Characteristics of the Total Suspended Matter Concentration in the Hongze Lake during 1984–2019 Based on Landsat Data
Round 1
Reviewer 1 Report
This manuscript mainly use Landsat series data to estimate the total suspended matter (TSM) concentration in Hongze Lake from 1984 to 2019 for the first time, and analyzed its spatial-temporal distribution and influencing factors. Based on TM and OLI data, this paper constructed the suspended matter concentration estimation models for Hongze Lake respectively, and analyzed the consistency of the two models. The inter-annual and seasonal variations of TSM concentrations in Hongze Lake from 1984 to 2019 were analyzed in detail, and the influence factors of TSM concentrations were analyzed in combination with natural and human activities. The article drawn some meaningful conclusions. Overall, the research theme of this paper well-fits the scopes of Remote Sensing. Reviewer appreciate all of author's effort for the manuscript. The paper is easy to read in general, and the model and results are solid. However, some questions are still needed to be answered before the manuscript could be published in the Remote Sensing.
Major comments
(1) In the section of materials, the source of the full text data needs to be explained, the economic data such as GDP needs to be supplemented, and it is also necessary to explain how the data is processed.
(2) The formula for the TSM concentration estimation model constructed for TM and OLI needs to be reflected in the text, not just in the figure.
Minor Comments
(1) In Sec1.Introduction, the last paragraph is a requirement for the format of the article and should be deleted.
(2) Sec 4.3
‘he specific dates are shown in Table 2.’ Table 2 should be changed to Table 1.
(3) Sec 4.3
‘looding from the Huai River into Hongze Lake resulted in a large amount of suspended particulate matter……’,’suspended particulate matter’ should be changed to TSM, after the first occurrence of abbreviations in professional terms, the abbreviations are used in the future. Full text needs to be checked.
(4) Table 1, The font size in the table should be the same as the title and needs to be modified.
(5) Sec 4.4
‘Authors should discuss the results and how they can be interpreted from the perspective of previous studies and of the working hypotheses. The findings and their implications should be discussed in the broadest context possible. Future research directions may also be highlighted.’ This passage should be deleted.
(6) Sec 5. Conclusions
‘This article analyzes the inter-annual variation of TSM concentrations in Hongze Lake and the influencing factors…’’analyzes’ should be changed to “analyzed”. The tense of the full text needs to be checked.
(7) Sec 4.4
‘resulting in a significantly higher TSM concentration in the Huai River estuary and the Huai River Lake area than……’, ‘the Huai River Lake area’ should be changed to Huaihe Bay, it should be unified with the full text.
(8) Sec 3.2
‘The constructed TSM remote sensing estimation models were applied to Landsat images from 1984 to 2019 to explore the 36-year suspended matter concentration changes in Hongze Lake.’ Since 2012 data is missing, it should be 35 years to be precise.
(9) The full text of the formula size and style need to be unified.
Author Response
The response letter can be found in the attachment
Author Response File: Author Response.pdf
Reviewer 2 Report
In this papers authors develop an algorithm for mapping the distribution of total suspended matter (TSM) in Hongze Lake, China, based on in situ measurement and Landsat NIR reflectance.
Introduction
The literature review is mainly based on the study area rather than a review of different approaches regardless the study area.
Line 78: analysis algorithms? You mean analytics
Lines 109-117: Template text!
Methodology
The description of the methodology is not extended the reader can barely understand how all these data (in situ, remote sensing, meteo) are used. A flowchart would make the methodology more comprehensive to the reader.
Lines-171-173: Authors are describing something in a wrong way. L1 data are Level 1 data which designates the level of process, while the Tiers are the inventory structure of L1 products. With these lines authors describe the Level 1, Tier 1 (L1TP). However, authors should also justify why did they choose L1 and not L2 products without the need for atmospheric correction.
Figure 2, what are the lines. Figure is not clear maybe use a histogram or a table
Results
Lines 196-210 are not results. Some part goes to introduction as literature review and other part goes to methodology.
Equation 1, lines 209-2010 and figure 3 are the core of the TSM algorithm but is totally unclear to me what is being done here. I assume that with equation 1 the Rrs for NIR(λ) was calculated for the point samples of water for June 2018 and April 2019 with the ASD Field Spec. Then this Rrs was correlated with the NIR reflectance from the image pixels corresponding to the location sample producing the diagrams Fig 3a and Fig 3b. If that is the case my question here is how the authors produced the correlation of Fig3b for Landsat TM given that during the sampling acquisition 2018-2019, Landsat TM was decommissioned?
