High-resolution drought simulations and comparison to soil moisture observations in Germany

Boeing, Friedrich; Rakovec, Oldrich; Kumar, Rohini; Samaniego, Luis; Schrön, Martin; Hildebrandt, Anke; Rebmann, Corinna; Thober, Stephan; Müller, Sebastian; Zacharias, Steffen; Bogena, Heye; Schneider, Katrin; Kiese, Ralf; Attinger, Sabine; Marx, Andreas

doi:https://doi.org/10.5194/hess-26-5137-2022

Articles | Volume 26, issue 19

https://doi.org/10.5194/hess-26-5137-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/hess-26-5137-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 26, issue 19

Research article

| Highlight paper

|

12 Oct 2022

Research article | Highlight paper |

| 12 Oct 2022

High-resolution drought simulations and comparison to soil moisture observations in Germany

Friedrich Boeing, Oldrich Rakovec, Rohini Kumar, Luis Samaniego, Martin Schrön, Anke Hildebrandt, Corinna Rebmann, Stephan Thober, Sebastian Müller, Steffen Zacharias, Heye Bogena, Katrin Schneider, Ralf Kiese, Sabine Attinger, and Andreas Marx

Download

Final revised paper (published on 12 Oct 2022)
Supplement to the final revised paper
Preprint (discussion started on 05 Aug 2021)

Interactive discussion

Status: closed

RC1:
'Comment on hess-2021-402', Anonymous Referee #1, 29 Nov 2021

The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2021-402/hess-2021-402-RC1-supplement.pdf

Citation: https://doi.org/10.5194/hess-2021-402-RC1
- AC1: 'Reply on RC1', Friedrich Boeing, 17 Feb 2022
  
  Dear reviewer,
  we thank you for your assessment of our work. Please find our detailed response in the supplement .
  Best regards
  
  Citation: https://doi.org/10.5194/hess-2021-402-AC1
RC2:
'Comment on hess-2021-402', Anonymous Referee #2, 25 Jan 2022

Review of “High-resolution drought simulations and comparison to soil moisture observations in Germany” This manuscripts analyses the relationship between soil moisture observations and estimations by models in Germany with focus on drought monitoring. The manuscript is well written and organised. Nevertheless, I would like to include some caveats related to the limitations of the validation approach and the usefulness of the new high spatial resolution data base in order to assess drought severity. I include specific details related to these issues (and others) below (numbers refer to the specific lines of the manuscript):

11- What is “vegetation period”? Is maybe “vegetative active period”?

Table 1- I would like to ask for a technical question. Do you think if the quality of the globcover map is sufficient for the modelling. How is considered the uncertainty of land cover information in the model? I find very high detail of information related to the improvement of the soil maps, map I have the impression that the land cover data is not considered so carefully and it can be strongly relevant to model soil moisture given different water consumption by ecosystem types (even at the scale of species), the role of root structure, root depth, etc.

150- I find very few information related to the meteorological data. There is not information on the number of stations used for each variable, the quality of the data, quality control processes, data gap filling, temporal homogeneity, etc., but also information related to the quality of resulting gridded data (e.g., cross-validation statistics would be useful). Meteorological data can be also an important source of uncertainty in the model outputs…

151-154- What about uncertainty of the Hargreaves-Samani equation to estimate Potential Evapotranspiration? It is widely known that temperature based methods show uncertainties related to physically based models like the Penman-Monteith equation. For example, wind speed and relative humidity may have large importance on PET, even more in non-stationary scenarios charecterised by decreased relative humidity over land and wind speed reduction.

172- Figure 1 > Figure 2.

231-235- The validation procedure is exclusively based on correlations. Nevertheless, if the main purpose of the manuscript is related to drought monitoring, I think more relevant to assess model outputs during periods of water deficits. For example, it would be useful to check the capability of models to identify duration and magnitude of the dry periods. High correlation could mask a poor goodness between observations and models during dry periods. I would suggest to include statistics focusing on the drought periods in addition to the non-parametric correlations.

170-210- The length of the observation series is not indicated in this section. This information is relevant to assess robustness of the relationship between observations and models. Have the series the same length? How is this considered in the assessment of the signification of the relationships? I think this issue is affecting the validation of the results over the entire section 3.1 since the length of the series affect the degrees of freedom of the correlation analysis. I see in table 3 that the length of the series is between 2 and 5 years, which is too low to provide a robust validation of the model outputs.

