VAWIlog: A Log-Transformed LSWI–EVI Index for Improved Surface Water Mapping in Agricultural Environments
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis paper proposes that the logarithmically transformed LSWI/EVI ratio (VAWIlog) effectively enhances the detection of surface water under vegetation cover, and by comparing the existing similar methods (e.g., LSWI-EVI difference, etc.), it is clear that the advantages of VAWIlog compared to other forms of transformation. however, there remain some questions.
1 Designing indices for mixed vegetation-water body scenarios for agricultural environments (rice paddies) fills in the gaps of traditional water body indices (e.g., NDWI). The generalizability of VAWIlog to non-agricultural environments (e.g., forested wetlands, shrub swamps) needs to be more discussed. Current validation is limited to rice fields, and extrapolation of conclusions needs to be done with caution.
2 Although the use of 100 sensors to obtain water level data provides adequate spatial and temporal coverage, sensor details are missing (e.g., field representativeness,) and need to be supplemented with schematic or textual descriptions
3 What is the basis for the statement that “the water table is considered to be free of water when it is 20 cm below the soil surface”?
4 Why LSWI was chosen as the basis instead of other water body indices (e.g., NDWI)
5 How to avoid overfitting the training data during grid search combined with balanced accuracy (BA) selection thresholds
6 Validated using samples, the spatial distribution of the samples is not described, additional spatial representativeness analysis is needed." No distinction was made between open and vegetated water bodies in the “with water” samples, which may confuse the assessment of indicator performance.
7 other minor comments
-Figure 3c Labeling text does not match legend (“right before harvest” vs. “maturity”)
-The specific application pathways of VAWIlog for irrigation management (Page 1) and GHG assessment (Page 17) should be emphasized rather than generalized.
-Dynamic Outlier Detection Algorithm (Page 4, Eq 1-3) Undefined window size (n=15) selection basis
Author Response
Dear Reviewer,
We sincerely appreciate the time and effort you devoted to reviewing our manuscript. Your thoughtful comments and suggestions have greatly contributed to improving the clarity, quality, and overall rigor of our work. We have carefully addressed each point raised, and we hope that our revisions and responses meet your expectations.
Please find our point-by-point responses to your comments below.
Comments 1: Designing indices for mixed vegetation-water body scenarios for agricultural environments (rice paddies) fills in the gaps of traditional water body indices (e.g., NDWI). The generalizability of VAWIlog to non-agricultural environments (e.g., forested wetlands, shrub swamps) needs to be more discussed. Current validation is limited to rice fields, and extrapolation of conclusions needs to be done with caution.
Response 1: Thank you for raising this important point regarding the applicability of VAWIlog beyond rice paddies. We fully acknowledge that the current validation is confined to agricultural settings, particularly rice fields. At this stage, we do not yet have access to ground-truth data in non-agricultural environments such as forested wetlands or shrub swamps. Therefore, we agree it would be premature to claim general applicability in those contexts.
To reflect this limitation and highlight future directions, we have added the following clarification in the second paragraph of the Conclusion (Lines 475–478):
"It is important to note, however, that the proposed index has thus far been formulated and validated primarily within rice paddy field environments. Future work is recommended to assess its applicability across other land cover types, including mangroves, shrubs, and non-rice agricultural systems.”
Comments 2: Although the use of 100 sensors to obtain water level data provides adequate spatial and temporal coverage, sensor details are missing (e.g., field representativeness,) and need to be supplemented with schematic or textual descriptions.
Response 2: Thank you for pointing this out. We agree that including more information on the water level sensors enhances transparency and reproducibility. In response, we have added a schematic diagram, now presented as Figure 2, illustrating the structure of the water level sensor and how the observation tube is installed in the field.
This figure is accompanied by a detailed caption and additional text describing the sensor design and its field deployment. We hope this addition provides greater clarity on the setup and promotes replicability by other researchers.
Comments 3: What is the basis for the statement that “the water table is considered to be free of water when it is 20 cm below the soil surface”?
Response 3: Thank you for highlighting this potential source of confusion. We would like to clarify that we do not claim the water table is “free of water” at 20 cm depth. Instead, the sensors are configured to measure water levels within the range of +10 cm (above soil) to -20 cm (below soil). This range is designed to capture fluctuations relevant to surface water presence in rice paddies, where 0 cm represents the soil surface.
The 20 cm depth does not imply an absence of water below this level, but rather reflects the limit of the measurement tube used in our study setup. Our definition of water presence and absence is strictly based on whether the water level is above or below the 0 cm soil surface reference.
