could be put at the top of the paragraph

Done.

L321 this section could be merged with the previous – consider the treatment effect and then the drivers

Done.

L430 “to the amount of data” – how much? – you have a lot right? What would be sufficient?

This statement refers to the general fact that, also according to cited literature, machine learning algorithms such as SVM and ANN perform better with larger datasets. While we cannot quantify “sufficient data” for the failing ANN analysis, we can do so for the indeed working SVM through stepwise k-fold-subsampling and will include the results within the revised manuscript version.

L432 this section is very long- be more concise, move parts that cite other work to other parts of the discussion (or a new section there) – moreover, this section starts with a description of gap-filling – is that the main finding? If so the paper should be reworked so it’s the dominant research question and the rest is a case study to test it. I’d tend to the think the focus should be on soils, management, and the resultant ET.

Thank you for this comment. We will rewrite/restructure the conclusion.

Fig 1 what do the colors in plot c represent? Where are the soil moisture measurements and adjacent field as indicated in text?

The colors indicate the non-manipulated and manipulated plots of each soil type. We will include this information in the revised manuscript. See also reply above.

Fig 2 clarify that this is incoming par and not absorbed, reflected, etc.

Done.

Fig 3 change Okt to Oct, describe here or in the text where are values below zero? Indicate here or in the text whether these are already quality-screened and what those methods were, what is the estimated detection limit?

Figure 3 shows only the measured ET fluxes quality-screened by soft and hard criteria without any gap-filled data. Due to the fact that negative values only occurred during gap filling with ANN-Br (see in detail above), there are no negative values in this figure. This will be clarified in the revised manuscript.

Fig 6 some stats perhaps could help tell us if these curves or cumulations are different, significantly

We will add the mean cumulative seasonal ET and the standard deviation to the figure. 

Fig 7 what are the colors? 

We will add a legend to explain the colors.

Fig 8 is an nova possible here, which are specifically different from each other?

Results of statistical tests indicating difference between treatments will be added.

Fig A1 can real data be plotted here?

Yes, we can. We will add real measured weekly data.

Technical comments

L13 Add the before ongoing

Done.

L16 change ingoing to incoming (and throughout)

Done.

L17 the paper doesn’t really address the full water budget – perhaps main water budget term would be better?

That's right. We now call it "seasonal cumulative ET".

L92, 114 change build to built

Done.

L95 change the aim was to the paper’s aim is…

Done.

L102 reword “This enabled to asses”

Done.

L103-4 be consistent with WUE vs WUEagro

Done.

L114 the description of CarboZALF repeats the intro (and this sentence is too long)

Removed this sentence. Done.

L120 “organic fertilized” interrupts the flow; it can be omitted

Done.

L129 just topsoil (not topsoils) – and was not were

Done.

L135 weighed not weighted

Done.

L146 and elsewhere the “by” is not needed

Done.

L155 remove comma after both

Done.

L163 channel not canal

Done.

L166 write out three, clarify “10 seconds = 10 records”

Done. 

L174 perhaps replace “further called” “what we term the”

We changed the formulation. It now reads: “During the period from November 24, 2020 to March 22, 2021, which we refer to as the “non-growing season”,  no plant growth was assumed due to average daily temperatures below 5°C (<3 consecutive days).”

L208 change I to 1 (i.e., “eye” to “one”)

Done.

L208-210 consider re-organizing like: “a simple stat approach: (1) MDV), two empirical… (2, 3), and two machine learning..(4,5) for consistent structure

Done.

L258 add values after NSE

We will add a table with values for different performance statistics (NSE, NRMSE, MAE, etc.) in the revised manuscript and thereon after refer to those.

L259 remove “the” and the s in parameters

Done.

L269 change photosynthetic to photosynthetically

Done.

L274 remove “in the observed period” or otherwise reword

Done.

L275-6 to where does this downward trend go (Describe/quantify in text so one doesn’t have to look at the figure to determine it)

Done. 

L282 has a unique date style compared to the rest of the text

Done. 

L285 “nearly no differences” – perhaps negligible?

Done.

L287 “clearly”?

Removed. Done.

L327 remove s from variables

Done.

L330 remove used

Done.

L331 change offers to offered

Done.

L333 consider fewer rather than less

Done.

L335 remove comma

Done.

L337 reword – no need for “As well as” twice

Done.

L337 add of after suitability

Done.

L365 change was to were

Done.

L397 change were spanning over to spanned

Done.

L402 add t to constraints

Done.

