A Union of Dynamic Hydrological Modeling and Satellite Remotely-Sensed Data for Spatiotemporal Assessment of Sediment Yields
Round 1
Reviewer 1 Report
General Comments
The manuscript examines a Union of Dynamic Hydrological Modeling and Satellite Remotely-Sensed Data for Spatiotemporal Assessment of Sediment Yields. In general, the main objective is to use remote sensing data for better SY estimation using SWAT hydrological model. They claim the existing research gaps related to this topic (remote sensing + SY + SWAT) are i) lack of robust model frameworks for SY estimation under varying spatiotemporal and hydrogeological conditions; ii) lack of merged and ready to use information and data series from remote sensing into K-factor and C-factor estimation methods.
However, I doubt its novelty and I think they missed the concept of hydrological dynamics. Besides of the hydrogeological factors, SY is also depends on the river flow velocity and peak flow magnitude. So, how your framework improved the simulated SY? For example, Kamaludin et al., 2013; Manjulavani et al., 2021, BGanasri, and Ramesh, 2016, Sidi Almouctar et al., 2021, applied similar procedure as you used in your general methodology framework. Therefore, you need some justification or clear statement about your contribution.
- Kamaludin et al., 2013. Integration of remote sensing, RUSLE and GIS to model potential soil loss and sediment yield (SY). Hydrol. Earth Syst. Sci. Discuss., 10, 4567–4596, 2013
- Manjulavani et al., 2021. SOIL EROSION AND SEDIMENT YIELD MODELING USING REMOTE SENSING AND GIS TECHNIQUES. International Journal of Management and Applied Science, ISSN: 2394-7926 Volume-2, Issue-10, Special Issue-1, Oct.-2016.
- BGanasri, and Ramesh, 2016. Assessment of soil erosion by RUSLE model using remote sensing and GIS - A case study of Nethravathi Basin
- Sidi Almouctar et al., 2021. Soil Erosion Assessment Using the RUSLE Model and Geospatial Techniques (Remote Sensing and GIS) in South-Central Niger (Maradi Region).
Overall, the paper is well written and addresses an important topic. I have a few comments that should be addressed.
The cited literature is limited in terms of studies doing similar research. while I have not done a detailed literature review, I know of a few studies that have done nearly the same investigation (other locations) that are note cited (e.g., Kamaludin et al., 2013; Manjulavani et al., 2021, BGanasri, and Ramesh, 2016, Sidi Almouctar et al., 2021). While citing literature is common, more importantly how does the remote sensing products uncertain to the results in other studies? not many studies include the uncertainty estimates for each modeling system component. It would be good to include a few studies and compare the magnitudes of the key uncertainty contributors. are the major contributors the similar?
This is a nice piece of work, it reads well and the illustrations are clear. With regards to the discussion - I think there is a case for comparing and contrasting these results with those of others. There is a wealth of research out there and it would be useful to have a better feel of how these results compare with them. Apart from that, I think this is fine.
In conclusion, this manuscript needs a minor modification in terms of the comparison, modeling procedure and the quality of the results in order to contribute to hydrological dynamics theory or methodology and factual information about the computational hydrology dynamics of a region. Therefore, I recommend minor revision.
Comments for author File: Comments.pdf
Author Response
Manuscript ID: remotesensing-1499203
Title: A Union of Dynamic Hydrological Modeling and Satellite Remotely-Sensed Data for Spatiotemporal Assessment of Sediment Yields
Response to Reviewers:
We are very grateful to the reviewers for the thorough and insightful review. The comments and suggestions provided have contributed a great deal to improving the manuscript. According to these, we have made efforts in revising the manuscript, with the details explained as follows:
Reviewer 1 Comments:
The manuscript examines a Union of Dynamic Hydrological Modeling and Satellite Remotely-Sensed Data for Spatiotemporal Assessment of Sediment Yields. In general, the main objective is to use remote sensing data for better SY estimation using SWAT hydrological model. They claim the existing research gaps related to this topic (remote sensing + SY + SWAT) are i) lack of robust model frameworks for SY estimation under varying spatiotemporal and hydrogeological conditions; ii) lack of merged and ready to use information and data series from remote sensing into K-factor and C-factor estimation methods.
Response: Thank you for your valuable comments.
However, I doubt its novelty and I think they missed the concept of hydrological dynamics. Besides of the hydrogeological factors, SY is also depends on the river flow velocity and peak flow magnitude. So, how your framework improved the simulated SY? For example, Kamaludin et al., 2013; Manjulavani et al., 2021, BGanasri, and Ramesh, 2016, Sidi Almouctar et al., 2021, applied similar procedure as you used in your general methodology framework. Therefore, you need some justification or clear statement about your contribution.
Response: The connection of sediment yield (SY) to river flow velocities and peak magnitudes of flow is valid as stated by the Reviewer. However, sediment transport is primarily influenced by the gravitational forces acting on the sediment as well as the hydrogeological factors about land use land cover, soil, and topography. Hence, the present study utilizes two novel functionalities of K-factor and C-factor that are two important variables in predicting the SY in addition to the variables of rainfall erosion index, Universal Soil Loss Equation (USLE) support practice factor, USLE topographic factor, and coarse fragment factor.