Comments 4: Why LSWI was chosen as the basis instead of other water body indices (e.g., NDWI)
Response 4: Thank you for raising this point. We agree that the rationale for selecting LSWI as the base index needed to be more clearly explained. To address this, we added the following explanation in the fourth paragraph of the Introduction (Lines 85~89):
“LSWI was chosen as the basis for this development because, unlike other water indices, it incorporates information related to both soil moisture and vegetation water content. By adjusting LSWI with EVI, which reflects vegetation greenness and canopy density, the proposed VAWI aims to suppress the vegetation signal while retaining the soil moisture component. This enables more accurate detection of surface water obscured by vegetation.”
Comments 5: How to avoid overfitting the training data during grid search combined with balanced accuracy (BA) selection thresholds?
Response 5: Thank you for the insightful question. We would like to clarify that the thresholding was conducted using a simple one-dimensional grid search, as noted in Line 237. The process involves identifying the optimal threshold that maximizes separability between water presence and absence conditions, based on balanced accuracy.
This approach is widely used in surface water mapping (e.g., with NDWI or MNDWI), and by keeping the optimization limited to a single dimension, we reduce the risk of overfitting. Moreover, as shown in Figures 5, VAWIlog demonstrates consistent separability across different vegetation densities.In applications without ground-truth data, we recommend plotting VAWIlog versus EVI and identifying the linear division across the triangular pattern, which provides a practical and visually guided thresholding approach.
Comments 6: Validated using samples, the spatial distribution of the samples is not described, additional spatial representativeness analysis is needed." No distinction was made between open and vegetated water bodies in the “with water” samples, which may confuse the assessment of indicator performance.
Response 6: Thank you for this thoughtful and multi-part comment. We have addressed it in two ways:
(1) Spatial Representativeness:
To better demonstrate the spatial coverage, we have updated Figure 1b to include an elevation distribution plot. This addition reflects the terrain variability across sampling locations, which is relevant for understanding differences in water dynamics. Corresponding descriptions were added in Lines 115–120 to explain the significance of this variability.
(2) Open vs. Vegetated Water Distinction:
While we do not explicitly categorize samples into open and vegetated water bodies, we address this issue by analyzing index performance across varying vegetation conditions, as represented by EVI values. This approach captures the transition from bare to fully vegetated stages, offering an effective proxy for open versus vegetated water. As shown in Figures 5 and 9, VAWIlog consistently distinguishes water presence even under increasing vegetation cover.
We hope this explanation clarifies our methodology and rationale.
Comments 7a: Figure 3c Labeling text does not match legend (“right before harvest” vs. “maturity”)
Response 7a: Thank you for catching this inconsistency. We have revised the labeling in the now Figure 4c to read “right before harvest” to match the legend and maintain consistency.
Comments 7b: The specific application pathways of VAWIlog for irrigation management (Page 1) and GHG assessment (Page 17) should be emphasized rather than generalized.
Response 7b: We sincerely appreciate this excellent suggestion. We have revised the Conclusion section (Page 19) to more explicitly highlight how VAWIlog—along with the synergistic use of Landsat and Sentinel-2—can contribute to irrigation management and greenhouse gas (GHG) monitoring. This addition strengthens the practical relevance and broader impact of our findings.
Comments 7c: Dynamic Outlier Detection Algorithm (Page 4, Eq 1-3) Undefined window size (n=15) selection basis.
Response 7c: Thank you for pointing this out. To justify the window size selection, we added the following explanation (Lines 138–144):
“Given the 15–16 minute interval between consecutive measurements, this window spans approximately 3.75 to 4 hours of data. This duration is sufficient to capture short-term patterns while remaining sensitive to abrupt anomalies.”
The updated paragraph reads:
“For this process, a window size of n = 15 is selected to detect unnatural fluctuations in water levels. Given the 15–16 minute interval between consecutive measurements, this window spans approximately 3.75 to 4 hours of data. This duration is sufficient to capture short-term patterns while remaining sensitive to abrupt anomalies. Water levels, regardless of influences such as irrigation, rainfall, or drainage, are generally expected to follow a consistent rising or falling trend. By using a relatively small window, the algorithm conservatively flags outliers occurring over brief periods.”
Reviewer 2 Report
Comments and Suggestions for AuthorsThis manuscript presents interesting research on how to decouple the signals of vegetation and surface water with combinations and transformations of remote sensing based LSWI and EVI, which is useful for applications in rice growing regions. The MS is well written, and the results are supportive to the core opinion of the research. I have some comments below for the authors to improve and I suggest a minor revision:
- Figure 4 is the major standing point of the transformed index, could the authors show more sub figures, dividing the whole growing seasons of rice into four stages on transplanting, tillering, flowering and before harvest. The authors could discuss on each growing stage which transformation is most useful.