L445 remove comma after demonstrate

Done.

The authors present evapotranspiration (ET), water use efficiency (WUE) and ET gap-filling strategies along a toposequence of erosion for one growing season of winter wheat in NE Germany. In addition, a comparison between topsoil addition/removal is included in the study. ET is measured with a novel automatic chamber system comprising large transparent cylindrical chambers which are lowered on a collar from above using a crane. The advantage of this system is the relatively large measurement area and less impact of the automatic chamber as compared to other chamber systems. The study can be recommended for publication, however, there are aspects which should be considered for revision. (1) a complete revision from a native English speaker would improve the readability of the manuscript (disclaimer: saying that without being a native English speaker myself). (2) The method section should include more information on how ET was measured. (3) The discussion and conclusion section would need complete revision; the prior needs more focus on the study, the latter is too long and should be condensed to the most important aspects. (4) the measured ET has not been compared to other ET methodologies; this discussion point should be further strengthened. See details below:

We greatly appreciate your constructive review of our manuscript. We took your comments into careful consideration and you will find detailed replies below each comment marked in italics for better readability.

In regard to your general suggestions: 

1. Thanks, a native speaking colleague will carefully read, spell check the revised version and advise us on writing style.
2. We agree that this has been kept rather short. We have taken this into account and expanded the Material and Methods section accordingly (see also specific comments below).
3. We again agree. In the revised version, the discussion section will be significantly changed with a stronger focus on our original aims and the hypotheses (that we have now included, see below) and including the required additional sections, e.g. on methodological comparability. The conclusions have been significantly shortened, focusing now on the major outcome of our study.
4. This is true. Within the field of research, we have seen an intensive discussion on the methodological constraints of different approaches to measure and model ET. We have briefly touched this subject already in the discussion (relating our results to previous findings of a nearby lysimeter study), but the revised version will include a much more detailed account on the subject, summarizing ET measurement methods and discussing their advantages and disadvantages based on recent literature and our findings.

Abstract

L33: please include month of harvest here

Done. (See next Comment for citation)

L34: the concluding sentence is not clear to the reader, what is meant by “small long-term differences”?

Small differences over the entire growing season are responsible for most of the differences in cumulative seasonal ET.

It now reads: “The key period contributing to ~ 70% of overall ET of the entire growth period was from April until June (harvest), however differences in the overall ET budget (ETsum) between soil types and manipulation resulted predominantly from small daily differences between the treatments over the entire growth period.”

 

Introduction

L39-40: Please include a reference showing that climate change and increase in human population are responsible for increased land use for agriculture.

Done.

L44/57: what is meant by “fertility” and “fertile” here? Probably better to use the term “plant production”, “yield” “productive” etc. because fertility and fertile refers to generative reproduction

Done.

L47: is it 10-40% of rainfall?

Yes it is. It now reads: “of precipitation”.

L48-50: not sure, if this is a trend. NE Germany has typically lower rainfall compared to the rest of Germany, should not be related to climate change

That’s right. This sentence is confusing due to the fact that Germany is part of the region with increasing precipitation in general. We changed it for a better understanding. It now reads: “In Germany, annual precipitation budgets of more than 800 mm occur in most regions in west and south Germany but only 400 - 500 mm y-1 in the northeast (e.g. areas in Brandenburg and Mecklenburg-Western Pomerania; Schappert, 2018).”

L82: should probably read “field scale” here, using eddy covariance on plot level is difficult.

Done.

L83: if this is the footprint of eddy-covariance: please keep in mind that the footprint (in field crops) is probably lower than millions of square meters – especially at daytime (which is the important time period for ET) and during the growing season. The measurement height must be quite high then to reach such a huge footprint.

It now reads: “At the plot scale for example, eddy covariance systems deliver high frequency estimates of evapotranspiration of a homogeneous system at the field scale while currently prevailing manual chamber approaches are able to precisely capture multi treatment effects (<1m²) on evapotranspiration but lack temporal resolution.”

L103: include “WUE_agro” in brackets here.

Done.

L109: should read “soil surface”

Done. (Not part of the manuscript any more)

L113: should read “study” instead of “investigations”

Done.

L124 and following: please remove brackets where not needed.

Done.

L126 and following: probably better to use the term “topsoil modification” as in L121.

Done.

L129/140: should read “fertilizer application” instead of “fertilization”

Done.

L129: should probably be 3 of 6 plots per soil type

Done.

L130-135: it is not clear to the reader why this modification of the soil was done? What is the reason for it? 

