Towards a Long-Term Reanalysis of Land Surface Variables over Western Africa: LDAS-Monde Applied over Burkina Faso from 2001 to 2018

Round 1

Reviewer 1 Report

This paper investigates the capability of LDAS-Monde in estimating climatology of key vegetation and soil wetness variables through the joint assimilation of satellite derived LAI and SSM products. Particular attention is paid to the quality of atmospheric forcings which can potentially impact the overall performance of land data assimilation. Finally, the assimilation results as well as the open-loop estimation are carefully evaluated. As to my evaluating of the work, I see places to improve the manuscript, and I would recommend this paper being published on Remote Sensing by taking care of the following comments and concerns.

General comments

1.      Although LAI and SSM observations are ‘jointly” assimilated, which is a key feature in this LDAS-Monde implementation, model estimated LSVs (e.g. LAI and SSM) are separately evaluated. The physical connections between LAI and SSM are not addressed, and the individual contributions of LAI and SSM observations to the estimation of different LSVs as well as the overall performance of DA remains unclear. I realize it may not be feasible to conduct a series of new experiments, but I do expect some discussions addressing these concerns.

2.      While the authors claim “the assimilation is able to improve the simulation of both SSM and LAI”, I noticed this is concluded through evaluation using the CGLS products, which are also the ones used for assimilation. The problem is data assimilation, when effectively implemented, is by design to pull model predictions toward observations. And thus, it is no surprise to see DA generated SSM and LAI are closer to observations than the open-loop do. Indeed, there is a paragraph presenting evaluation on ET and GPP using independent datasets, but it is just not strong enough to demonstrate the capability of DA given that: 1) improvements of DA over open-loop are not so significant, and degradations are also observed at some circumstances (as seen from Fig. 12); 2) both improvements/degradations of DA over open-loop in time and space are not well interpreted or discussed. I would therefore suggest the authors to extend the current section 3.2.2 to include more detailed results/discussions by considering the above and, if possible, to reevaluate LAI and SSM with in-situ observations or other independent datasets.

Specific comments

1.      L80: quality ‘of’ atmospheric forcings …

2.      L90: does BF refer to ‘Burkina Faso’? need full name here

3.      Figure 1a: does ASCAT_SWI stand for saturation? If yes, then the unit cannot be m^3m^-3.

4.      L94: omit space after ‘project’

5.      L129-136: a diagram shows the general flow and design of the DA will be very helpful.

6.      L136: I always got confused with soil layering, better to have a diagram to clearly show the detailed layering scheme.

7.      L192-193: better to show latitudes and longitudes on maps in Figure 1.

8.      L202-205: during rescaling, do you exclude those modeled soil moistures when there is no effective ASCAT retrievals?

9.      L213: need a space before ‘As’

10.   L215: This ‘corresponds’ to

11.   Figure 5: it looks DA barely impacts model predictions at the SH and SS regions where seasonal averaged LAI is less than ~0.4. It would be interesting to see why DA performs differently in regions with sparse and dense vegetations. I expect the authors to spend a short paragraph to address this.

12.   Figure 11: lack of captions for subplots c)-d)

13.   L432: what does ‘10-member ensemble’ mean? Please clarify.

14.   L453-642: problems with numbering of references.

Author Response

We thank anonymous Reviewer#1 for his/her review of the manuscript and for highlighting the relevance of the study. Reviewer#1 has made several fruitful comments/corrections/suggestions that led to an improved version of the manuscript. A point-by-point response to Reviewer#1's comments is attached below (author-coverletter-3840670.v1.pdf).

Sincerely

Clement Albergel on behalf of the co-authors

Author Response File: Author Response.pdf

Reviewer 2 Report

In this paper, the authors applied LDAS-Monde land data assimilation system to generate the land surface reanalysis in Burkina Faso. The performance of their land analysis was thoroughly evaluated by several data.

General comments:

The topic of this paper is suitable to Remote Sensing. This paper is based on the authors’ previous intensive development of land data assimilation and their results are solid and interesting. However, I believe that there are many things to do to reach the full potential of this paper. I recommend the editor to reconsider this paper after major revisions.

Most importantly, I recommend the authors to run LDAS-Monde with both ERA-I and ERA5 and compare their performances. In the first part of this paper, the authors revealed that ERA5 outperformed ERA-I in terms of the representation of precipitation and downward solar radiation in Burkina Faso. In the second part, the authors evaluated the performance of LDAS-Monde driven by ERA5. The connection between two parts is unclear for me. I believe that the potential readers may be interested in how the improvement of ERA5 contributes to the calculation of land surface variables by an offline land surface model with and without data assimilation. Please discuss this point in the revised version of the paper by driving their land surface model with ERA-I.

