Using data assimilation to optimize pedotransfer functions using field-scale in situ soil moisture observations

Cooper, Elizabeth; Blyth, Eleanor; Cooper, Hollie; Ellis, Rich; Pinnington, Ewan; Dadson, Simon J.

doi:https://doi.org/10.5194/hess-25-2445-2021

Articles | Volume 25, issue 5

https://doi.org/10.5194/hess-25-2445-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/hess-25-2445-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 25, issue 5

Research article

|

10 May 2021

Research article |

| 10 May 2021

Using data assimilation to optimize pedotransfer functions using field-scale in situ soil moisture observations

Elizabeth Cooper, Eleanor Blyth, Hollie Cooper, Rich Ellis, Ewan Pinnington, and Simon J. Dadson

Download

Final revised paper (published on 10 May 2021)
Preprint (discussion started on 06 Aug 2020)

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'Comments on “Using data assimilation to optimize pedotransfer functions using large-scale in-situ soil moisture observations.” By Cooper et al.', Anonymous Referee #1, 15 Aug 2020
- AC1: 'Response to reviewer 1', Elizabeth Cooper, 23 Oct 2020
RC2: 'Review of hess-2020-359, Using data assimilation to optimize pedotransfer functions using large-scale in-situ soil moisture observations by Cooper et al.', Anonymous Referee #2, 20 Aug 2020
- AC2: 'Response to reviewer 2', Elizabeth Cooper, 23 Oct 2020
RC3: 'Comments on Cooper et al. “Using data assimilation to optimize pedotransfer functions using large-scale in-situ soil moisture observations.”', Anonymous Referee #3, 11 Sep 2020
- AC3: 'Response to reviewer 3', Elizabeth Cooper, 23 Oct 2020

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

ED: Publish subject to revisions (further review by editor and referees) (01 Nov 2020) by Bob Su

AR by Elizabeth Cooper on behalf of the Authors (09 Nov 2020) Author's response Manuscript

ED: Publish subject to revisions (further review by editor and referees) (24 Nov 2020) by Bob Su

AR by Elizabeth Cooper on behalf of the Authors (26 Nov 2020) Author's response Manuscript

ED: Referee Nomination & Report Request started (29 Nov 2020) by Bob Su

RR by Anonymous Referee #2 (05 Dec 2020)

RR by Anonymous Referee #3 (08 Dec 2020)

RR by Anonymous Referee #1 (12 Dec 2020)

Suggestions for revision or reasons for rejection

Soil Physics is fundamentally important

Great thanks to the authors for their efforts in addressing the comment. On the other hand, this reviewer is not convinced with the response by only adding “we assume that the soil texture values from the HWSD are correct …” while without evolving the fundamental view in terms of soil physics. This reviewer will motivate his observation with the following main point:

In the previous comment, this reviewer mentioned that “…This actually means that the better match between predicted SM and the COSMOSUK SM measurement, as demonstrated in this study, can be achieved with any other soil texture information input (e.g., SoilGrids, or FAO-UNESCO). But then, this is very dangerous then, as it will lead to speculation that the in-situ measured soil information is not important…”

The main reason behind the above comment is that although the current study showed very promising results in terms of improved soil moisture estimates, it was, however, achieved via optimizing constants in the underlying PTF functions.

Such an approach seems to overlook the fundamental importance of soil physics. The point is the improved performance of LSM is most probably not due to the realistic representation of soil physics, but might be potentially like ‘getting results right for wrong reasons.’ Unless the author can prove this point, this reviewer will suggest the authors continue their efforts in taking soil physics into account. For example, the PTF parameters (constants), being the focus of this study, are not soil physical property that can be measured directly, the optimization of which means no specific contributions to the understanding of soil physical processes. Perhaps, a better target for doing this exercise is to estimate directly the soil texture, which can be easily obtained either from the field or the soil database.

Some minor points as below for the author's further considerations:
1. In case the author want to pursue further their original direction, this reviewer would suggest them to have a look at the impact of applying different soil databases on their findings;

2. Instead of using only one PTF, this reviewer would suggest the author consider comprehensively other PTFs as well;

3. No matter which above point was considered in the further revision, this author would like to see the comparison between the estimated soil hydraulic (and thermal if possible/feasible) properties with either in-situ or lab measurements (e.g., soil water retention curve, hydraulic conductivity curve, etc.).