The manipulative field experiment was originally established to study the feedbacks of a dynamic disequilibrium in the carbon cycle of arable lands. Deep tillage or soil erosion lead to an admixture of subsoil material into the plough layer (Doetterl et al. 2016) which alters topsoil properties (SOC, clay content etc.). The resulting changes in the main rooting zone might reduce crop growth (Oettl et al. 2021). We mimic these common landscape processes in our manipulation experiment under controlled conditions (Vaidya et al. 2021). 

In the present manuscript we highlight the feedbacks of topsoil changes on the evapotranspiration. 

What can be taken from the comparison of modified vs. non-modified? 

From the comparison of modified vs. non-modified we obtain new knowledge on crop growth and evaporation feedbacks after a strong erosion event or a deep ploughing measure. 

This needs further explanation in the Introduction, Method, and Discussion section.

After topsoil removal (1.2 t per plot) we added the equivalent mass (1.2 t) of the respective subsoil horizons (E, Bt, Ck) taken from a large soil pit nearby. Subsequently we mixed the added subsoil material with the remaining local Ap horizon.

All of this information will be part of the revised manuscript.

L134: How much soil was added to each of the soil types?

See comment above!

L140: what was the amount of macronutrients applied?

We will include a table with detailed fertilization information in the appendix and refer to it in this section.

L141-142: could this be a problem when comparing non-modified vs. modified plots? 

Due to initial changes in the topsoil structure (after adding subsoil material) germination differed between manipulated and non-manipulated. To achieve similar plant densities in all plots we replanted seedlings as described. By doing so, we avoid pure germination effects on the processes / fluxes studied.

Did the modification improve germination of the crop?

No, see above.

L143: When was glyphosate applied? The term “Round-up” can be removed.

Done. It now reads: “For general plant protection and soil treatment, herbicides were applied to the field prior to the growing season (e.g. glyphosate; September 3, 2020).”

L144-152: It would be recommendable to explain the whole system here, i.e., analyzers and sensors, variables taken, measurement rate, length of measurement, how many measurement per day etc.

We agree that some information is missing or mentioned in the wrong place. All information concerning the system (mainly from chapter "2.5.1 ET Flux calculation") and other missing information will be incorporated in the revised manuscript.

L144: Did the chambers have fans?

Two small axial flow fans (5.61 m3 min-1) were used to homogenize the chamber headspace air (will be included in the description of the gantry crane).

L150-151: not sure what is meant here with “parallel”

Both chambers measure simultaneously/parallel with a fixed distance, so that always the same two plots of the two rows are measured. Since the "parallel" here is confusing/obsolete, we have removed it.

L155-159: please include which sensors were used. Also, a discussion on inside and outside PAR and Temp is helpful here.

We'll add information about the Temp/PAR sensors and the effects on inside/outside measurements in the revised manuscript. Both chambers have an average chamber light transmission of ~76 % (74% for chamber 1 and 78% for chamber 2), but Pape et al. (2009) reported only minor differences in light transmission for PMMA over the measured PAR spectrum (400-700 nm). Temperature differences occurred within the chamber but were minimized by the short measurement time and ventilation (<1.5°C).

L159: Is the SM measurement comparable when measured in an adjacent field without the treatment effects?

Sorry for the misleading description, it was not an adjacent field it was a plot on the same field. We have now marked the adjacent plot in Fig. 1 (< 25m away). We’ll add the location to the site map.

L161: should read “plots”

Done.

L166 the area of measurement probably depends on the measurement height here, please include.

That’s correct. We added the fixed height. Now it reads “The upward sensor was fitted with a cosine-correction diffusor for measurements of the incident radiation, while the downward sensor installed 1.8m above the ground had a 25° cone field of view, thus covering an area of 0.5 m2 during measurements (Görres et al. 2014). ”

L166/167: it is not clear how many measurements were taken with this information, please re-phrase.

Done. In now reads: „Each plot was measured once a week for 30 seconds, resulting in one mean value including 30 measurement points. “ 

L172: Please describe further what is meant by “stagnation”.

Stagnation occurred during the winter due to plant inactivity (no plant growth). It now reads: “Since only weekly plot-wise RVI data were available, daily RVI data were obtained by fitting a sigmoidal function for initial plant growth in the fall up to stagnation due to plant inactivity in the winter and a polynomial function for shoot elongation and later senescence during spring growth and summer maturation, respectively (Fig. A1).”