The Reviewer is right about the similarity of the references such as Kamaludin et al., 2013; Manjulavani et al., 2021, BGanasri, and Ramesh, 2016, Sidi Almouctar et al., 2021. However, these studies have used Revised Universal Soil Loss Equation (RUSLE) whereas the present study used the USLE equation. The referenced studies have used a combination of remotely sensed data, field sampling data, and statistical interpolation techniques to predict SY whereas the present study uses two completely remotely-sensed data-driven equations to predict SY. These two equations of K-factor (Eq. 2) and C-factor (Eq. 1) were used for the very first time in modeling multiple watersheds to predict sediment yields.
(Preetha and Al-Hamdan 2022) (1)
(Preetha and Al-Hamdan 2018) (2)
The novel functions of K-factor and C-factor which were developed using satellite remotely-sensed data that is spatiotemporally dynamic are expected to tackle the uncertainties in SY predictions arising from the conventionally used static input variables for estimating K-factor and C-factor. Thus, they would improve the SY predictions.
Preetha PP, Al-Hamdan AZ (2022) Synergy of remotely sensed data in spatiotemporal dynamic modeling of the crop and cover management factor. Pedosphere 32(3): 381–392.
Preetha PP, Al-Hamdan AZ (2018) Multi-level pedotransfer modification functions of the USLE-K factor for annual soil erodibility estimation of mixed landscapes. Model Earth Syst Environ 1-13. https://doi.org/10.1007/s40808-018-0563-5
The aforementioned clarification has been added to the text of the manuscript in section ‘2.3. Development of Connective Algorithm’.
Kamaludin et al., 2013. Integration of remote sensing, RUSLE and GIS to model potential soil loss and sediment yield (SY). Hydrol. Earth Syst. Sci. Discuss., 10, 4567–4596, 2013
Manjulavani et al., 2021. SOIL EROSION AND SEDIMENT YIELD MODELING USING REMOTE SENSING AND GIS TECHNIQUES. International Journal of Management and Applied Science, ISSN: 2394-7926 Volume-2, Issue-10, Special Issue-1, Oct.-2016.
BGanasri, and Ramesh, 2016. Assessment of soil erosion by RUSLE model using remote sensing and GIS - A case study of Nethravathi Basin
Sidi Almouctar et al., 2021. Soil Erosion Assessment Using the RUSLE Model and Geospatial Techniques (Remote Sensing and GIS) in South-Central Niger (Maradi Region).
Response: Thank you very much for the valuable references provided by the Reviewer. It is acknowledged.
Overall, the paper is well written and addresses an important topic. I have a few comments that should be addressed.
Response: Thank you for your valuable comments.
The cited literature is limited in terms of studies doing similar research. while I have not done a detailed literature review, I know of a few studies that have done nearly the same investigation (other locations) that are note cited (e.g., Kamaludin et al., 2013; Manjulavani et al., 2021, BGanasri, and Ramesh, 2016, Sidi Almouctar et al., 2021). While citing literature is common, more importantly how does the remote sensing products uncertain to the results in other studies? not many studies include the uncertainty estimates for each modeling system component. It would be good to include a few studies and compare the magnitudes of the key uncertainty contributors. are the major contributors the similar?
Response: The Reviewer is right about the similarity of the references such as Kamaludin et al., 2013; Manjulavani et al., 2021, BGanasri, and Ramesh, 2016, Sidi Almouctar et al., 2021. However, these studies have used Revised Universal Soil Loss Equation (RUSLE) whereas the present study used the USLE equation. The referenced studies have used a combination of remotely sensed data, field sampling data, and statistical interpolation techniques to predict SY whereas the present study uses two completely remotely-sensed data-driven equations to predict SY. These two equations of K-factor (Eq. 2) and C-factor (Eq. 1) were used for the very first time in modeling multiple watersheds to predict sediment yields.
(Preetha and Al-Hamdan 2022) (1)
(Preetha and Al-Hamdan 2018) (2)
The novel functions of K-factor and C-factor which were developed using satellite remotely-sensed data that is spatiotemporally dynamic are expected to tackle the uncertainties in SY predictions arising from the conventionally used static input variables for estimating K-factor and C-factor. Thus, they would improve the SY predictions.
