Comment on “STH-net: a model-driven soil monitoring network for process-based hydrological modelling from the pedon to the hillslope scale ” by Martini et al. 

The manuscript presents the data set of a new soil monitoring network, providing soil water content measurements from TDR probes, soil temperatures, and meteorological forcing. The paper is well written and the provided data set is of high relevance for hydrology.

I only have few minor comments that need further clarification or additions to the data set.

Specific comments:

Measurement interval vs. reported data interval: The measurements are taken in a 10 minute interval. However, the data is reported in an hourly resolution. What is the motivation for this?

Please provide more details on the averaging. Line 197 only states “The original files are averaged to hourly values […]”. For cumulative data like the precipitation I assume it is not the average, but the sum of the the values in the previous hour. For the other values, is it the average of values half an hour before and half an hour after the reported time? Please clarify.

Also, the averaging interval of 12 hours to smooth the water content data seems quite long. This could impact the accuracy of the timing of e.g. the arrival of an infiltration front at a sensor location and reduce the information contained in the data set. For example, on October 4/5, 2019 the water content for T08 increases by about 0.1 in 10 hours (in the reported smoothed data). This indicates that the 12h smoothing interval may have a significant impact on some situations with rapid changes in the water content. Please briefly discuss the impact of the chosen smoothing, consider shorter smoothing intervals or possibly even provide an additional water content data file without this smoothing.

Temperature correction of water content measurements: The relative dielectric permittivity of water depends on the temperature. Temperature measurements are available. However, the manuscript does not mention if the dielectric permittivity for water is calculated temperature dependent. If so, please mention this and provide the details, including information on how the more sparse temperature measurements are inter- and extrapolated to the locations of the TDR probes.

There is very little information on the piezometer (e.g. manufacturer and model). The data itself is missing in the data set, although available since March 2020. Is it the goal to include the data into the data set? If so, I would recommend to add the data as part of this revision.

Overall, there is limited information about the different probes and sensors. I would recommend to add a table, that lists measurement range, uncertainty, and resolution for the different sensors.

Is the time reported in UTC or local time? Please mention this.

Line 144: “ε_soil is set to 4.6.” Please motivate why this value was chosen.

Technical comments:

Line 67-70: “Specifically, the approach followed at the site accounts for the soil spatial variability through detailed soil mapping and is designed to provide in situ data of, to our knowledge, the best quality available to date, with high temporal resolution and dense coverage in the vertical direction, about the soil water dynamics in the vadose zone and of its boundary conditions.”

There is no further information about the soil mapping in the manuscript. Is this part of a previous publication for this site? If so, please give a reference here.

In line 83 “Wollschläger et al., 2018” should be “Wollschläger et al., 2017” (at least according to the References).

Lines 112-114: “Each of the instrumented soil profiles located on the hillslopes features seven TDR probes installed at the depths of 0.1, 0.2, 0.3, 0.4, 0.5, 0.6 and 0.7 m, whilst the profiles at P4 and P5 feature additional TDR probes at the depths of 0.8, 0.9, 1 and 1.1 m in order to cover the deeper soils.”

P3 seems to also have a probe in the depth of 0.8 m.

Some of the figures have a rather low quality/resolution. This should be improved.

Comment on “STH-net: a model-driven soil monitoring network for process-based hydrological modelling from the pedon to the hillslope scale” by Martini et al.

The paper is generally well written and presented and describes a dataset will be useful for many wishing to understand soil processes better. A have some general comments and specific points which I have made below.

General comments

I am not sure the title for the paper is suitable. The title says the monitoring network is model driven, to me this suggests you used a model to find areas of highest uncertainty and instrumented these. How about shortening to “STH-net: a soil monitoring network for process-based hydrological modelling from the pedon to the hillslope scale”

Seems to be a lot made about the high resolution data collected at 10 minute intervals, yet the data set is averaged to hourly – not sure why this is. A user can average up if they need but they can’t disaggregate once you present averaged data. 12 hour smoothing and hourly averaging may remove a lot of important information, particularly around rainfall events when responses are quick.

The section on groundwater level measurement is very sparse. There are no details of the type of sensor used and no corresponding figure like the other sensors types get. This dataset also appears to be missing on the data sharing site. Details need to be improved and data added or any mention of the piezometers should be removed.