L182-183: Please include the exact number of measurements. It is not clear how many measurements per day were made, how often was measured, how much original data was available.

The measurement regime with 2 to 3 measurement days per week and soil type (each measurement day featuring up to 24 single chamber measurements per plot) will be explained in more detail in the revised manuscript. In addition exact numbers of measurements conducted by the system will be reported. 

L184: Were nighttime data included in the calculations? This may impact mean values.

All fluxes (day and night) of sufficient data quality (see section on hard and soft criteria) are included. It is true that this affects the diurnal/seasonal ET in comparison to approaches where only daytime data is available making this rather a benefit of the approach. 

L189: please include the step how ET was converted from mmol to kg (for WUE).

We have now added an equation for the conversion:

ET_flux [mmol m^(-2) s^(-1) ] =   (ET_flux [mm/d] / (t × 1000) )  * (1/M_H2O )      

        

L195-197: please include how much data was there originally, how much data was excluded as outlier per treatment, and later, how much data was gap-filled.

We will include a table in the appendix with information on the number of measurements per treatment and the corresponding gap-filling percentages.

L197 please write out “interquartile range”

Done.

L206: should read “2.5.2”

Done.

L244/311: why use SVM alone and not all gap-filling strategies and discuss it?

We did all the calculations for the five different gap-filling strategies. We'll add plots like figure 6 for all methods to the supplement and discuss them in more detail in the revised discussion section. 

L250: why not include NEE/GPP here and calculate WUE as GPP/ET? The LI850 would give the corresponding CO2 flux for it? The PAR and Temp could help to calculate respiration.

We fully agree that a detailed account WUE based on detailed gas exchange not only of ET but also of NEE and GPP is absolutely relevant. Of course the LI850 is capable and indeed we assess CO2 cycling as well as its partitioning (see e.g. Vaidya et al., 2021) along with ET measurements. For the sake of a reasonable word limit and given the already dual focus (1. Gap filling of ET and 2. Impact of erosion and land-use on ET) of the manuscript, we have decided to explore the treatment based differences as well as short term dynamics of WUE based on gas exchange of both carbon and water in a separate manuscript. We profoundly belief that this is necessary to be able to holistically discuss the expected system responses in water use efficiency.

L259-260: would a multilinear model work to estimate/explain ET? Is there difference among treatments or can the data be combined?

A multilinear model would represent the NLR approach implemented already (including RH, Temp., RVI and soil moisture, PAR), which however performed less well than the SVM to estimate ET. Difference between relations to single parameters are shown within the figure in terms of different colored regression lines.  

Results

L264-265: should be part of Method section

Done.

L283-291: not sure about the approximation sign here ~: RVI values are shown with accuracy of two digits. 

That was a mistake. We were talking about the mean values of all three treatments. We changed the sign now.

L293-295: should be part of Method section, and in more detail.

Done.

L296-300: please include actual ET values here.

We agree that true values of ET are missing here. We will add this in the revised manuscript.

L308: what is meant by “allocation problems”?

LUT allocation problems occur when no class is found for certain environmental parameters. The LUT approach will then use the mean value. We changed the formulation. It now reads: “Considering the k-fold-cross validation (Fig. 5, table 2), ANN_Br and SVM perform well, while LUT is not suitable due to allocation problems that arise when no class is found for the given climate conditions and the mean is used.”

L309: what is the difference between Table 1 and 2. The table description does not tell.

Table 1 shows the calibration statistics and Table 2 shows the validation statistics as explained in the description.

L311: could it be a solution to set negative values to zero? Were these nighttime values?

Computationally we could set the values to zero. However, this would violate general assumptions about the possibility of occurrence of negative ET fluxes. 

The problem in our specific case is not the general occurrence of negative fluxes but the occurrence during times of non-saturated atmospheric condition during the day (morning or late afternoon), where plant activity (e.g. stomatal opening and photosynthesis) must be assumed and the unreasonable magnitude of those fluxes (up to -6 mm/d). 

We therefore decided against it. The revised version of the manuscript will contain a more detailed explanation.

L319-320: should be part of section 3.6.

We have now merged the two sections as suggested by reviewer one.

Discussion

L328: Micro Bowen ratio systems have a relatively small footprint and can be used for small scales.

It is true that the Micro Bowen Ratio is not appropriate as an example at this point. We will delete it and discuss the differences elsewhere. 

L330: “used, new” please re-phrase: the reader may think: “either it is used or new”

Done.

L332: Please re-phrase, because eddy covariance ET flux is averaged over 30-minute intervals and measures at 20 Hz.

