Next Article in Journal
Using Remotely Sensed Information to Improve Vegetation Parameterization in a Semi-Distributed Hydrological Model (SMART) for Upland Catchments in Australia
Previous Article in Journal
Object Detection and Image Segmentation with Deep Learning on Earth Observation Data: A Review—Part II: Applications
 
 
Article
Peer-Review Record

Water Volume Variations Estimation and Analysis Using Multisource Satellite Data: A Case Study of Lake Victoria

Remote Sens. 2020, 12(18), 3052; https://doi.org/10.3390/rs12183052
by Yi Lin 1,2, Xin Li 1,2, Tinghui Zhang 1,2, Nengfang Chao 3, Jie Yu 1,4, Jianqing Cai 5,* and Nico Sneeuw 5
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Reviewer 4:
Remote Sens. 2020, 12(18), 3052; https://doi.org/10.3390/rs12183052
Submission received: 13 August 2020 / Revised: 7 September 2020 / Accepted: 16 September 2020 / Published: 18 September 2020

Round 1

Reviewer 1 Report

The authors have revised the manuscript following some comments. However, several comments are not responded or revised satisfactorily. Therefore, further revision is required.

(1) “Water level above lowest level” seems not a good term. It can be replaced by “water depth above lowest level” or simply water level above sea level. Please specify the lowest level in the abstract and main text.

(2) Use WLALL or ΔH throughout the manuscript, instead a mixture of them. Similar for WVALL or ΔV. Meanwhile, when described by text, WLALL should be spelled in full as “water volume above lowest level” instead of “water volume” or “water volume variation”.

(3) The input features / variables / indexes are not training samples for SVM, and should be clearly described.

(4) SVM Parameters should be optimized in this study, instead of “later study work”.

(5) In most cases, “lake surface height” and (lake surface hegth in Fig. 5) should be “lake surface level”.

(6) The fitting curve in Fig.8(a) is not a line.

(7) Either values or units of monthly precipitation is wrong in Fig. 12. Please check if the annual precipitation agrees with the observed one.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Dear authors, 

The manuscript was improved in many points (especially the figures). It is probably not possible to implement all suggestions given by the reviewers, however most of they were done.

 

 

 

Author Response

On basis of anonymous reviewers' constructive suggestions, the authors have revised the manuscript for improvement. We appreciate your work very much.

Reviewer 3 Report

This manuscript has significantly improved compared to the first version that I revised a while ago.

I'd like to start my (few) comments mentioning something about the Response 12 provided in the previous round. More specifically, regarding the part "It is a laborious and time-consuming task to download all the images over the 15 years and to identify the appropriate images after
the cloud recognition", I recommend the AppEEARS portal (https://lpdaac.usgs.gov/tools/appeears/). There, it is possible to select an area and a time period, and download a single (netcdf) file; thus much more practical than downloading individual images. In the case of MODIS images, I'd suggest to use only those with good quality, given by the quality flags (or Quality Assurance - QA flag).

 

Regarding the present manuscript, it seems that the authors are using, at least, two different software to produce the figures, which makes it difficult to have a more consistent ‘graphical pattern’ throughout the paper. I understand that sometimes this is inevitable, but I noticed that most of the Excel line plots were set differently. Yes, I agree this is esthetics, but it definitely forces me to rate the Quality of Presentation as ‘Low’. In most of those plots (e.g. Fig 9, 10 etc), the figure could be smaller if the font size is increased a bit.

Also, the resolution of some plots seems to be low. For example, it is absurd the difference in the quality between the plots in Fig 11 and 12 compared to Fig 10.  Perhaps that is caused by a file compression carried out by RS but please check.

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

This paper contributes with an extensive and detailed analysis
of the Lake Victoria water level and volume changes by using a
sufficient amount of data derived by multiple remote sensing resources.

The paper is well written and follows a clear logic line.

Some improvement can be made to the figures.

Hereafter, some suggestions for improvements are listed:

28. Note that the study makes use also of Envisat data.

97. Please specify over which period this change happened.

119. Substitute: 'river'---->'lake'

Figure 1. please add a second panel to Figure 1 with an an optical
unlabelled image of Lake Victoria

228. Substitute: Jaosn--->Jason

266. Add some references to literature.

291. What is the amount of the two average biases?

Figure 6. In Figure 6 the period from 2010 to 2017 is missing and
in the text a missing panel "a" is recalled several times.

Figure 6. I note a costant overestimation of the WLALL respect to
GRLM ENVI (specially from 2008). Is this shift within the precision
given for the estimations?

306. Substitute: presents after the procession --->shows the result
after processing

Table 4. This table seems inessential because the Figure 6 already
show all these values and in particular it is not clear how those
specific dates have been chosen. Furthermore, the average error is
clearly higher from 2008 onwards, why? Is this error eventually
within the precisions of the data products?

318. Double check the value of the maximum surface area variation.

Figure 7. Just keep one of the two panels of your choice.
They contain redundant information.

Figure 8. Panel (a) does not show a linear regression.
Panels (a) and (c) are the same.

Figure 10. You do not recall panel (b) of Figure 10.
Can you provide at least the slopes of these linear fittings?

438. Substitute: that the ---> that the relationship between the

456. Again, note that the study makes use also of Envisat data.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have revised the manuscript following most of comments from the reviewers.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.


Round 1

Reviewer 1 Report

The topic is interesting. However, several drawbacks of the study prevent its possible publication in the present form.