Has the accuracy of the WXT520 rainfall been assessed in any way? My experience in using such sensors shows that they work well in some climate conditions and not in others. Seeing as the rain is the primary driver of any modelling it is important to establish this as an accurate measure. These devices have been tested by others and show not produce some serious over-estimates (see Basara et al 2009)

Specific comments

L13 – change to “…dynamics are being…”

L29-31 – sentence needs to be reworded or split in two so that it makes sense

L35-36 – reword to “hence numerical models are needed for the comprehensive representation of the system state and fluxes so that the hydrological system can be better understood.” ???

L38 – change remarked to noted

L41 – makes no sense. Delete ‘however’?

L56 – change remarked to noted

L71 – change to “...aim to provide physical models…”

L131 – delete safety

L134- 135 – suggest change to “The TDR probes were custom made and have three 0.2 m-long rods. They were calibrated through measurements in air and in water…”

Refs

Basara, J. B., Illston, B. G., Winning Jr, T. E., & Fiebrich, C. A. (2009). Evaluation of rainfall measurements from the WXT510 Sensor for use in the Oklahoma City Micronet. The Open Atmospheric Science Journal, 3(1).

The submission describes the installation and data collection of a set of eight profiles of TDR soil water content sensors, temperature sensors, and groundwater wells. The installation and instrumentation are well described. The figures do a nice job of providing an overview of the site and the data being collected there.

The manuscript seems to fit many of the criteria listed in the ESSD guidelines, including describing open access data and having a thorough description of the data products. However, the manuscript also falls short in a few areas. The data management section provides an overview of some of the processing applied to the data, but I did not see those scripts listed on the URL provided in the paper. Also, I did not notice any discussion of uncertainties. I think this discussion could be fairly brief, but one example would be to do a sensitivity analysis for Eq. (2), which requires an estimate of porosity and soil dielectric. Given the emphasis of the journal, the authors should work to better quantify uncertainties in the data.

With attention to these details, the manuscript should be acceptable.

Specific Comments:

L2: The title seems more aspirational than descriptive. I saw nothing related to “model-driven” or “process-based” in the abstract, and while the introduction does a bit better in this regard, the authors might consider editing the title.

L13: dynamics “are”… Also this sentence is very long; might consider breaking it up.

L28: This “what” gives rise?

L32: Provide more details or support on what is meant by the strongest gradients are in the vertical direction.

L34: Could simplify as “heterogeneity of soil properties across spatial scales”

L37: Another long sentence that might be better split up.

L42: Define intermediate and large scales. Which do you consider the hillslope to fall into? You later also use ‘hillslope scale’ (L58), so try to be consistent with these terms.

L56-63: This paragraph was fairly unfocused, and in particularly the part about CZOs seems tangential to the manuscript (e.g., referencing chemical and biological processes).

L69: Awkward sentence.

L85: a subcontinental what?

L90: either “a meadow” or “meadows”

L102: Elaborate on pedological features.

L105: I did not see a scale on this figure.

L124: It would be helpful to label the horizons, if possible.

L135: What is meant by self-produced? That the probes produced them?

L150: Would be good to briefly comment on the data gaps in P4 and P5 (same for Figure 5).

L166: I would use “monitoring well” instead of “piezometers” here and elsewhere, since the two terms are not really synonymous. Also, the sensor used to measure water table height is not described here.

L179: “were” installed.

Review of the manuscript ‘STH-net: a model-driven soil monitoring network for process-based hydrological modelling from the pedon to the hillslope scale’ by Martini et al. (essd-2020-363)

General comment

I’m the fifth (!) reviewer of this manuscript and I have very little to add to this manuscript. This is a great site, a very detailed and dataset useful for several hydrological applications. The manuscript is overall well written and illustrated. I only noted some vague terms in Section 1 and I invite the Authors to be more specific in their statements. In addition to the comments by the previous reviewers, I have only few minor comments below.

Specific comments

1. As one of the previous reviewers, I’m not convinced about the “model-driven” part in the title…it looks as it would exclude the need of any purely experimental analysis.

35. I agree but why only models are necessary? On the contrary, I would say that a strong observational basis is necessary in order to understand these non linearities.

