Predictive Analysis of Hydrological Variables in the Cahaba Watershed: Enhancing Forecasting Accuracy for Water Resource Management Using Time-Series and Machine Learning Models

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

General Comments

This study presents a hybrid modeling approach that integrates the physically based SWAT model with the Prophet time-series model and a multi-output regression-based machine learning model to forecast hydrological variables in the Cahaba watershed under present and future climate scenarios. While the hybrid SWAT-Prophet-ML framework demonstrates potential for near-term prediction of seasonal variables such as ET and PET, the manuscript currently falls short in several areas related to scientific framing, methodological robustness, and model validation. I appreciate the authors’ efforts in combining data-driven and process-based approaches, which is an important direction in hydrological modeling. However, significant revisions are necessary before this manuscript can be considered for publication. My detailed comments are listed below.

 

Major Comments

1. The introduction lacks a clearly defined scientific problem and innovation point. While the authors provide a comprehensive background on SWAT, Prophet, and ML approaches, the introduction fails to articulate what specific research gap this study addresses and how the proposed hybrid model advances beyond existing literature. For example, the novelty of combining Prophet with ML for post-SWAT calibration must be framed in terms of added scientific value, not just technical integration.
2. The methodological rationale for integrating Prophet and machine learning with SWAT is underdeveloped. The study employs SWAT outputs to train Prophet for water balance variables and then applies machine learning on Prophet results to estimate hydrological responses. However, the logic behind this two-step architecture remains unclear. Why not directly forecast hydrological outputs with ML? What unique value does Prophet bring in this chain, particularly when its assumptions of periodicity and stationarity may not hold under future climate scenarios?
3. The machine learning component is not adequately justified or interpreted. The choice of using multi-output regression with polynomial features is presented without justification or comparison to other common methods in hydrology, such as random forest, gradient boosting, or LSTM networks. Moreover, the paper does not provide any feature importance analysis or SHAP values to understand model behavior. As a result, the predictive value remains a black box.
4. Evaluation metrics are limited and do not sufficiently assess extreme event capture. The study relies on RMSE, R², and NSE, which are useful but insufficient for assessing the performance of predictive models under extreme hydrological conditions. Additional evaluation metrics (e.g., peak error, bias in high-flow events, or categorical performance in extreme years such as 2011 and 2019) would help assess the model’s reliability for disaster planning.
5. Model performance under future climate scenarios is not convincingly validated. The authors acknowledge that SWAT-Prophet-ML performs poorly under future conditions, but the discussion lacks depth. The limitation is attributed to lack of training data, but no efforts are made to test solutions (e.g., training with scenario-based synthetic data, data augmentation, or physical constraints). The hybrid model’s scalability under changing regimes remains questionable.

 

Specific Comments

6. The x-axis of Figures 5, 6 is labeled from 0–40, which does not correspond to recognizable time units. Please label axes with actual years to facilitate interpretation.
7. Although the text mentions large uncertainty bounds in Prophet predictions, these are not visualized in the figures. Adding confidence intervals or error bands would enhance the transparency of model performance and allow readers to assess reliability.
8. Phrases such as “Prophet is a powerful tool” or “novel model” appear too frequently and are not academically objective. I recommend revising the text to adopt a more concise and formal tone. Consider professional editing or consultation with a native English speaker.
9. The conclusions reiterate previous content but do not clearly summarize the main findings or limitations. A more structured summary (e.g., bullet points or short paragraphs addressing present vs. future performance, variable-specific findings, and model limitations) is recommended.
10. Given that the model results hinge on customized implementation of Prophet and ML algorithms, code and input data (or at least SWAT outputs) should be made available via a repository (e.g., GitHub or Zenodo) to ensure transparency and reproducibility.

Author Response

Response to Reviewer’s Comments on Earth Manuscript - 3694442

 

The authors are grateful for the Reviewers and Editor for their insightful comments and recommendations.

 

 

1. Comments and Suggestions for Authors – Reviewer 1

General Comments

This study presents a hybrid modeling approach that integrates the physically based SWAT model with the Prophet time-series model and a multi-output regression-based machine learning model to forecast hydrological variables in the Cahaba watershed under present and future climate scenarios. While the hybrid SWAT-Prophet-ML framework demonstrates potential for near-term prediction of seasonal variables such as ET and PET, the manuscript currently falls short in several areas related to scientific framing, methodological robustness, and model validation. I appreciate the authors’ efforts in combining data-driven and process-based approaches, which is an important direction in hydrological modeling. However, significant revisions are necessary before this manuscript can be considered for publication. My detailed comments are listed below.