In Fig3c and 3d authors perform a validation comparing the estimated TSM and the actual TSM. Again, the same question applies for Fig3d. But my major concern here is that the make a validation with a sampling that has already be used in fitting, i.e. the 10 points are part of the 32 and this can be seen from the distribution of the plots of Fig3 and the Fig1 which have total 32 points. That is a major flaw of the process and a split into train and validation of the sampling dataset should have been performed before fitting.
Next authors perform a spatial change through time. They present the results based on the average of all images for each year. This could make sense if substantial and comparable number of images for each year was available but based on the fig 2 where the available data was ranged from 2 images to 11 images, statistics like averaging , trends, concentration variation between each year, maximum, minimum etc make no sense. TSM shows an annual variation based on various factors hence the acquisition of 2 images in a year does not reflect the status for the whole year. For example if in Year 2010 the 2 images acquired in autumn where the lower concentrations might expected and then the 2011 the 4 images are acquired in spring and summer where the higher concentrations are expected, the average of each year does not reflect the yearly trend.
Finally in the Discussion section authors perform an analysis between the interannual concentrations and other factors, but given the limitations discussed above from my point of view is not applicable.
Authors should study more careful similar studies i.e.
Spatiotemporal Distribution of Total Suspended Matter Concentration in Changdang Lake Based on In Situ Hyperspectral Data and Sentinel-2 Images https://doi.org/10.3390/rs13214230
where the majority of their limitations are not met. Furthermore, give that this is a site specific research, the results should be compared with other TSM results for the same study area. i.e.
Remote sensing of total suspended matter concentration in lakes across China using Landsat images and Google Earth Engine (https://doi.org/10.1016/j.isprsjprs.2022.02.018)
Based on the above, I regret that I have to reject the manuscript.
Author Response
The response letter can be found in the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The manuscript titled, “Characteristics of Total Suspended Matter Concentration in Hongze Lake during 1984 – 2019 Based on Landsat Data” by Du et al. presents a regional algorithm to estimate the concentrations of total suspended matter (TSM) in Hongze Lake (China). The algorithm is formulated with the near-infrared band of the Landsat sensors in order to ensure its applicability to Landsat data and utilize long-term data records of Landsat sensors in support of investigating spatiotemporal variations in TSM in Hongze Lake. Overall, the scientific approach presented in this manuscript is valid, and the results could be of significant interest to the scientific community. However, I have a couple of major comments/suggestions about the methodology to obtain Rrs from Landsat imagery and the selection of an appropriate TSM model for the study area. See my comments below,
Major comments:
1) TSM algorithm
Linear regression doesn’t seem to be a good candidate for the TSM algorithm based on the data that are used in this study. Currently, the linear model fits reasonably to data with high TSM concentrations (~ > 20 mg/L); however, it doesn’t pass through the data with TSM concentrations < 20 mg/L. Based on Figs. 3a and 3b, the proposed TSM model should overestimate the low TSM concentrations which can also be seen to an extent in Fig. 3c and 3d for two stations with low TSM concentrations.
The intercept of the linear model often provides important information about the validity of the model mainly based on the underlying relationship between dependent and independent variables. For example, the model in Figure 3a. has an intercept of 13.02 mg/L, which suggests that in the case of zero Rrs at NIR the model would yield the TSM concentration of 13.02 mg/L. Hence, the TSM model shows that there are particles with a total concentration of 13.02 mg/L even if there is no light emerging out of the water surface. The back-scattering of light due to mineral particles is generally the main contributor to the observed variations in Rrs signal at NIR in turbid waters. Conceptually, this result from the TSM model seems invalid because the zero Rrs suggests either the water is strongly absorbing all light that is back-scattered in the water column or there are not enough particles to create a strong back-scattering of light out of the water. The proposed model would make more sense if the intercept is close to zero.
In such a case, the non-linear model such as the power function or exponential function should provide robust fits to the data and conceptual strength to the model. Would you please explain on what bases the linear regression model is chosen?