73. Yes, true, but it’s not described how it is possible to move from the hillslope the catchment scale. Please, elaborate on this.

Minor comments and technical corrections

16. What are the most relevant variables? A bit vague, please specify.

22. What does “high quality” mean? Please, specify. Also at 69.

27. “Larger” compared to what? Please, specify.

32. “Grandient”…in which terms? Soil types? Soil properties? Any reference? This is confusing, please clarify.

33. Which “phenomena”? Vagueness again, please specify.

40. "Landscape” is not so appropriate for hydrological applications…what about catchment?

42. What does “intermediate” refer to? Specify. Analogously, what does “larger” mean?

44. Why only “non linearities”? Please explain.

47. Small-scale = plot scale?

59-64.  Long sentence, please rephrase.

Fig. 4 and 5. Include precipitation in the top panel. Increase the font size of legend and axis labels.

Fig. 6. “Rainfall” should be “Rainfall rate (mm/hr- or whatever time step it is displays)”. Increase the font size of legend and axis labels.

MS: Data are new (2019-2020), methods seem appropriate, and data will be useful. However, some text could be provided about how these data might be useful beyond immediate application to hydrological modeling, for example, to biogeochemical modeling, for oxygen, organic carbon, redox-active species, contaminants, or nutrients? There are also a few missing references I identify below. The piezometer data are missing. Reference to those data should be removed entirely if they are not provided in the dataset.

Data quality: I think the 10-minute and hourly averaged data should be provided, unless clearly justified why 10-minute data are not useful. I also think the code for the data processing should be provided for reproducibility. There is also no mention of data missingness, how missing data values were imputed or treated. A table showing the percentage of data coverage (completeness for each variable) might be helpful if there was significant data loss. And if data are imputed a binary flag column showing which rows were imputed could be helpful.  In some cases, significant data were lost but there is no explanation. Why is there missing data from P2 for the first two months, and gaps from April to August in 2020 at P4? Finally, there is no header for the timestep variable in the Meteo text file, which creates a shift in variable names when the file is read into data analysis software.

Line-by-line Comments

Line 32: I like this point about heterogeneity, but wouldn’t processes be more localized if the steepest gradients were observed laterally?

Line 60: The citations Brantley et al., 2017 and White et al., 2015 (referring to CZOs) are missing from the references.

Line 64: Please state explicitly whether TERENO is part of the CZO network, or whether it is modeled after it?

Figure 4: What is going on between April and mid-August 2020 for P4 and P5 soil water content? data are missing

Line 163: Why are groundwater data not shown? (only available after March 2020)

Figure 7: Comment on similarity to other classification schemes, perhaps also include US soil order names in the text.

Line 200: How was signal vs. noise determined?

I’m yet another referee!  I agree with most of the comments submitted by the referees and am glad to see the author’s response to address them.  The manuscript is written well and the dataset will be of interest to the hydrometeorological community.  I just have a few points that I would like to be included in the revision.

1. The authors agreed to provide more information on the soil description, which is great. Figure 7 shows that texture class varies vertically at all stations, but it is difficult to extract the values as they overlap.  It will be valuable to have a table for individual silt, sand, and clay % at each station for all depth, so different soil texture classes can be derived for different models.  In addition, as a referee mentioned, dielectric permittivity, porosity, and soil bulk density are important parameters in the models that would be helpful to include as a list or reference.
2. Line 73 mentions that this experimental site aims to understand catchment-scale processes and the experimental site is a catchment size of 1.44 km2. Line 101 says the extent of the hillslope is 250x80 m, which seems to match the spatial maps (figure 1).  I understand that P1-P8 are chosen according to the topographic gradient to capture the soil water dynamics along the transect of the gauging station in hillslope scale, but I think that it would be helpful to show where the hillslope is embedded within the catchment to get a bigger picture.
3. The datasets feature variables capturing soils heterogeneity; however, any additional information on vegetation and the site environment is helpful in a modeling perspective. The text mentions that the land cover is grass—can you add general description on how tall, growing season, time of the mowing, root depth, and LAI if available?  The annual precipitation indicates it doesn’t rain much, but does it snow and accumulate in winter?