Response: The comments from the reviewer are acknowledged and revised accordingly. We thank the reviewer for recognizing our hybrid approach. We have clarified scientific framing, strengthened methodological explanations, and expanded model validation using calibrated SWAT outputs and scenario-based evaluations.

 

 

Major Comments

The introduction lacks a clearly defined scientific problem and innovation point. While the authors provide a comprehensive background on SWAT, Prophet, and ML approaches, the introduction fails to articulate what specific research gap this study addresses and how the proposed hybrid model advances beyond existing literature. For example, the novelty of combining Prophet with ML for post-SWAT calibration must be framed in terms of added scientific value, not just technical integration.

Response: The comments from the reviewer are acknowledged and revised accordingly. We thank the reviewer for this insightful comment. We have revised the introduction to clearly define the research gap, limited integration of time-series decomposition with physical models, and highlighted our innovation using Prophet to extract structured seasonal features from SWAT outputs, enhancing ML prediction of nonlinear hydrological responses. The section is as follows:

 

Despite the growing application of hybrid models in hydrological forecasting, there remains a critical gap in integrating physically based models with time-series decomposition techniques that retain seasonal structure while enabling nonlinear response estimation. Existing literature often applies machine learning models directly to raw hydrological data or SWAT outputs, which can lead to poor generalization, especially in the presence of high-frequency noise or complex seasonality (Lange et al., 2020). This study addresses this gap by introducing a novel SWAT-Prophet-ML framework that leverages the structural interpretability of Prophet for trend and seasonality extraction from SWAT-generated water balance variables and uses those outputs as features in a multi-output regression model with polynomial transformations to predict key hydrological responses. This two-stage decomposition regression pipeline enhances feature expressiveness while reducing the complexity typically faced in fully empirical models. The innovation lies not in the use of individual tools such as SWAT, Prophet, or ML but in the methodological synergy. Prophet helps transform temporally structured but noisy SWAT outputs into stable, decomposed series that improve machine learning performance in multi-variable regression. Furthermore, this model provides a foundation for modular extensions, such as scenario-based training for future climate adaptation, and allows for flexibility in physical-data integration (Zounemat-Kermani et al., 2021).

 

 

The methodological rationale for integrating Prophet and machine learning with SWAT is underdeveloped. The study employs SWAT outputs to train Prophet for water balance variables and then applies machine learning on Prophet results to estimate hydrological responses. However, the logic behind this two-step architecture remains unclear. Why not directly forecast hydrological outputs with ML? What unique value does Prophet bring in this chain, particularly when its assumptions of periodicity and stationarity may not hold under future climate scenarios?

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the reviewer’s observation. We have clarified that Prophet serves as a decomposition tool to extract trend and seasonality from SWAT outputs, reducing noise and enhancing the input structure for ML. This improves prediction stability and interpretability over directly applying ML to raw outputs. The section is as follows:

 

The Prophet model is used not for forecasting in isolation, but as a preprocessing mechanism to decompose these SWAT outputs into trend and seasonal components. This step offers two distinct advantages: (1) it reduces high-frequency variability and captures domain-relevant periodic behavior, and (2) it provides structured input features that enhance the performance and stability of the subsequent machine learning model. Direct use of raw SWAT outputs in regression often introduces noise and undermines model accuracy, particularly in capturing multi-output hydrological responses like surface runoff, groundwater contribution, or sediment yield. Finally, multi-output regression with polynomial features is employed to learn complex nonlinear mappings between decomposed water balance variables and hydrological responses. Polynomial transformations improve the representational capacity of the model without overfitting to noise, and multi-output regression preserves interdependence among output variables—a common challenge in traditional single-output ML models. This sequential approach—physics-informed simulation → structured decomposition → nonlinear learning—was chosen specifically to balance forecast stability, interpretability, and generalization, especially for applications in watershed-scale water resource planning.