As per the suggestion of the Reviewer, we have included a few studies and compared the intensities and magnitudes of the key uncertainty contributors of the pieces of literature and the present study. Nevertheless, the remote sensing products were uncertain to the results in other SY studies owing to specific variables that were directly or indirectly connected to the spatiotemporal variation of SY such as land cover, soil moisture content, and topographical characteristics (slope gradient, slope length). Few studies evaluated the relations of C-factor, K-factor, and SY to land use change (Mancino et al. 2014), urban land covers (Mattheus and Norton 2013), dense forestation (Prasannakumar et al. 2012), vegetated zones (Gitas et al. 2007), and crop cover management (Qian-kun et al. 2015) that have not been accounted in the conventional USLE equations for SY predictions. Hence, the present study used annually varying land use land cover data for the study areas to understand the vitality of dynamic land cover modeling for realistic soil loss predictions. Besides, enhanced vegetation index which details the vegetative property of a hydrologic unit was considered as the primary variable in the novel C-factor modeling. Additionally, the variables of soil moisture content (Li and Meng 2013), slope length (Asmamaw and Mohammed 2013), and slope gradient (Mugagga et al. 2012) which are relevant for soil loss estimation were used as the independent variables affecting SY for the watersheds of the study.
Preetha PP, Al-Hamdan AZ (2022) Synergy of remotely sensed data in spatiotemporal dynamic modeling of the crop and cover management factor. Pedosphere 32(3): 381–392.
Preetha PP, Al-Hamdan AZ (2018) Multi-level pedotransfer modification functions of the USLE-K factor for annual soil erodibility estimation of mixed landscapes. Model Earth Syst Environ 1-13. https://doi.org/10.1007/s40808-018-0563-5
Mancino G, Nole A, Ripullone F, Ferrara A (2014) Landsat TM imagery and NDVI differencing to detect vegetation change: assessing natural forest expansion in Basilicata, southern Italy. iForest 7:75–84. doi: 10.3832/ifor0909-007
Mattheus C, Norton M (2013) Comparison of pond-sedimentation data with a GIS-based USLE model of sediment yield for a small forested urban watershed. Anthropocene 2:89–101. https ://doi.org/10.1016/j.ancen e.2013.10.003
Gitas IZ, Douros K, Minakou C, Silleos GN, Karydas CG (2007) Multi-temporal soil erosion risk assessment in N. Chalkidiki using a modified USLE raster model. EARSeL eProc 8:40–52. http://www.eproceedings.org/static/vol08_1/08_1_gitas1.pdf
Prasannakumar V, Vijith H, Abinod S, Geetha N (2012) Estimation of soil erosion risk within a small mountainous sub-watershed in Kerala, India, using Revised Universal Soil Loss Equation (RUSLE) and geo-information technol-ogy. Geosci Front 3(2):209–215. https://doi.org/10.1016/j.gsf.2011.11.003
Qian-kun G, Bao-yuan L, Yun X, Ying-na L, Shui-qing Y (2015) Estimation of USLE crop and management factor values for crop rotation systems in China. J Integr Agric. 14(9):1877–1888. https://doi.org/10.1016/S2095-3119(15)61097-8
Li L, Meng QY (2013) Reviews of phosphorus transport and transformation in soil under freezing and thawing ac-tions. Front Ecol Environ 6:1074–1078
Asmamaw LB, Mohammed AA (2013) Effects of slope gradient and changes in land use/cover on selected soil phys-ico-biochemical properties of the Gerado catchment, north-eastern Ethiopia. Int J Environ Stud. 70:111–125.
Mugagga F, Kakembo V, Buyinza M. 2012. Land use changes on the slopes of Mount Elgon and the implications for the occurrence of landslides. Catena. 90: 39–46.
The aforementioned clarification and the references have been added to the text of the manuscript in the respective sections.
This is a nice piece of work, it reads well and the illustrations are clear. With regards to the discussion - I think there is a case for comparing and contrasting these results with those of others. There is a wealth of research out there and it would be useful to have a better feel of how these results compare with them. Apart from that, I think this is fine.
Response: Thank you. It is acknowledged. The results of the study are compared with those of others as suggested by the Reviewer.
Hence, the connectivity of the dynamic estimates of C-factor and K-factor into SWAT realistically predicted SY when compared to the dynamic assessment using the singular C-factor model. It was supported by the findings of Mahmoodabadi (2011) and Vemu and Pinnamaneni (2012).
Mahmoodabadi M (2011) Sediment yield estimation using a semi-quantitative model and GIS-remote sensing data. Int. Agrophys. 25:241-247.
Vemu S, Pinnamaneni UB (2012) Sediment yield estimation and prioritization of watershed using remote sensing and GIS. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. XXXIX-B8:529–533. https://doi.org/10.5194/isprsarchives-XXXIX-B8-529-2012
The aforementioned clarification has been added to the text of the manuscript in section ‘3.2. Results of Connective Algorithm Validation’.
In conclusion, this manuscript needs a minor modification in terms of the comparison, modeling procedure and the quality of the results in order to contribute to hydrological dynamics theory or methodology and factual information about the computational hydrology dynamics of a region. Therefore, I recommend minor revision.
Response: Thank you for your valuable comments.
Author Response File: Author Response.docx
Reviewer 2 Report
In my opinion, this study presents a quite interesting attempt to exploit remotely sensed data to reproduce erodibility- and land-cover-related dynamics into the SWAT model. Unfortunately, the way the work is presented disadvantages it. The manuscript is too long and redundant; many sentences are actually repeated. In this way, the attention of the reader is not really focused on the main findings of this research. Hence, major efforts are needed to properly present this research. More comments in the following:
- The abstract is confused and disorganized. I suggest to re-write it.