Figures 6 and 7. Under my opinion, I do not think that this information is providing an useful output to determine the goodness of providing additional spatial resolution to assess drought severity. Large scale statistics are aggregating the information, being normal that both databases at 4km and 1 km of spatial resolution provide similar results. I think the relevant information of the 1 km modelling approach is not the general large spatial pattern but the local differences that could emerge given higher spatial resolution. This is something interesting to be analysed (e.g. using spatial statistics: the variance between grid cells, the differences between areas characterised by diversity of land cover/soil characteristics) to determine if higher spatial resolution is providing relevant information for drought monitoring and management. Observing Figures 6 and 7 I would say that the higher spatial resolution is really not needed as it basically identifies the same patterns that 4 km grids.

Citation: https://doi.org/10.5194/hess-2021-402-RC2
- AC2: 'Reply on RC2', Friedrich Boeing, 17 Feb 2022
  
  Dear reviewer,
  
  we thank you for your assessment of our work. Please find our detailed response in the supplement .
  
  Best regards
  
  Citation: https://doi.org/10.5194/hess-2021-402-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

ED: Reconsider after major revisions (further review by editor and referees) (05 Mar 2022) by Adriaan J. (Ryan) Teuling

AR by Friedrich Boeing on behalf of the Authors (13 Apr 2022) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (20 Apr 2022) by Adriaan J. (Ryan) Teuling

RR by Anonymous Referee #2 (05 May 2022)

RR by Rene Orth (16 May 2022)

Suggestions for revision or reasons for rejection

Review of Boeing et al., HESS-2021-402
“High-resolution drought simulations and comparison to soil moisture observations in Germany“

This study investigates the performance of the land surface modelling system underlying the
German Drought monitor. The authors compare two model versions with different spatial resolution
against a comprehensive set of soil moisture measurements obtained through different methods
and from different regions across Germany. They find overall similar performance with slight
improvements in the cold season, while additionally regional changes in the results can be observed,
probably following changes in the underlying soil type quantification.

-------------------

Recommendation:
I think the paper requires major revisions.

The topic of the study is interesting and timely. Drought monitoring in Germany is of increasing
relevance given the recent droughts, and their impacts on agriculture, forestry and water resources.
Accurate drought monitoring, as well as respective analyses and prediction, requires accurate
land surface models which can represent relevant processes. In this context, the current study
is a relevant contribution as it compares the modelling system of the German drought monitor
with a comprehensive set of soil moisture measurements obtained through different techniques,
and from different regions throughout the country. Further, the study illustrates the advantages
of performing simulations at higher spatial resolution which is increasingly relevant for regional-
scale drought assessments.
I think that the manuscript is a good match match for the readership of HESS, and appreciate the efforts
made by the authors to improve the manuscript in response to the previous comments of both reviewers.

However, before it is ready for publication some concerns particularly related to the understanding
of differences between modelled and observed soil moisture dynamics should be resolved,
as detailed below.

--------------------

General comments:

(1)
I agree with both reviewers that additional explanations/analyses are necessary for understanding the differences
between the modelled and observed soil moisture dynamics, and feel that the author's responses in this
context fall short in several aspects.
(i) Why does the updated model version compare better with observations in fall and winter, but not so much during
the growing season?
(ii) Following the main purpose of the German drought monitor, the validation of the modelled soil moisture against
observations needs to be performed also with an exclusive focus on drought periods in addition to the overall agreement
already assessed in the paper. Even though there is less data available during drought periods, I feel it should still
be sufficient for a meaningful analysis, particularly given the relatively low threshold used in the definition
of droughts, and the fact that the time series length does not systematically affect the determined agreement
between modelled and observed time series (Figure R1).
(iii) To better understand the small-scale differences between both model versions it would be helpful to relate the
obtained differences (e.g. in Figure 8) to the underlying differences between the soil type maps and land use
classifications to establish some cause-effect relations.
(iv) The role of updates in the land use vs. soil datasets, as raised by both reviewers, is not really clear even in
the updated version of the manuscript. While the authors state that changes in the land use dataset do not affect the
results, it remains unclear how they arrived to this conclusion.
(v) Finally, it would ask the authors to add some discussion on the poor agreement between models and observations
in winter months (and partly also in May). I realize that this might also be related to poor quality of the measurements
in the cold season, but wonder what this lack of validation means for the applicability of the Drought monitor
during these months of the year?

(2)
In this study drought is defined to occur 20% of the time. This seems arbitrary and is not related
to actual impacts of drought on ecosystems and/or society. Further, this seems quite a high percentage
in my opinion; if vice versa another 20% of the time then would be associated with conditions with too much water,
this effectively means that 40% of any time series considered in this study are regarded as extreme
conditions.

