A New Remote Sensing Method to Estimate River to Ocean DOC Flux in Peatland Dominated Sarawak Coastal Regions, Borneo

Round 1

Reviewer 1 Report

My major criticism of the study is the experimental design.  Water samples and in situ optical measurements were used to develop the algorithm.  Unfortunately a second field data collection from the 45 stations was not conducted to validate the DOC algorithm.  The same data used to develop the DOC algorithm should not be used to validate the DOC algorithm.  The study would benefit from an attempt to split the station data into algorithm development and the the remanding data into a field data validation of the DOC algorithm.  

The authors state that historical river gauge data was used to validate the TMPA monthly discharge estimates for the river basins.  The study would be improved by additional discussion of the there steps to estimate monthly discharge from each river basin and the validation of these estimate from river gauge data.

The manuscript would benefit from minor English grammar editing.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper addresses an important topic regarding estimating river to ocean DOC flux by combining the Landsat-8 derived DOC concentration and river discharge information. However, the analyses sound superficial and there is no appropriate validation. The validation should be based on independent data to examine the accuracy of image-derived DOC concentrations. At the moment, only a correlation coefficient between in-situ spectra and in-situ DOC concentrations is provided that tells nothing about the performance of the method. The literature review is also poor and previous works should be addressed. The structure of the paper also needs improvement. Although not rejecting, I recommend an extensive revision of the paper before reconsidering for a possible publication. Here are my detailed comments:

T

he acronym DOC should be spelled out in the first appearance in the abstract.

 

Linea 37-38: “The present study presents the first application of Landsat 8 imagery to estimate DOC concentration in turbid waters for estuarine and coastal waters.” >>> this statement is not precise. There are studies based on Landsat-8 estimating DOC concentration. For instance:

 

Monitoring dissolved organic carbon by combining Landsat-8 and Sentinel-2 satellites: Case study in Saginaw River estuary, Lake Huron

 

I suggest to properly cite the previous works and define the objectives/contribution of this study.

 

It should be clarified why only a single ratio model is used for approximating DOC concentration. There are other spectral features that can be examined or at least discussed:

Novel Spectra-Derived Features for Empirical Retrieval of Water Quality Parameters: Demonstrations for OLI, MSI, and OLCI Sensors

 

Moreover, the combination of spectral features (e.g., more than one band ratio) should be examined. There are relevant studies on feature combination:

Multiple Optimal Depth Predictors Analysis (MODPA) for river bathymetry: Findings from spectroradiometry, simulations, and satellite imagery

Multispectral bathymetry using a simple physically based algorithm

 

Lines 99-100: “We focused on linear and power models, images from Landsat-8 and band-ratio combinations of B2/B3, B3/B2, B3/B4 and B4/B3.” >> Some band combinations are examined and some not. I suggest using a systematic approach to find the optimal band combination for the ratio model. The analysis should be shown in the form of a matrix representing the R2 derived from each band ratio. You can refer to:

Spectrally based remote sensing of river bathymetry

Optimal band ratio analysis of worldview-3 imagery for bathymetry of shallow rivers (case study: Sarca River, Italy)

 

Section 2-6: it is a bit unclear the way max DOC is calculated and its role in the flux modeling. I suggest to formulate this part and provide some visualizations.

 

Section 3-1: this section is related to the dataset and it should be moved there.

 

Table 2 and Figure 3: the band ratio models and the associated correlation coefficients are valid only for the in-situ data. The in-situ spectral data are used for performing the regression. What about real landsat-8 data? Moreover, what is shown is just correlation of regressing in-situ spectra vs. in-situ DOC concentration. There is no validation neither for the in-situ analysis nor for the real data. The validation should be performed based on independent samples. The retrievals from real landsat-8 data also should be validated.

 

Section 3-2: better to show some maps for better understanding what has been done. Moreover, some part of the text is related to the method. In general, the paper requires some re-organization as the texts are mixed up in different sections and there are also repetitive material.

 

Line 206: “Uncertainty in DOC flux products generated in this study would be less than _15%:. Who did you end up with this number? The validation and uncertainty assessment are unclear.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors improved the manuscript by providing some indications of the accuracy by splitting the in-situ data into rating and validation sets. However, both spectral and DOC concentration data are from the field. The validation of the image-derived DOC concentration is very limited (only 3 matchups with a time difference with satellite overpass). I understand that in the tropical area the cloud cover can be a problem for using optical images but this limitation could be mitigated by using other case studies or more images. Given the limited validation performed in this study, it should be clearly noted in the paper.

 

I disagree with the argument of the authors that the references should be limited to those that conducted a study in tropical areas. There are many principals and methods which are common in different studies concerned with remote seeing of in-water constituents independent from the case study. The literature works should be acknowledged and properly cited and 37 is a relatively low number for the list of references. 

 

Without performing an experiment, it can not be concluded that a single ratio model would work better than multiple feature regressors. At least, this should be discussed.

 

An experiment has been added regarding examining different band ratios in a systematic way to find the optimal one but without referring to the relevant works. That's why I remind that the references should be cited properly.

Author Response

Please find the answer to reviewer 2 in the attached file. 

Author Response File: Author Response.pdf