- The Material and Methods section is too long. The information contained in this section could be adequately conveyed in a half of its current size. Too many technical details are provided; even though tools as like Python programming language or GIS are surely precious for scientists working in geosciences, repeating the words "Python" and "GIS" 8 and 9 times, respectively, in the manuscript provides an idea on the redundancy characterizing it. There is even a description of the Numpy package (lines 315-318).
- It is difficult to have an idea of the improvements obtained by passing from Case 1 to Case 2 and Case 3...the scores of Table 5 should be indicated in the figure.
- Figs 7 and 8. It seems that colors vary between 0 and a maximum that changes year per year. If this is correct, please uniform the colorbar in order to be able to evaluate the differences among various years. Also, the layout should be adjusted (i.e., huniform space between figures and between the colorbar and the table of values).
- Finally, the authors mention several times the link of this research with irrigation management. I think this aspect should be properly explained in the manuscript.
Author Response
Manuscript ID: remotesensing-1499203
Title: A Union of Dynamic Hydrological Modeling and Satellite Remotely-Sensed Data for Spatiotemporal Assessment of Sediment Yields
Response to Reviewers:
We are very grateful to the reviewers for the thorough and insightful review. The comments and suggestions provided have contributed a great deal to improving the manuscript. According to these, we have made efforts in revising the manuscript, with the details explained as follows:
Reviewer 2 Comments:
In my opinion, this study presents a quite interesting attempt to exploit remotely sensed data to reproduce erodibility- and land-cover-related dynamics into the SWAT model. Unfortunately, the way the work is presented disadvantages it. The manuscript is too long and redundant; many sentences are actually repeated. In this way, the attention of the reader is not really focused on the main findings of this research. Hence, major efforts are needed to properly present this research. More comments in the following:
Response: Thank you for your valuable comments. It is thoroughly checked and revised as suggested by the Reviewer.
The abstract is confused and disorganized. I suggest to re-write it.
Response: The abstract is revised as suggested by the Reviewer.
The existing frameworks for water quality modeling overlook the connection between multiple dynamic factors affecting spatiotemporal sediment yields (SY). This study aimed to implement satellite remotely-sensed data and hydrological modeling to dynamically assess the multiple factors within basin-scale hydrologic models for a realistic spatiotemporal prediction of SY in watersheds. (2) A connective algorithm was developed to incorporate dynamic models of crop and cover management factor (C-factor) and soil erodibility factor (K-factor) into the Soil and Water Assessment Tool (SWAT) with the aid of Python programming language and Geographic Information System (GIS). It predicted the annual SY in each hydrologic response unit (HRU) of similar land cover, soil, and slope characteristics in watersheds between 2002 and 2013. (3) The modeled SY closely matched the observed SY using the connective algorithm with the inclusion of the two dynamic factors of K and C (predicted R2 (PR2): 0.60-0.70, R2: 0.70-0.80, Nash Sutcliffe efficiency (NS): 0.65-0.75). The findings of the study highlight the necessity of excellent spatial and temporal data in real-time hydrological modeling of catchments.
The aforementioned clarification has been added to the text of the manuscript in the respective section.
The Material and Methods section is too long. The information contained in this section could be adequately conveyed in a half of its current size. Too many technical details are provided; even though tools as like Python programming language or GIS are surely precious for scientists working in geosciences, repeating the words "Python" and "GIS" 8 and 9 times, respectively, in the manuscript provides an idea on the redundancy characterizing it. There is even a description of the Numpy package (lines 315-318).
Response: The information contained in the section “Material and Methods” is revised and rewritten to a half of its current size as suggested by the Reviewer.
- Materials and Methods
This study is a union of two dynamic factors of the USLE equation into SWAT to advance SY predictions using satellite remotely-sensed data. The C-factor and K-factor models were developed using dynamic remotely-sensed MODIS data of enhanced vegetation index, the fraction of photosynthetically active radiation, leaf area index, and soil moisture. A GIS and Python programming language-based connective algorithm was proposed for the incorporation of the C-factor and K-factor models into SWAT. Then sensitivity analysis was performed to evaluate their influence on SY estimation in watersheds. The spatial distributions of the K-factor, C-factor, and SY at annual time scales were generated and analyzed for every hydrologic response unit (HRU) of similar land cover, soil attributes, and elevation characteristics within the watersheds of the study. The impact of the connective framework of K-factor and C-factor on SY is evaluated for three cases: (1) using the traditional factors of CUSLE and KUSLE, (2) using the dynamic C-factor and the traditional KUSLE, and (3) using the dynamic K-factor and the dynamic C-factor. The study thus enhances the spatiotemporal predictions of SY in the SWAT model. Besides, the agreement between the real-world observations and the spatiotemporal predictions from the satellite remotely-sensed data-based connective system of K-factor and C-factor would strengthen their reliability and global usability for real-time water quality modeling. It would also help to analyze the interconnectivity of the USLE factors and watershed erosion for better water use management.