4. furthermore, perhaps the author can also clarify: instead of updating soil texture with the LaVEnDAR system, which only has two variables (e.g., sand and clay) to be updated, the 12 PTFs constants were targeted. Would not be more computationally effective to update 2 instead of 12? Or perhaps this reviewer missing something here.

Hide

ED: Publish subject to revisions (further review by editor and referees) (14 Dec 2020) by Bob Su

AR by Elizabeth Cooper on behalf of the Authors (11 Jan 2021) Author's response Manuscript

ED: Referee Nomination & Report Request started (19 Jan 2021) by Bob Su

RR by Hong Zhao (30 Jan 2021)

RR by Anonymous Referee #1 (15 Feb 2021)

Suggestions for revision or reasons for rejection

1. Authors’ response: We agree completely that accurate representation of soil physics processes in land surface models is absolutely crucial. When using a land surface model, there is inevitable uncertainty in the soil texture. Textures taken from any global dataset, as in this study, are likely inappropriately coarse. On the other hand, soil texture measurement taken at a point may also be unrepresentative of the scales on which land surface models are run.

Referee: I fully agree with you on this. Perhaps, could you help to put the above explanation also into the updated manuscript?

2. The usual approach is to use a pedotransfer function that has been optimised on very small scale measurements; here we are trialling the opposite approach and show that we can produce a consistent improvement to soil moisture outputs without generating completely unphysical soil physics parameters.”

Referee: Indeed, this is the contribution of this study. Nevertheless, we should make it more specific (e.g., in the ‘Conclusion’) that this is realized with the Data Assimilation approach.

3. We agree that using in-situ soil texture measurements would have been better. Unfortunately we do not have access to this data at the sites we used.

Referee: Could you help to make the above point clear in the updated manuscript?

4. In addition, we wanted to investigate the possibility of using a global soil texture dataset so that our newly parameterised pedotransfer function (PTF) could then be applied anywhere.

Referee: I think this deserves further thinking. The PTF parameters calibrated over 16 UK sites to be applied at a global scale? This seems to me unrealistic …. Or I am happy to see I am proved wrong …

5. We agree that there is danger here of getting the right answer for the wrong reason. However, we would argue that our results are robust for the following reasons: We have optimised a pedotransfer function taking into account soil texture and COSMOS-UK measurements from 16 independent sites simultaneously. We do not optimise the PTF on a per site basis. The fact that one newly optimised PTF improves the fit to data across all 16 sites suggests that this is a systematic improvement to the PTF –i.e. to the mapping between soil texture as reported in the HWSD and the soil physics parameters. The new soil physics parameters which are obtained with the newly optimised PTF are within the range of physically reasonable values. We did not constrain these values in any way so it would have been possible to generate physically unrealistic values; in this case we would have shared the reviewer’s concern about getting a better match to soil moisture measurements for spurious reasons.

Referee: Many thanks for your further explanations. Could you help to include the above text in the update?

6. As an extra test, we re-ran our experiment using soil texture information from SoilGrids. We found that the textures were indeed substantially different than those in the HWSD at some of our locations. However, we found that our conclusions were broadly consistent no matter which soil texture database we used. In both cases we were able to show a better fit to soil moisture data at all sites after optimising our chosen ptf, as shown in fig.1. We again found that the largest contribution to this improvement was in the alpha component of the KGE metric, i.e. we increased the range of soil moisture values available to JULES. We also again found that the value of vsat increased at all sites, and that the values of Ksat and sathh decreased across all sites after data assimilation; see fig.2.

Referee: This is good to know. Could you help to include the relevant results for SoilGrids? Or at least in the appendix (with some brief explanation). Readers (incl. this reviewer) will be more confident in the approach you proposed in this study with this result included.

Referee Report: PDF

Hide

ED: Publish subject to minor revisions (review by editor) (04 Mar 2021) by Bob Su

AR by Elizabeth Cooper on behalf of the Authors (13 Mar 2021) Author's response Author's tracked changes Manuscript

ED: Publish as is (21 Mar 2021) by Bob Su

AR by Elizabeth Cooper on behalf of the Authors (22 Mar 2021) Author's response Manuscript

Short summary

Soil moisture estimates from land surface models are important for forecasting floods, droughts, weather, and climate trends. We show that by combining model estimates of soil moisture with measurements from field-scale, ground-based sensors, we can improve the performance of the land surface model in predicting soil moisture values.