- Please move line 255~line 265 to the Introduction section, which is a review of current water indexes.
- The font in Figure 8 is too small, please increase them with revised MS.
- Please move line 319~327 to the data and method section.
- Line 396, please change the format of reference 45 to Author (year).
- Line 409, figure 10 should be figure 13.
Author Response
Dear Reviewer,
We sincerely appreciate the time and effort you devoted to reviewing our manuscript. Your thoughtful comments and suggestions have greatly contributed to improving the clarity, quality, and overall rigor of our work. We have carefully addressed each point raised, and we hope that our revisions and responses meet your expectations.
Please find our point-by-point responses to your comments below.
Comment 1: Figure 4 is the major standing point of the transformed index, could the authors show more sub figures, dividing the whole growing seasons of rice into four stages on transplanting, tillering, flowering and before harvest. The authors could discuss on each growing stage which transformation is most useful.
Response 1: Thank you very much for highlighting the importance of Figure 4 and for the valuable suggestion. While we fully appreciate the idea of analyzing the index transformations across the four rice growth stages, we respectfully decided not to pursue this approach for the following reasons:
-
Our aim is to evaluate the transformed indices based on their response to vegetation characteristics as captured by EVI, rather than on fixed phenological stages. This allows the method to remain generalizable across diverse rice-growing systems and geographic contexts, where the timing of phenological stages may vary.
-
Nevertheless, as illustrated in Figure 4, the proposed index, VAWIlog, consistently performs well across increasing EVI values, reliably distinguishing between water presence and absence.
We hope this explanation clarifies our reasoning and we sincerely appreciate your understanding.
Comment 2: Please move line 255~line 265 to the Introduction section, which is a review of current water indexes.
Response 2: Thank you for this helpful suggestion. We agree that the content in lines 255–265 fits better in the Introduction. In response, we have integrated this content into the third paragraph of the Introduction to improve flow and maintain consistency. The revised text can be found in lines 53–70 of the updated manuscript.
Comment 3: The font in Figure 8 is too small, please increase them with revised MS.
Response 3: Thank you for pointing this out. We have revised Figure 8 (now Figure 9) and increased the font size for better readability in the updated manuscript.
Comment 4: Please move line 319~327 to the data and method section.
Response 4: Thank you for the thoughtful suggestion. To address this, we have added a new subsection, Section 2.4: "Comparative analysis against established water indices and Dynamic World V1", within the Data and Methods section. This provides an overview of the indices used for benchmarking, along with background information on Dynamic World V1, and statistical significance testing done.
Comment 5: Line 396, please change the format of reference 45 to Author (year).
Response 5: Thank you for catching this formatting inconsistency. We agree that since the reference initiates the sentence, using the Author (year) format is more appropriate. We have revised the citation accordingly, while maintaining the numbered style required by the MDPI template.
Comment 6: Line 409, figure 10 should be figure 13.
Response 6: Thank you for identifying this error. We have corrected the figure reference from Figure 10 to (now Figure 15) in the revised manuscript.
Reviewer 3 Report
Comments and Suggestions for AuthorsThis is a very meaningful study. The authors propose a new water body extraction index and verify its reliability from multiple dimensions. The index can, to some extent, make up for the limitations of satellite optical remote sensing imagery in identifying water bodies under the canopy, and is expected to improve the accuracy of water body identification and has the potential to be promoted and applied on a larger spatial scale. The authors are requested to consider the following details:
The abstract lacks quantitative results. For example, the absolute accuracy of water body identification using the new index, and the degree of improvement in identification accuracy compared with traditional indices, etc.
Please provide product information about the water level sensors, such as brand, version, accuracy, data collection methods, etc., so that other scholars can repeat and verify the experimental results.
The Harmonized Landsat and Sentinel-2 (HLS) project lacks specific information on period, temporal resolution, number of bands, etc.
When comparing the differences between different indices, it is recommended to add significance tests to reflect the statistical significance of the comparison results.
Please supplement the quantitative results about the limitations of the index. For example, VAWIlog tends to underestimate when detecting open water bodies, especially rivers and reservoirs; in forest areas, VAWIlog tends to overestimate the existence of water bodies...
The authors suggest using NDWI to detect open water and water under tree canopies. What is the theoretical basis for this? Do all indices based on satellite optical sensors cannot currently identify water bodies under the canopy, and does NDWI just have relatively higher identification accuracy?