 

 

The machine learning component is not adequately justified or interpreted. The choice of using multi-output regression with polynomial features is presented without justification or comparison to other common methods in hydrology, such as random forest, gradient boosting, or LSTM networks. Moreover, the paper does not provide any feature importance analysis or SHAP values to understand model behavior. As a result, the predictive value remains a black box.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the valuable feedback. We now justify the use of multi-output regression with polynomial features for its interpretability and reduced overfitting risk, and we have added future plans to benchmark against RF, GBM, and LSTM models with SHAP-based interpretability analysis. The section is as follows:

The polynomial regression model was chosen to provide a balance between model complexity, computational efficiency, and explanatory power, especially when used in conjunction with seasonal-decomposed Prophet outputs. As future work, we plan to extend this framework by benchmarking it against ensemble models (e.g., Random Forest, XGBoost) and sequence models (e.g., LSTM) and incorporating SHAP (SHapley Additive exPlanations) values or permutation-based feature importance to enhance model interpretability and quantify individual variable contributions to hydrological outputs. This will help transition the hybrid model from a predictive tool to a diagnostic and decision-support framework in climate-sensitive watershed management.

Importantly, to address model interpretability, a limitation in many black-box models, we will employ SHAP (SHapley Additive exPlanations) for tree-based models and attention-based visualizations for LSTM models to quantify feature influence on individual predictions. This approach aims to provide both predictive strength and domain interpretability, which are critical for informing climate-resilient water policy decisions. By benchmarking the existing polynomial model against these alternatives, we aim to validate the robustness of our proposed hybrid structure and identify cases where more advanced architectures yield substantial benefit over simpler regression approaches.

 

 

Evaluation metrics are limited and do not sufficiently assess extreme event capture. The study relies on RMSE, R², and NSE, which are useful but insufficient for assessing the performance of predictive models under extreme hydrological conditions. Additional evaluation metrics (e.g., peak error, bias in high-flow events, or categorical performance in extreme years such as 2011 and 2019) would help assess the model’s reliability for disaster planning.

Response: We acknowledge the reviewer’s concern about the limitations of RMSE, R², and NSE in capturing extreme hydrological events. However, these metrics are widely accepted in hydrological modeling for evaluating overall model performance across different flow conditions. Our study focuses on simulating continuous streamflow, not exclusively extreme events, making these metrics appropriate for establishing baseline performance. The evaluation period includes extreme years like 2011 and 2019, so the metrics still reflect model behavior under such conditions. Nonetheless, we recognize the value of additional extreme-event-focused metrics and have noted this as a direction for future research in the revised manuscript.

 

 

Model performance under future climate scenarios is not convincingly validated. The authors acknowledge that SWAT-Prophet-ML performs poorly under future conditions, but the discussion lacks depth. The limitation is attributed to lack of training data, but no efforts are made to test solutions (e.g., training with scenario-based synthetic data, data augmentation, or physical constraints). The hybrid model’s scalability under changing regimes remains questionable.

Response: The comments from the reviewer are acknowledged and revised accordingly.  We acknowledge this limitation and have included the section “4. Discussion and Future Work” to include future improvements such as scenario-based synthetic training, data augmentation, and incorporation of physical constraints to enhance the hybrid model’s generalization and scalability under changing climate regimes. The section details are as follows:

 

4. Discussion and Future Work

4.1. Enhancing Model Generalization Under Future Climate Regimes

The reduced performance of the SWAT-Prophet-ML framework under future climate scenarios, as discussed in Section 3.5.1, is primarily due to the model’s reliance on historical data distributions, which do not reflect the increased variability, shifts in precipitation patterns, and altered temperature regimes expected in future conditions. Prophet and polynomial regression are fundamentally trained on stationary trends, making them vulnerable to failure under non-stationary, out-of-distribution inputs.

To address this, we propose several enhancements to improve generalization under future conditions:

• Scenario-Based Synthetic Training: Future iterations of this model will incorporate SWAT-generated outputs under multiple Representative Concentration Pathway (RCP) scenarios as training inputs, expanding the model’s exposure to a broader range of climatic conditions.
• Data Augmentation: We plan to generate perturbed versions of climate input variables (e.g., rainfall intensity shifts, temperature anomalies) using Gaussian noise or bootstrapping methods to improve robustness to rare or extreme events.
• Physically Informed Constraints: Incorporating mass balance principles directly into the loss function or model architecture (e.g., conservation-aware ML or physics-guided neural networks) will ensure hydrologic plausibility, even when data distributions deviate from historical norms.
• Transfer Learning Techniques: Pretraining models on global climate-simulated datasets and fine-tuning on watershed-specific data could improve adaptability to novel climate regimes, especially in data-scarce basins.

These directions will not only strengthen the model’s applicability to changing climate scenarios but also improve its scalability for deployment in diverse hydrological settings globally.