2.1. Dynamic Models of C-factor and K-factor
2.1.1. C-factor Model
A pedotransfer functionality of C-factor was developed based on remotely-sensed geospatial data including enhanced vegetation index (EVI), a fraction of photosynthetically active radiation (SR in %), leaf area index (LAI), soil moisture content (AWC in %), slope gradient (S), and percentage area for every HRU of similar land use, soil, and slope characteristics in the watershed (A). The MODIS EVI data are of 250-meter resolution and contain the best possible pixel value for 16 days. LAI and SR data product is a 500-meter resolution product on a sinusoidal grid for an 8-day observation period. AWC data product represents the soil moisture content (%) in the surface from 0-10 cm, which determines the nutrients such as nitrogen, phosphorus, and organic carbon in soils (Li and Meng 2013). The values of the A and S were obtained from the model outputs of watershed delineation identified for the corresponding HRUs.
The remotely-sensed environmental data, including the EVI, AWC, LAI, and SR, were obtained per pixel from the processed MODIS imageries for the Southeastern United States (USGS 2019a; USGS 2019b; USGS 2019c) (Table 1). The remotely-sensed datasets of EVI, AWC, LAI, and SR were generated by the spatial superposition of the remotely-sensed data for the spatial level of HRU in the watersheds.
Table 1. Input data used in SWAT modeling and estimation of C-factor and K-factor.
Data |
Data sources |
Information |
Period |
DEM (S, LS) |
NED |
Raster, Annual, 30 m |
2002-2013 |
Land cover |
USGS: MODIS: LP DAAC |
Raster, Annual, 500 m |
2002-2013 |
Soil (BD, Psoil) |
USDA |
Raster, Annual, 60 m |
2002-2013 |
EVI
AWC
LAI
SR
Meteorological data
Hydrological data |
USGS: MODIS: LP DAAC
USGS: MODIS: LP DAAC
USGS: MODIS: LP DAAC
USGS: MODIS: LP DAAC
U.S. National Weather Service, NOAA
USGS |
Raster, 16-Day, 250 m
Raster, Monthly, 250 m
Raster, 8-Day, 500 m
Raster, 8-Day, 500 m
Daily , 250 m Daily , 250 m
Monthly |
2002-2013
2002-2013
2002-2013
2002-2013
2002-2013
2002-2013 |
|
|
|
|
The C-factor model used for the vigorous estimation of CUSLE in this study is given here:
(Preetha and Al-Hamdan 2022) (1)
where a and b are parameters that determine the shape of the exponential curve of C and EVI. A value of 1.5 and 1 was assigned for a and b, respectively and the coefficients presented in Eq. (2) were employed which yielded the best fit for the C-factor model for the study areas (R2 = 0.68, PR2 = 0.51, p < 0.05).
2.1.2. K-factor Model
The dynamic functionality of the K-factor in the study was developed using the topographic factor (LSUSLE), crop and cover management factor (dynamic C-factor), and soil properties of moisture content (AWC in %), bulk density (BD in g/cm3), and permeability (Psoil in mm/h). The topographic factor, LSUSLE was calculated using the equation by Neitsch et al. (2011). The values of the variables, such as slope length, L (m), and slope steepness or slope gradient, S (m/m) were obtained from the model outputs of watershed delineation identified for the corresponding HRUs. The slope, S, was calculated from the DEM of the watershed. The soil properties of BD and Psoil were obtained from the soil attribute characterization module (.sol) of the SWAT model. They were calculated by the spatial join of the soil map (soil type) and HRU map (HRU ID) in the SWAT model. Thus the developed model of K-factor serve as a dynamic and realistic improvement of the KUSLE equation in terms of capturing the HRU wise as well as annual variations in soil erodibility and is given here:
(Preetha and Al-Hamdan 2018) (2)
where C-factor is the crop and cover management factor developed using Eq. (1).
2.2. SWAT Model
This study uses the basin-scale hydrologic model, SWAT as a platform for the implementation of dynamic functionalities to enhance the soil loss predictions in watersheds. The SWAT is a physically-based, distributed parameter model that operates on the interface called ArcSWAT. It is used for the long-term analysis of hydrologic components and predicts the transport of sediments and contaminants in the watershed scale with varying soils, land uses, and management conditions (Arnold et al. 1998). The concept behind the modeling of spatial units in SWAT is the assortment of HRU, which are the portions of a sub basin that possess unique land cover, slope gradient, and soil characteristics. SWAT uses USLE to calculate SY (Neitsch et al. 2001; Neitsch et al. 2011). The SY from USLE in each HRU in a given day, denoted by SYUSLE (metric ton/ha) are obtained from the SWAT model using the equation
SYUSLE = 1.292 EIUSLE KUSLE CUSLE PUSLE LSUSLE CFRG (3)
where EIUSLE is the rainfall erosion index in 0.017 m-metric ton cm/ cu.m hr, KUSLE is the K-factor in 0.013 metric ton cu.m hr/ cu.m metric ton cm, PUSLE is the USLE support practice factor, LSUSLE is the USLE topographic factor, and CFRG is the coarse fragment factor.