Author Response
Dear Reviewer,
We sincerely appreciate the time and effort you devoted to reviewing our manuscript. Your thoughtful comments and suggestions have greatly contributed to improving the clarity, quality, and overall rigor of our work. We have carefully addressed each point raised, and we hope that our revisions and responses meet your expectations.
Please find our point-by-point responses to your comments below.
Comment 1: The abstract lacks quantitative results. For example, the absolute accuracy of water body identification using the new index, and the degree of improvement in identification accuracy compared with traditional indices, etc.
Response 1: Thank you for this insightful comment. We agree that including quantitative results in the abstract is standard and helps establish the practical value of the proposed method. In response, we have revised the abstract to include performance metrics of VAWIlog, as shown in Lines 13–14:
"...with a balanced accuracy (BA) of 0.69 and a producer’s accuracy (PA) of 0.80. This reflects an average improvement of 25% over conventional methods.”
This addition provides a clearer summary of the index’s performance and improvement over existing approaches.
Comment 2: Please provide product information about the water level sensors, such as brand, version, accuracy, data collection methods, etc., so that other scholars can repeat and verify the experimental results.
Response 2: Thank you very much for this helpful suggestion. We fully agree that providing detailed information on the water level sensors is essential for ensuring reproducibility and transparency. In response, we have added relevant product details in Lines 121–124, including sensor specifications and accuracy. Additionally, we included a schematic illustration of the sensor setup, now presented as Figure 2, which shows the sensor components, observation well dimensions, and installation method.
We hope this addition facilitates replication and further adoption of the methodology by other researchers.
Comment 3: The Harmonized Landsat and Sentinel-2 (HLS) project lacks specific information on period, temporal resolution, number of bands, etc.
Response 3: Thank you for bringing this to our attention. Upon review, we found that while information on period and temporal resolution was already included, the number of spectral bands used was not explicitly stated. We have now added the following clarification in Lines 172–175:
“The spectral bands used in the study include the visible bands (Red, Green, and Blue), as well as the Near-Infrared (NIR) and Shortwave Infrared (SWIR1 and SWIR2) bands, which are commonly used in vegetation and water-related indices.”
The full description of the HLS dataset and its application in this study can be found in Lines 104–175.
Comment 4: When comparing the differences between different indices, it is recommended to add significance tests to reflect the statistical significance of the comparison results.
Response 4: Thank you very much for this meaningful suggestion. We agree that statistical testing strengthens the validity of our comparative analysis. In response, we conducted a Wilcoxon signed-rank test, a non-parametric paired test, to assess the significance of the performance differences between VAWIlog and well-established indices.
Details of the procedure are included in the newly added Section 2.4: Comparative Analysis against Established Water Indices and Dynamic World V1, and the results are presented in Figure 7b, with the following description added in Lines 324–325:
“Statistical testing using the Wilcoxon signed-rank test further confirmed that these improvements are statistically significant (p-value < 0.01), as shown in Figure 7b.”
Comment 5: Please supplement the quantitative results about the limitations of the index. For example, VAWIlog tends to underestimate when detecting open water bodies, especially rivers and reservoirs; in forest areas, VAWIlog tends to overestimate the existence of water bodies...
Response 5: Thank you for this important comment. We agree that it is essential to highlight the limitations of VAWIlog quantitatively. In response, we have added Figure 14, which illustrates the underestimation of VAWIlog in open water conditions and its overestimation in forested land covers (or trees), using NDWI as a reference. These patterns are also visually evident in Figure 13.
Supporting explanations and references to the new figure are provided in Lines 415–418.
Comment 6: The authors suggest using NDWI to detect open water and water under tree canopies. What is the theoretical basis for this? Do all indices based on satellite optical sensors cannot currently identify water bodies under the canopy, and does NDWI just have relatively higher identification accuracy?
Response 6: Thank you for raising this important point. We would like to clarify that NDWI is used in this study solely for demonstration purposes in open water mapping. The intention is not to claim that NDWI can more reliably detect open water compared to other indices, but rather to demonstrate a broader framework that combines indices with complementary strengths.
To avoid misunderstanding, we revised the manuscript to include the following clarification in Lines 445–450:
“Building on the observed behavior of VAWIlog and conventional open-water indices across different land cover types, we propose a combined approach that integrates VAWIlog, any established water index (such as NDWI, MNDWI, AWEI, WRI, or others), and the Dynamic World V1 dataset to produce more comprehensive surface water maps. While NDWI is used here for demonstration purposes, the framework is designed to accommodate any open-water index suited to the target landscape.”
This emphasizes the flexibility of the proposed method and clarifies that NDWI is not uniquely favored.