(1) The range of water level change during the study period is very small as compared with the whole range of lake level. Similar for water area and water volume. Therefore, the results are limited to these narrow ranges of water level / area / volume changes, and may not be applicable to a broader scope.

(2) The units for water volume and precipitation are WRONG. I’m not sure about the reliability of the results.

(3) The manuscript is a case study, and Lake Victoria should be clearly specified in the title and the beginning of the abstract.

(4) What are inputs and outputs of the support vector machine (SVM)? How was SVM parameters obtained?

(5) “The fluctuations in lake water volume for the period 2003-2017 shown in Figure 10(a-b) can be divided into four periods”. Therefore, trend or average in each sub-period should be give separately, instead of a single trend in the whole period.

(6) From Fig. 6, there exists systematic error between WLALL and G-REALM. Similar trends of these two variables does not necessarily indicate a good agreement.

 

Reviewer 2 Report

Dear authors,

The manuscript presents the study of the changes of the water volume (the relative water volume) of African largest Lake Viktoria. The water volume is estimated from 15 year series of water level and surface area calculated from the satellite altimetry data of Jason-1/-2/-3 and MODIS imagery. On my opinion, the text of the manuscript need to be further improved. Recently, the text lack of sufficient explanation of the data, methods applied as well as the results (the specific comments are included in the text attached to the review). 

I would suggest to pay attention to these references while discussing the results:

Vanderkelen, I., van Lipzig, N.P.M., Thiery, W., Modelling the water balance of Lake Victoria (East Africa) - Part 1: Observational analysis, Hydrology and Earth System Sciences, 22, 5509-5525

Schwatke, C., Dettmering, D., Bosch, W., and Seitz, F.:DAHITI – An innovative approach for estimating waterlevel time series over inland waters using multi-missionsatellite altimetry, Hydrol. Earth Syst. Sci., 19, 4345–4364,https://doi.org/10.5194/hess-19-4345-2015, 2015.

Comments for author File: Comments.pdf

Reviewer 3 Report

General comments

There is a certain mismatch between the title, the objective and the results.
First, we have an important word in the Title: 'multi-scale'; then, in the introduction, you mention 'spatiotemporal analysis' but nothing about such multi-scale analysis. When I get to the Methods section, I find a sub-section for each item composing the so-called 'technical route', except the 'multi-scale analysis'. So I get to the results without knowing what to expect about such analysis. Finally, while reading the Results, I see the outcomes of the items depicted in Figure 3 but nothing about a spatial analysis, as advertised in the Introduction.

As mentioned below, I do not agree with the validation of your estimates using G-REALM data, as both use the same source of data.

 

Specific comments

Abstract

L33-34: this sentence seem to be incomplete

Data


Section 3.2
L164-167: this sentence is a bit confusing and needs some polishing. For instance, you wrote: "For large lakes such as Lake Victoria, we use ... information of Lake Victoria,...". Although I can understand what you meant, there are better way of writing it. Moreover, in L166, 'more data' than what? Landsat? Please clarify.
About the first part, I suggest something like "Water body information of large lakes, such as Lake Victoria, can be extracted using ..."

L168: Can you make the part 'less cloud cover' less subjective? Can you put a number on it? What was the criterion adopted?

Method

The numbering (i, ii, etc) is not shown in Figure 3. Adopting a numbering-based route in the text and not showing it the in flowchart (which is divided in 3 parts) is a bit confusing.

Section 3.1 - This doesn't make any sense to me. G-REALM uses altimetry data to monitor lake elevation, which is exactly what you did in step (i), and using the same data (i.e. from Jason 1, 2 and 3). In other words, you use a altimetry-based estimate to validate your altimetry-based estimates.

Section 4.2 - As acknowledged in the text, there are some months with poor data availability. How this issue was addressed? For example, how did you estimate the lake area for November of 2005, 2006...2017?

According to Eq 3, you assume that all pixels have the same exact area; however, in reality (i) individual pixels may overlap neighboring pixels, (ii)  the 500 m size is the nadir pixel size and (iii) pixel size is expected to increase as you go from the nadir pixel toward the limb. A more accurate approach to determine pixel size would consider the pixel position and include the computation of the along-track pixel size.

Given the large extent of the study area, it is possible that both approaches would not be so different. Please verify and, if that's the case, provide a brief explanation justifying your choice.

Results

Figure 5: x-axis label ('Date') in (b) and (c) are not necessary

Figure 6: from (a) we can see a clear that WLALL is overstimating G-REALM but (b) is masking that due to the different ranges in 'x' and 'y' axis. Please adopt the same upper and lowers limits in both axis (e.g [-1 2.5]). This should make it easier to see that the bias in the estimates. Moreover, plotting the curves using different ranges (Fig 6a) is quite misleading. Having that said, I agree that both curves have similar tendency but not sure they 'agree well' as you wrote. If you meant, they agree in terms of shape, that's ok, although this idea is already contained in the tendency part. But they are quite different, numerically speaking (~2 times ??). 
Moreover, given that both estimates are altimetry-based, why not simply using G-REALM data?

Section 5.2
Figure 7: are those the (i) actual min and max areas or the (b) min and max areas of the available data? If (a), how did you estimate the water level for the missing months? If (b), please make it clearer in the text.

Section 5.3.1
L296 - the unit of RMSE is missing

Section 5.3.3

L358: if you plan to use wavelet analysis to explore the time-frequency structure of a signal, a minimum explanation would be expected in Methods, as such analysis is not so trivial to all readers.

 

 

 

Back to TopTop