(3)
There are many small language errors (such as missing articles or wrong grammar) throughout the manuscript.
This was also pointed out by reviewer #1, but apparently the authors mainly corrected the issues
explicitely pointed out by the reviewer. Several errors are left in the manuscript, as stated below
in my specific comments.
The authors need to take special care of these points, and of an accurate and correct language
throughout the entire manuscript when revising the manuscript.

(4)
The processing of the in-situ soil moisture measurements is not entirely clear.
For example I did not get how the measurements, which represent different soil depths,
are aggregated to the layers considered in the analysis.
Further, I understand that in-situ data are averaged across stations in case they are
located in the same grid cell. But what happens if the time series length between the stations
is different? This could induce spurious signals in the aggregated time series which
may affect the comparison with the model data.
I appreciate that this is mentioned as a limitation by the authors in line 443, but feel
that this chould be addressed by e.g. normalizing the in-situ time series to make them
more comparable.

I do not wish to remain anonymous - Rene Orth.

------------------

Specific comments:

abstract: it would be good to mention the analyzed soil depth

line 9: replace "older" and "present" with more infomative terms

line 15: here you could cite Orth et al. 2022

line 36: Why is relative in quotation marks? And I feel that "relatively" would fit better.

line 54: "of" is not needed

line 58: would change "usually" to "sometimes" as many Fluxnet stations do not measure soil moisture at all

line 59: I would recommend to cite here Koster et al. 2019 and to include its findings in the discussion here.

line 67: I guess this is about *reflected* neutrons?

line 83: "...observed soil moisture that constitute the basis...", please improve phrasing

line 85: "...higher spatial resolved soil...", please improve phrasing

line 105: insert "the" before "US", and explain abbreviation

line 107: "The aggregation has based a set of upscaling rules...", please improve phrasing

line 128: Germany does not extent down to 45N

lines 136/137: I do not understand this.

lines 145-148: I do not understand this.

line 153: insert "the" before "plausible variable range"

line 154-156: Please provide more information on the calculation of the variograms,
including more motivation for doing this.

line 156-158: no actual surface radiation is used as an input?

line 202: superflous bracket

line 235: insert "the" before "sample size"

line 236: "data were masked to...", please improve phrasing, and provide additional details

lines 244, 263: Why are these reference time periods different?

line 294: typo in "represented"

line 295/296: "model performances does...", please improve phrasing

line 296: "correlations with length of time series...", please improve phrasing

line 331: "generally higher", please explain

line 344: variable rooting depths is not a process

line 381: insert "the" before "total soil column"

line 392: "2019 total soil", please improve phrasing

line 399: remove "to"

line 433: change "to be determined for" into "across"

lines 448-454: This is a very important point, good to see this here!

Table 1: What is meant with geology dataset?

Table 2, caption: please explain abbreviations dfj, jja, mam, son

Figure 1, caption: "Soil physical properties clay...", please improve phrasing

Figure 4, caption: panel d) is not mentioned

Figure 5, caption: "positive correlations are marked with x", I do not understand this
as there are also x symbols denoting negative correlations in the figure (?)

References:

Koster, R.D. et al. Length Scales of Hydrological Variability as Inferred from SMAP Soil Moisture Retrievals,
J. Hydrometeorol. 20(11), 2129-2146, doi: 10.1175/JHM-D-19-0070.1 (2019).

Orth, R. et al. Contrasting biophysical and social impacts of hydro-meteorological extremes,
Env. Res. Lett. 17, 014044, doi: 10.1088/1748-9326/ac4139 (2022).

Hide

ED: Publish subject to revisions (further review by editor and referees) (16 May 2022) by Adriaan J. (Ryan) Teuling

AR by Friedrich Boeing on behalf of the Authors (27 Jun 2022) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (05 Jul 2022) by Adriaan J. (Ryan) Teuling

RR by Rene Orth (10 Aug 2022)

ED: Publish subject to minor revisions (review by editor) (25 Aug 2022) by Adriaan J. (Ryan) Teuling

AR by Friedrich Boeing on behalf of the Authors (29 Aug 2022) Author's response Author's tracked changes Manuscript

ED: Publish as is (08 Sep 2022) by Adriaan J. (Ryan) Teuling

AR by Friedrich Boeing on behalf of the Authors (14 Sep 2022) Author's response Manuscript

Short summary

In this paper, we deliver an evaluation of the second generation operational German drought monitor (https://www.ufz.de/duerremonitor) with a state-of-the-art compilation of observed soil moisture data from 40 locations and four different measurement methods in Germany. We show that the expressed stakeholder needs for higher resolution drought information at the one-kilometer scale can be met and that the agreement of simulated and observed soil moisture dynamics can be moderately improved.