My other comments are shown below.

Major points:

L78: Several papers have already generated and evaluated large-scale land surface reanalysis by land data assimilation systems (e.g., Yin et al. 2019; Sawada 2018). Please cite those literatures and discuss the new contributions of this paper.

L244: Although LDAS-Monde is well documented in the other papers, it may be worth mentioning the length of the assimilation window and the configuration of observation errors, which are important to interpret the results.

L259: Although ubRMSD is often used to compare model-simulated and observed soil moisture, bias and RMSD itself are also important to evaluate precipitation data. Please also report these metrics.

L282: Please report bias and RMSD for radiation.

L361-372: Could you please show the estimated Jacobian and interpret the results shown in Figure 10 using the structure of the Jacobian? I believe it is interesting.

L394-395: I believe that Figure 12 indicates the degradation of evapotranspitation estimation by LDAS-Monde. Please modify this sentence. Since GPP is improved, it is straightforward to assume that the estimation of transpiration is improved considering the photosynthesis-transpiration connections. Therefore, the degradation of evapotranspiration may be related to the estimation of soil evaporation. Since the authors did not modify surface soil moisture, no improvement of soil evaporation is expected. Is it correct? Please elaborate more to discuss why evapotranspiration is not improved.

Minor points:

L114: The first sentence of this paragraph may be grammatically wrong.

References

Sawada, Y. (2018). Quantifying drought propagation from soil moisture to vegetation dynamics using a newly developed ecohydrological land reanalysis. Remote Sensing, 10(8). https://doi.org/10.3390/rs10081197

Yin, J., Zhan, X., Liu, J., & Schull, M. (2019). An Inter‐comparison of Noah Model Skills with Benefits of Assimilating SMOPS Blended and Individual Soil Moisture Retrievals. Water Resources Research, 55. https://doi.org/10.1029/2018WR024326

Author Response

We thank anonymous Reviewer#2 for his/her review of the manuscript and for highlighting the relevance of the study. Reviewer#2 has made several fruitful comments/corrections/suggestions that led to an improved version of the manuscript. A point-by-point response to Reviewer#2's comments is attached below (author-coverletter-3840679.v1.pdf).

Sincerely

Clement Albergel on behalf of the co-authors

Author Response File: Author Response.pdf

Reviewer 3 Report

Review Comments for “remotesensing-451162”

In this manuscript (MS), the authors present an interesting work regarding the assimilation of remotely-sensed SM and LAI to improve the predictions of land surface variables. I think their work fits the scope of Remote Sensing. However, I found there are several problems existed in this MS.

One major issue with this MS is that it’s difficult to see the novelty. The introduction section of this MS is not well written. The authors failed to identify the scientific questions through literature survey, and they failed to demonstrate the novelty or motivation of their work here neither. Statements in line 65-69 are not enough. The findings of this MS are obvious – assimilating variables could improve the performance of LSM, the authors therefore must justify the contribution they made to the community.

The authors presented many results in the MS but failed to provide in-depth discussion. This is also related to my previous comment that the authors didn’t figure out the novelty or motivation of their work.

I encourage the authors to improve the introduction and discussion sections to better demonstrate the novelty of their work.

I also found some minor issues:

(1)   How many variables from ERA5 were used to drive the model?

(2)   Why atmospheric variables were interpolated to 0.25 degree which may introduce uncertainties, why didn’t set up the model at the resolution of atmospheric variables?

(3)   Fig 8 and 9, are these results monthly or seasonal?

(4)   The authors didn’t provide information regarding the study area, e.g. area, mean precipitation, etc.

Author Response

We thank anonymous Reviewer#3 for his/her review of the manuscript and for highlighting the relevance of the study. Reviewer#3 has made several fruitful comments/corrections/suggestions that led to an improved version of the manuscript. A point-by-point response to Reviewer#3's comments is attached below (author-coverletter-3852039.v1.pdf).

Sincerely

Clement Albergel on behalf of the co-authors

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Thank the authors for taking efforts to improve this manuscript. In the revision/response, more details on the DA implementations are provided, and most of the concerns on evaluation results are well addressed. I am now pleased to recommend this paper being published in Remote Sensing.

Reviewer 3 Report

I think the revised version can be accepted for publication in Remote Sensing.