2.3. Development of Connective Algorithm
The study developed a connective algorithm to simulate spatiotemporal sediment yields in SWAT by incorporating two models of C-factor and K-factor. These two equations of K-factor (Eq. 4) and C-factor (Eq. 2) were used for the very first time in modeling multiple watersheds and to predict sediment yields. The novel functions of K-factor and C-factor which were developed using satellite remotely-sensed data that is spatiotemporally dynamic are expected to tackle the uncertainties in SY predictions arising from the conventionally used static input variables for estimating K-factor and C-factor. Thus, they would improve the SY predictions. The proposed algorithm to incorporate dynamic models of C-factor and K-factor into SWAT was based on modeling a yearly variation scheme for the traditional equations of CUSLE and KUSLE in the SWAT model. The algorithm adopted an input-oriented functionality and runs on annual temporal conditions in the level of sub basins and HRUs. The newly incorporated components included modification of SWAT equations (SWAT-Equation), addition and editing of SWAT input files (SWAT-Edit), and extraction of SWAT output files (SWAT-Extract). SWAT-Equation was used to modify the equations in the routines and subroutines. SWAT-Edit was used to add new model input files as well as to edit the existing model input files of SWAT such as .crop (land management and vegetation characterization module) and .sol. Additionally, SWAT-Edit edits the model input files of SWAT simultaneously for .crop and .sol modules. First, the remotely-sensed factors that take part in the algorithm were fed into the respective input modules of SWAT and saved using SWAT-Edit. Second, each row pointing to one HRU was modified to the new functions for C-factor and K-factor (Eqs. (2) and (4)) using SWAT-Equation. Then SWAT was called to simulate SY for the study watersheds for the required temporal conditions. Further, the spatiotemporal maps for the CUSLE, KUSLE, K-factor, C-factor, and sediment yields were processed. The described processes are shown in Figure 1.
Figure 1. The workflow of the proposed algorithm for obtaining dynamic C-factor, K-factor, and SY.
2.4. Validation of Connective Algorithm
2.4.1. Sensitivity Analysis
The developed algorithm connecting models of the C-factor and K-factor into SWAT was validated using sensitivity analysis of SY predictions in three different cases. The sensitivity analysis involved three cases of SY predictions for annual conditions in the HRU level using spatiotemporal C-factors and K-factors. Case 1 represented SY with the traditional values of CUSLE and KUSLE. Case 2 represented SY with dynamic C-factor and traditional KUSLE. Case 3 represented SY with the dynamic C-factor and dynamic K-factor, respectively. All cases were correlated to observed sediments at the monitoring stations to understand the effect of the C-factor and K-factor functions in realizing the spatial and temporal dynamics in SY estimation.
2.4.2. Performance Evaluation
In this study, the performance of the developed algorithm in advancing spatiotemporal SY predictions was evaluated using three statistical methods. The coefficient of determination R2 ranges typically from 0 to 1 and is expressed as
R2 = (4)
where (5) and (6)
y and f represent the observed data and predicted data, respectively. SStot, and SSres, represent the total sum of squares proportional to the variance of the data and the sum of squares of residuals, respectively.
Next, the predicted R2 (PR2) was employed, which indicates the wellness of a model in predicting responses to new observations. One of the most commonly used coefficients of determination in hydrology is the Nash–Sutcliffe model efficiency coefficient (NS).
(7)
where Xsimulated_average is the mean simulated stream flows, and SY of the watersheds averaged per HRU.
2.5. Spatiotemporal Analysis
2.5.1. Spatial Interpolation and Mapping
In the present study, the inverse distance weight (IDW) method was used for computing the spatial patterns of the remotely-sensed variables of EVI, AWC, LAI, and SR for all the HRUs in the watersheds annually. The raster data was used to obtain the HRU wise spatial maps of EVI, AWC, LAI, and SR. Each HRU was organized into grids where an HRU contains a value representing EVI, AWC, LAI, and SR, respectively (Risal et al. 2016; Sadeghi et al. 2017). Later, they were used to estimate and geospatially map the modified C-factors and K-factors.
2.5.2. Temporal Trend Detection
The t-test was employed as a parametric trend detection test to understand the HRU wise annual trends of C-factor, K-factor, and SY (Yue and Wang 2004; Onoz and Bayazit, 2012).
2.6. Case Study Areas
Two case studies were chosen as representations of the water quality constituent of SY evaluated using dynamic soil erodibility and cover management factors with changing space and time. Two basins, the Tampa Bay watershed (TBW) in Florida and the Winyah Bay watershed (WBW) in South Carolina, were used in this study (Figure 2). The TBW is located on the Gulf coast of west-central Florida and has an area of approximately 590,522 km2. The second study area of WBW is approximately 575,590 km2 and encompasses the neighboring areas of the tidal waters of the estuary. The TBW highlights the hydrological responses of a coastal zone with rapid urban development, and the WBW implies the impact of coastal development on increasing soil erosion.