This paragraph has been shortened, making the sentence no longer part of the manuscript.

L334: it is not clear from the current version of the manuscript, if the ET is measured reliably (I suspect so, but it is not clear to the reader at this point). Please include a Discussion section about it.

We agree. While we have tried to discuss this to some degree (see the relation to a nearby lysimeter study, which however can´t be directly compared since the data is from previous years), this section was not very extensive and did not give the subject enough credit. The revised version will include a much more detailed account on the subject, summarizing ET measurement methods and discussing their advantages and disadvantages based on recent literature (see general comment 4).

L340: it is not clear what “long term” means here.

"Long-term" here means "over the entire growing season". We have changed it for better understanding. 

L244: it is not clear what “more dynamically” means.

In this context, "more dynamically" describes the naturally occurring diurnal effects on ET due to temperature and relative humidity, compared to the effects of RVI (plant growth), which occur slowly over the entire growing season. We'll explain this in more detail in the revised discussion section.

L345: why not include VPD in the study here. It should be possible to calculate it. It may give more insight the crop-atmosphere interaction.

It is indeed possible to calculate the gap filling approaches using the vapor pressure deficit (VPD), but temperature and relative humidity should be excluded to avoid autocorrelation. We calculated all the gap filling approaches using VPD, but upon comparison, we found that the calculation using temperature and relative humidity gave more accurate ET estimates.

L346: how does PAR impact ET? There is significant relationship, but an explanation is missing.

PAR has two effects on ET: increasing PAR leads to both an increase in photosynthesis and affects stomatal conductance, and an increase in temperature on the surface of the plants. This leads to an increase in ET with increasing PAR. We will discuss this in more detail in the revised manuscript.

L347: exclude “in summary”

Done.

L349: only one measurement from an adjacent field available.

That is correct, we will discuss the implication of our limited soil moisture data set in more detail.

L382-383: what is meant by “static differences”?

In this context, "static differences" means differences that develop slowly over the growing season such as plant growth. 

L395: should read “ET”

Done.

L395-398: should be part of a separate section on the accuracy of the ET measurements.

Yes, we agree and will discuss this in more detail in a separate paragraph in the revised manuscript.

L405-410: please re-phrase this statement.

We will re-phrase this statement in the revised manuscript.

L427-429: why not include wind speed in this study? Were no measurements available? Strong winds are mentioned L447, so maybe the crane is equipped with an anemometer?

The crane is equipped with an anemometer and wind speed data are available. However, we could not find any reasonable effect on ET.

L412: please adjust LUT then for this study. There is room for improvement, as can be seen in Figure 5. It seems that a few events are off, which if excluded improve the R2 substantially.

This sentence may be misleading. Studies with other datasets (as cited here) could produce valid results with the LUT. With our dataset, the only way to avoid allocation problems was to use fewer classes. This resulted in a loss of variability, making diurnal differences invisible and seasonal ET estimates less accurate. Points in figure 5 show the allocation problems and are thus a vital result of the validation.

Conclusion – in general too long and rather a discussion than a conclusion. Please consider moving L433-444 and L454-464 (and others) to the discussion section, and re-write the conclusion as a summary of the most important findings.

We agree. We will rewrite it in light of the comment and include the mentioned passages in the discussion.

L466: Why not include it in this study?

It is not entirely clear to us which statement this refers to. We did not have any further data available in regard to either long-term ET data or the Measurements of stable water isotopes.

Figure 3: “replicates summarized” – is it the mean or the sum of three replicates?

We will remove "replicates summarized". Figure 3 shows all the ET values calculated from the measurements of all three replicates per treatment.

Table B2 and Figure 7 could be combined.

This is a great idea. Thank you very much.

Figure 8. what is the standard deviation and what is the mean in the figure? Also, please discuss the effect of DM and ET on WUE in the Discussion section in more detail. E.g., it seems quite interesting that modified and non-modified Regosols have similar ET but different DM which impacts WUE.

In the revised version, we will rework this plot to better reflect the mean and standard deviation. The effect of DM and ET on WUE will be discussed in more detail in the revised discussion.

Fiugre A1/Figure 6: what kind of variation is shown in the graph?

In both figures, the lines describe the mean values of the three replicates with the shadow (variation) as the standard deviation. We will change the description in the revised version.

Figure 6: seasonal ET, is this the cumulative ET shown here? 

Yes it is the “seasonal cumulative ET”. We will use this term here and in the whole manuscript.