2) Application of the TSM algorithm to the Landsat data
Section 2.3 suggests that the FLAASH-based atmospheric correction is applied to the Landsat imagery. However, the FLAASH-based atmospheric correction would provide dimensionless surface reflectance, whereas the model is formulated with remote sensing reflectance (Rrs; unit is sr-1). Both surface reflectance (unitless) and Rrs (sr-1) are different quantities which are often misunderstood and used interchangeably. The dimensionless FLAASH-corrected reflectances can be converted to remote sensing reflectances through division by the value of pi [i.e., ρ(λ) = Π × Rrs(λ)]. I wonder if such conversion is done before applying the TSM algorithm to the FLAASH-corrected Landsat imagery. The detailed information on FLAASH-based atmospheric correction over turbid waters is given in the following paper,
Moses, W.J., Gitelson, A.A., Perk, R.L., Gurlin, D., Rundquist, D.C., Leavitt, B.C., Barrow, T.M. and Brakhage, P., 2012. Estimation of chlorophyll-a concentration in turbid productive waters using airborne hyperspectral data. Water research, 46(4), 993-1004.
Minor comments/suggestions:
Line 74: Please add a reference. References [22-26] don’t include any reference on work related to remote sensing of turbidity.
Line 78: “analysis” should be “analytic or analytical”.
Lines 84-87: Without showing the results that other algorithms fail to provide reasonable estimates of TSM in Hongze Lake, it is difficult to prove this claim.
Lines 109-117: These lines seem some instructions on how to write the scientific article and does not belong to this manuscript.
Line 120: “N’” and “E’” should be “N” and “E”.
Line 126: “70 % of the total freshwater inflow to the lake”?
Line 144: Add a space between Rrs and (sr-1). Also, this line should end with “,” instead of “.”.
Line 145; Eq. 1: What value of r was used. Also, skylight correction is not accurate because r depends on a variety of factors. Thus, an additional step of residual correction is also applied to Rrs(λ) obtained from Eq. 1 by assuming zero water-leaving reflectance at NIR or at longer wavelengths (Mobley 1999). I wonder if the authors have applied such corrections in this study.
Also, please replace “*” by “×” in Eq. 1.
Mobley 1999. Estimation of the remote-sensing reflectance from above-surface measurements. Applied Optics 38, 7442-7455.
Line 160: “monthly precipitation” is written twice in this sentence.
Line 161: “maximum daily precipitation”.
Line 162: Please add “and” before “average wind speed. I wonder if all minimum and maximum meteorological variables have been used in this study. As per my understanding, only average wind speed and monthly precipitations are used in this analysis. If true, there is no point to list all the meteorological variables. Please keep only the ones that are utilized in this study.
Figure 2: It is stated in line 171 that the 2012 image data were missing, however, Fig. 2 shows that 3 images are obtained for the year 2012. Please clarify this.
Also, line 167 suggests that Landsat TM data were available up to 2011; however, the grey dashed line (for TM) extends up to 2015. Please clarify this.
Please add the description on the dashed lines in the figure caption. It is not clear what the dashed lines represent.
Line 190; Eq. 2: The correct form should be [(retrieved – measured)/measured] ×100 instead of [(retrieved – measured)/retrieved] ×100.
Line 196 – 197: This line needs to be supported by adding some references.
Lines 200 – 202: References 40 and 41 do not represent this statement. References 42 and 43 are not listed in “References”.
Lines 279, 390, 394, 420: Either space is needed, or it is missing in these sentences.
Fig. 10: It has never been mentioned in the text.
Table 1: Correct the symbol for the comma.
Line 463 – 466: These are instructions about how to write the Discussion section and it doesn’t belong to this manuscript.
Author Response
The response letter can be found in the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Other researchers acquire images for certain period of the year hence the data are comparable between each year. For example Du et al (2020) acquire images between April and October while Zheng et al (2015 ) acquire images only July, August and September. You do not mention in your manuscript for which months do you acquire images per year. A table should be inserted as Annex with dates (or at least the month) of the images in order for the reader to be aware any potential limitations of your trend analysis arise from your data
Du, Y., Song, K., Liu, G., Wen, Z., Fang, C., Shang, Y., ... & Zhang, B. (2020). Quantifying total suspended matter (TSM) in waters using Landsat images during 1984–2018 across the Songnen Plain, Northeast China. Journal of Environmental Management, 262, 110334.
Zheng, Z.; Li, Y.; Guo, Y.; Xu, Y.; Liu, G.; Du, C. Landsat-Based Long-Term Monitoring of Total Suspended Matter Concentration Pattern Change in the Wet Season for Dongting Lake, China. Remote Sens-Basel 2015, 7, 13975-13999, doi:10.3390/rs71013975.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
Thank you for replying to my comments. I see the improvement in the revised version of this manuscript. I am happy to recommend this manuscript for publication in Remote Sensing.
Minor comment:
The formula for MAPE is changed which should also change the MAPE in Fig. 3 and section 3.1. Please update these values in the manuscript.
Author Response
Please see the attachment
Author Response File: Author Response.pdf