the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Jul 2023
Improved representation of soil moisture simulations through incorporation of cosmic-ray neutron count measurements in a large-scale hydrologic model
Abstract. Profound knowledge of soil moisture and its variability plays a crucial role in hydrological modeling to support agricultural management, flood and drought monitoring and forecasting, and groundwater recharge estimation. Cosmic-ray neutron sensing (CRNS) have been recognized as a promising tool for soil moisture monitoring due to their hectare-scale footprint and decimeter-scale measurement depth. Different approaches exists that could be the basis for incorporating CRNS data into distributed hydrologic models, but largely still need to be implemented, thoroughly compared, and tested across different soil and vegetation types. This study establishes a framework to accommodate neutron count measurements and assess the accuracy of soil water content simulated by the mesoscale Hydrological Model (mHM) for the first time. It covers CRNS observations across different vegetation types in Germany ranging from agricultural areas to forest. We include two different approaches to estimate CRNS neutron counts in mHM based on the simulated soil moisture: a method based on the Desilets equation and another one based on the Cosmic-ray Soil Moisture Interaction Code (COSMIC). Within the Desilets approach, we further test two different averaging methods for the vertically layered soil moisture, namely uniform vs. non-uniform weighting scheme depending on the CRNS penetrating depth. A Monte Carlos simulation with Latin hypercube sampling approach (with N = 100,000) is employed to explore and constrain the (behavioral) mHM parameterizations against observed CRNS neutron counts. Overall, the three methods perform well with Kling-Gupta efficiency > 0.8 and percent bias < 1 % across the majority of investigated sites. We find that the non-uniform weighting scheme in the Desilets method provide the most reliable performance, whereas the more commonly used approach with uniformly weighted average soil moisture overestimates the observed CRNS neutron counts. We then also demonstrate the usefulness of incorporating CRNS measurements into mHM for the simulations of both soil moisture and evapotranspiration and add a broader discussion on the potential and guidelines of incorporating CRNS measurements in large-scale hydrological and land surface models.
- Preprint
(27116 KB) - Metadata XML
-
Supplement
(3105 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2023-1548', Anonymous Referee #1, 29 Aug 2023
Please, see the file attached for my comments.
- AC1: 'Reply on RC1', Eshrat Fatima, 02 Nov 2023
-
RC2: 'Comment on egusphere-2023-1548', Anonymous Referee #2, 12 Sep 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1548/egusphere-2023-1548-RC2-supplement.pdf
- AC2: 'Reply on RC2', Eshrat Fatima, 02 Nov 2023
-
CC1: 'Comment on egusphere-2023-1548', Markus Köhli, 13 Sep 2023
In their manuscript "Improved representation of soil moisture simulations through incorporation of cosmic-ray neutron count measurements in a large-scale hydrologic model"
the authors Fatima et al. describe the combination of several years of data from CRNS instruments at selected experimental sites and hydrological modeling. The reviewer appreciates the approach of the authors to move towards a more comprehensive picture of how CRNS can help to understand hydrological dynamics, howeverm there are significant concerns about the methodology, data representation, completeness in the representation of CRNS and lack of transparency in the choice of models and parameters in the paper. The reviewer suggest addressing these issues to improve the quality and validity of the research.General:
- Without any apparent reason or justification this paper selectively uses the soil depths 0-5 cm, 5-25 cm and 25-60 cm. Except for the top soil layer, none of these classes are representative for either hydrological processes or measurement depths of CRNS. What is the reason for the authors to use that scheme? This sampling is too coarse and not deep enough to be acceptable and needs to be refined. This sampling can and will lead to non-obvious systematics and therefore would create misleading results. Specifically as the authors focus on statistical analysis methods this whole approach seems questionable.
- Using just the N0 method and the COSMIC operator in order to draw the general conclusions the authors would like to present is with respect to the state of the art not justified. Either narrow down the scope to a more exemplary analysis or include other models like the UCF method from Franz et al. or the UTS method from Köhli et al., which are both not even mentioned in the overview, yet mentioned in the discussion.
- The study heavily focuses on modeling and statistical analysis. Throughout the manuscript the reader encounters a large amount of seemingly arbitrary choices, which in the end provides the impression that results are selected and tinkered in order provide a realistic picture. The authors want to show that they uses methods which are accepted in the community, from the statistical and the modeling perspective, then, however, they introduce a significant amount of modifications and ad hoc assumptions which may put the whole approach into question. Exmaples are: In which way should a sophisticated hydrological model help, if the authors choose to simply take only three layers, for which the choice of hydrological parameters is also not really transparent? What is the reason to take only the 1% of the model runs with the best KGE (that number clearly depends on the initial parameter ranges the authors arbitrarily chose)? As the choice of the neutron model and their parametrization is not at all according to any standard, what does such tight constraint say other than the whole analysis is tuned to fit one specific ad hoc assumption. In which way is the KGE modified by the authors introducing a bias on this analysis? Why are 99 % of the model runs excluded if there are significant deviations from the models and the data, even visible in the plots presented? To be clear on that point: If significant deviations can be observed between model and data, any tight statistical constraint (in matching them incongruently) will lead inevitably to wrong results.