2.6.1. Data and Modeling
The elevation data was rasterized as a 30 m resolution digital elevation model (DEM) from the National Elevation Dataset (NED) provided by the United States Geological Survey (USGS). The soil data concerning texture, depth, and drainage attributes were rasterized from vector maps supplied by the Web Soil Survey (WSS) under the United States Department of Agriculture (USDA 2018). The study area watersheds were delineated using the DEMs. The annual land cover maps of the watershed HRUs were developed by a supervised classification analysis based on remotely-sensed MODIS images. The developed classifier was trained for the study area watersheds which produced their supervised classification with fifteen land cover classes (Preetha and Al-Hamdan 2018). The annually classified land cover maps were employed as the land cover unit into the SWAT model for enhancing the hydrological predictions in SWAT.
Figure 2. Location of the study areas of TBW and WBW with the employed monitoring stations.
The models were developed in SWAT for both study areas. A watershed delineation was performed, and several sub basins were obtained. Then the overlay of land cover, soil, and the slope was carried out, producing the HRUs within the sub basins. The threshold values used for land cover, soil, and slope while defining the HRUs were 7%, 12%, and 12%, respectively. Later, climate data such as precipitation, temperature, solar radiation, relative humidity, and wind speed were incorporated, model setup was completed, and simulation runs were performed. The stream flows and sediment loads from each HRU were calculated separately using input data for weather, soil, topography, vegetation, and land management practices. Later, they were merged to determine the total loadings from the sub basin as well as from the HRUs. More details and descriptions of the water balance, soil erosion, and water quality process equations can be found in the SWAT technical documentation (Neitsch et al. 2011). Twelve years (2002-2013) of daily weather data were collected from ground weather stations as well as from downscaled and projected data from Climate.gov (Neitsch et al. 2005; NOAA, 2017) with a spatial resolution of 250 m. The observed monthly stream flows and sediment concentrations (Xobserved), obtained from USGS for the years 2002-2013 at the respective monitoring stations were used to calibrate and validate the SWAT model. All the datasets for the study were collected for the years from 2002 to 2013 (Table 1). Simulations were done for land cover, soil type, and climate condition of 2002 to 2013 using the weather data with the existing management practices, which generated the simulated stream flows and SY from 2002 to 2013 (Xsimulated). Table 2 explains the descriptive statistics of remotely-sensed and modeled soil properties for K-factor estimation in TBW and WBW. The soil texture is stable for a relatively long period of five (2009-2013 in WBW) to eight (2002-2009 in TBW) years in certain parts of the watersheds. However, the majority of the HRUs showcased annual variations in the soil property of AWC, and some of the HRUs showed for BD and Psoil, respectively.
Table 2. Descriptive statistics of remotely-sensed and modeled soil properties for K-factor estimation in (a) TBW and (b) WBW.
(a)
Descriptive Statistics |
AWC (%) |
BD (g/cm3) |
Psoil (mm/hr) |
LS (m) |
A (sq.km) |
Mean |
13.82 |
1.20 |
497.02 |
23.88 |
258.32 |
Median |
12.27 |
1.25 |
537.50 |
19.75 |
36.34 |
Mode |
24.60 |
0.87 |
487.50 |
0.00 |
0.18 |
Minimum |
0.00 |
0.29 |
0.12 |
0.00 |
0.09 |
Maximum |
37.49 |
1.78 |
875.00 |
403.48 |
13356.90 |
(b)
Descriptive Statistics |
AWC (%) |
BD (g/cm3) |
Psoil (mm/hr) |
LS (m) |
A (sq.km) |
Mean |
23.33 |
1.17 |
121.90 |
22.94 |
276.95 |
Median |
22.84 |
1.14 |
116.75 |
7.03 |
35.73 |
Mode |
18.25 |
1.44 |
39.25 |
0.30 |
0.09 |
Minimum |
17.47 |
0.20 |
2.80 |
0.00 |
0.01 |
Maximum |
32.80 |
1.55 |
887.50 |
1131.53 |
10426.52 |
It is acknowledged. The unnecessary mentions about “Python" and "GIS" have been deleted from the manuscript. The term “Numpy package” and its detailing have been deleted from the manuscript.
The aforementioned clarification has been added to the text of the manuscript in section “2. Materials and Methods”.
It is difficult to have an idea of the improvements obtained by passing from Case 1 to Case 2 and Case 3...the scores of Table 5 should be indicated in the figure.
Response: The scores of Table 5 are indicated in Figure 4 as suggested by the Reviewer.
Figure 4. The observed versus predictions of SY using CUSLE-KUSLE (Case 1), C-factor-KUSLE (Case 2), and C-factor-K-factor (Case 3) for the years between 2002 and 2013 for a) TBW (n=89) and b) WBW (n=139).