- Instead of overloading the manuscript with a multitude of different statistical measures, the authors should focus on providing a reasonable basis for comparing model and data. The authors rather present in the manuscript their own struggles and the reader does not learn anything from that way of analyzing a problem the authors fabricated in an intransparent way.General figure layout:
- the neutron data is plotted as quite large dots, which scatter significantly. Either smaller dots should be chosen or some type of smoothing.- l88: "the physics-based model COSMIC" - COSMIC is not physics-based, it is not based on a comprehensive physical interaction picture. It selectively takes specific processes and invents arbitrary mathematical representations for them.
- l99+: "What is the best approach to simulate CRNS neutron counts in a hydrological model considering the heterogeneity of vertical soil moisture profiles?" - This paper provides an exemplary data analysis which is insufficient for generalizations of the mentioned type
- l159+: "Simulations from mHM revealed that the sensitivity to the highest soil water content was observed at 5 cm depth (...)" - this sentence is highly confusing, grammatically and in the context of the manuscript as for example there is no information provided anywhere about a layer specifically at 5 cm depth.
- l165+: "Theoretically, the N0 parameter, which represents the neutron count rate level of the particular CRNS probe used for rather dry soil at the local conditions, should be site-specific" - please describe theories which underline the theoretical reasoning that N0 is site specific. The mentioned references do not provide that information. In case the N0 equation is an inadequate representation of the neutron count rate a site-specific behavior would be a result.
- l172: "may be impacted by factors such as soil chemistry" - within the field of CRNS researchers claim that this method would be independent of soil chemistry. The reviewer is curious how the authors come to this assumption
- l172: "heterogeneity" - which type of 'heterogeneity' do the authors refer to? In case 'heterogeneity' refers to topographical heterogeneity the whole approach of this analysis is questionable.
- l182: "derived from neutron particle physics modeling" - in which way are these parameters derived from particle physics and empirical (as mentioned above) at the same time?
- l200: The weighting scheme as presented is incomplete. (5)-(7) only take into account the weight of one depth. As to the model the authors used, the weight needs to be calculated by an integral over the depth weighting function for the height of the soil profile, not just the wighting function by itself.
- l203: What is the depth z_i? As the authors use soil horizons of considerable height, how do the authors calculate z_i?
- l210+: The COSMIC model only mimicks the mentioned subset of physical processes, in no way it represents them. Analytically COSMIC only represents an exponential N(theta) function, with an arbitrary parameter adaptation (8). The underlying mentioned physical processes are not responsible for the signal generation within CRNS as it lacks the spatial neutron transport, which CRNS claims to use as a unique feature compared to other methods.
- (11) is missing
- l241: explain the term "geometric integral"
- l275: What are "COSMOS models"?
- l293: "COSMIC is physically based, a loss of the physical meaning of the parameters in question would be very critical." - as described above COSMIC takes an incomplete subset of physical processes in order to justify its model. In this sentence the authors echo the critics which have been mentioned with respect to this model. Without the representation of neutron transport for example, any possible source-only model can hardly justify itself to be correct.
- Fig. 4: Why is the depiction of the hisograms so coarse? Please enlarge the scale to the relevant range
- Fig. 4: As many columns have the same height, the reviewer questions the representativeness of the results. The exact same height could mean that either the model provides on the basis of the 10000 data sets the same values or most of the results were discarded and the authors want to draw conclusions from just a few values.
- l323: Given the fact that the authors chose to represent the soil in very coarse layers, the "crowding cows" in the otherwise very wet catchment, seem to be a distracting and out of scope reasoning by the authors and at that point do not strengthen the scinetific quality of the material presented.
- Fig. 5 and Fig. 6. Both are plotted in such a tiny way, that it is hard to identify the different lines from each other and blur the relevant deviations.
- l379+: citep missing. "which are typically 5–60 cm and sensitive to shallow soil moisture" the reference tells that CRNS actually measures deeper than that and for that reason the choice of only simulating up to 60 cm seems wrong or incomplete.
- l445: "Overall, the three methods (NDes,U, NDes,W, and NCOSMIC) in mHM were able to consistently simulate the neutron count" - the results were inconsistent and the variability was only partially covered.Citation: https://doi.org/10.5194/egusphere-2023-1548-CC1 - AC3: 'Reply on CC1', Eshrat Fatima, 02 Nov 2023
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 562 | 296 | 43 | 901 | 64 | 23 | 30 |
- HTML: 562
- PDF: 296
- XML: 43
- Total: 901
- Supplement: 64
- BibTeX: 23
- EndNote: 30
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1