The aforementioned clarification has been added to the text of the manuscript as Figure 4.
Figs 7 and 8. It seems that colors vary between 0 and a maximum that changes year per year. If this is correct, please uniform the colorbar in order to be able to evaluate the differences among various years. Also, the layout should be adjusted (i.e., huniform space between figures and between the colorbar and the table of values).
Response: The comment by the Reviewer is correct. The color bar is changed from a minimum (0 ton/ha) to a maximum (varying yearly) to be able to evaluate the differences among various years as suggested by the Reviewer. The layout is uniformly adjusted for Figures and Tables as suggested by the Reviewer.
Figure 7. Spatiotemporal distributions of the annual SY (ton/ha) in each HRU of TBW from 2002 to 2009 incorporating the dynamic models of C-factor and K-factor.
Figure 8. Spatiotemporal distributions of the annual SY (ton/ha) in each HRU of WBW from 2009 to 2013 incorporating the dynamic models of C-factor and K-factor.
The aforementioned clarification has been added to the text of the manuscript as Figures 7 and 8.
Finally, the authors mention several times the link of this research with irrigation management. I think this aspect should be properly explained in the manuscript.
Response: It is acknowledged. The unnecessary mentions about irrigation management have been deleted from the manuscript.
Author Response File: Author Response.docx
Reviewer 3 Report
- Line 208. There are many indexes mentioned with different spatial resolution. Please provide more information about the spatial resolution of meteorological data, and also specify the spatial resolution (HRU) used in this study.
- In this study, the simulation time scale is season or even month in the SWAT model. USLE model is mostly in more long term average (year). Please cite relative studies to support that the time scale is proper.
- The XY scatter graphs are recommended to show the relationship between the observed and simulated data in both Fig. 3 and 4.
Author Response
Manuscript ID: remotesensing-1499203
Title: A Union of Dynamic Hydrological Modeling and Satellite Remotely-Sensed Data for Spatiotemporal Assessment of Sediment Yields
Response to Reviewers:
We are very grateful to the reviewers for the thorough and insightful review. The comments and suggestions provided have contributed a great deal to improving the manuscript. According to these, we have made efforts in revising the manuscript, with the details explained as follows:
Reviewer 3 Comments:
Line 208. There are many indexes mentioned with different spatial resolution. Please provide more information about the spatial resolution of meteorological data, and also specify the spatial resolution (HRU) used in this study.
Response: The information about the spatial resolution of meteorological data and the spatial resolution (HRU) used in this study is included as suggested by the Reviewer.
Twelve years (2002-2013) of daily weather data were collected from ground weather stations as well as from downscaled and projected data from Climate.gov (Neitsch et al. 2005; NOAA, 2017) with a spatial resolution of 250 m. The spatial resolution of the HRUs in the study ranged from 9 km to 16 km with an annual temporal resolution.
The aforementioned clarification has been added to the text of the manuscript in the respective sections.
In this study, the simulation time scale is season or even month in the SWAT model. USLE model is mostly in more long term average (year). Please cite relative studies to support that the time scale is proper.
Response: Relative studies are cited to support that the time scale is proper as suggested by the Reviewer.
The SWAT model simulations were employed in the levels of seasons, months, and years in the study models. The soil loss predictions were estimated seasonally and monthly in the study watersheds as opposed to the conventional SY predictions on a long-term timespan. This is supported by the studies of Xu et al. (2009), Zhang et al. (2012), and Swami and Kulkarni (2016).
Swami V, Kulkarni S (2016) Simulation of runoff and sediment yield for a Kaneri watershed using SWAT model. Jounal of Geoscience and Environment Protection 4:1-15. doi: 10.4236/gep.2016.41001.
Xu ZX, Pang JP, Liu CM, Li JY (2009) Assessment of runoff and sediment yield in the miyun reservoir catchment by using swat model. Hydrol. Process. 23:3619–3630.
Zhang A, Zhang C, Fu G, Wang B, Bao Z, Zheng H (2012) Assessments of impacts of climate change and human activities on runoff with swat for the huifa river basin, northeast china. Water Resour. Manag. 26,:2199–2217.
The aforementioned clarification has been added to the text of the manuscript in section ‘3.1. SWAT Calibration and Validation Results’.
The XY scatter graphs are recommended to show the relationship between the observed and simulated data in both Fig. 3 and 4.
Response: The XY scatter graphs are revised to show the relationship between the observed and simulated data in both Figures 3 and 4 as suggested by the Reviewer.
Figure 3. The observed versus simulated stream flows in the calibration and validation periods of SWAT model for a) TBW (n=89) and b) WBW (n=139).
Figure 4. The observed versus predictions of SY using CUSLE-KUSLE (Case 1), C-factor-KUSLE (Case 2), and C-factor-K-factor (Case 3) for the years between 2002 and 2013 for a) TBW (n=89) and b) WBW (n=139).
The aforementioned clarification has been added to the text of the manuscript as Figure 3 and Figure 4.
Author Response File: Author Response.docx