 

4.2. Comparative Evaluation with Ensemble and Deep Learning Models  

While this study establishes the feasibility and accuracy of the SWAT-Prophet-ML framework using multi-output regression with polynomial features, future work will expand the modeling pipeline to include state-of-the-art ensemble and deep learning models such as Random Forest (RF), Gradient Boosting Machines (GBM), and Long Short-Term Memory (LSTM) networks. These models have been extensively validated in hydrology for their ability to capture high-dimensional nonlinear relationships and temporal dependencies (Mosavi et al., 2018; Ji et al., 2021). Comparative performance evaluation will be conducted using cross-validated RMSE, NSE, and R² metrics, and statistical significance of performance differences will be assessed using paired tests. Importantly, to address model interpretability, a limitation in many black-box models, we will employ SHAP (SHapley Additive exPlanations) for tree-based models and attention-based visualizations for LSTM models to quantify feature influence on individual predictions. This approach aims to provide both predictive strength and domain interpretability, which are critical for informing climate-resilient water policy decisions. By benchmarking the existing polynomial model against these alternatives, we aim to validate the robustness of our proposed hybrid structure and identify cases where more advanced architectures yield substantial benefit over simpler regression approaches.

 

4.3. Uncertainty in Prophet Forecasts and Practical Implications

The Prophet model inherently produces uncertainty estimates via 95% prediction intervals (PIs), which account for variability in the trend, seasonality, and model residuals. These intervals are especially useful for communicating the reliability of forecasts in real-world water resource management scenarios. In this study, prediction intervals were generated alongside point forecasts for key water balance variables, including precipitation, ET, PET, and snowmelt, across both present and future climate scenarios.

While the intervals offer valuable information about forecast spread and model confidence, several limitations affect their practical interpretability:

• Under present climate conditions, Prophet’s uncertainty remain relatively narrow for smooth variables like ET and PET, enhancing confidence in monthly water demand and planning decisions.
• However, for highly variable phenomena such as precipitation and snowmelt, especially under future climate projections (2030–2042), the uncertainty intervals widen substantially. This reflects not only inherent input variability but also the model’s inability to anticipate regime shifts or outliers outside the historical data distribution.
• Because the downstream ML model in the SWAT-Prophet-ML pipeline depends on Prophet outputs, errors or overconfident predictions from Prophet may propagate, potentially affecting the reliability of surface runoff, water yield, or sediment yield predictions.

From a practical standpoint, these wide uncertainty intervals reduce the confidence of hydrologists and decision-makers in using model outputs for fine-grained policy recommendations, especially for extreme event forecasting. For example, large uncertainty in precipitation forecasts may hinder reservoir operations or flood mitigation planning.

To address these issues, future work will incorporate:

• Quantile regression forests or Bayesian models to generate more robust uncertainty estimates.
• Calibration of Prophet's uncertainty intervals using historical residual validation;
• Monte Carlo dropout or ensemble-based simulations to propagate uncertainty through the full pipeline (SWAT → Prophet → ML);
• Expressing results not just as deterministic outputs but as probabilistic forecasts that better support risk-informed decision-making.

 

 

Specific Comments

The x-axis of Figures 5, 6 is labeled from 0–40, which does not correspond to recognizable time units. Please label axes with actual years to facilitate interpretation.

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have updated the x-axis in Figures 5 and 7 to display actual years for clearer interpretation.

 

 

Although the text mentions large uncertainty bounds in Prophet predictions, these are not visualized in the figures. Adding confidence intervals or error bands would enhance the transparency of model performance and allow readers to assess reliability.

Response: We thank the reviewer for the valuable comments which have been addressed in the revised manuscript. Figures 3 and 5 present the observed vs. predicted estimates of water balances and hydrological responses for the Cahaba River Basin under present climate conditions; therefore, error bars have been added as suggested. Figures 4 and 6 illustrate projected estimates under future climate scenarios, and error bars are not included here, as there are no observed values for direct comparison. However, confidence intervals have been added to all four figures (Figures 3–6) to provide a clearer representation, in line with the reviewer’s recommendation.

 

 

Phrases such as “Prophet is a powerful tool” or “novel model” appear too frequently and are not academically objective. I recommend revising the text to adopt a more concise and formal tone. Consider professional editing or consultation with a native English speaker.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have revised the manuscript to adopt a more concise and formal academic tone, reducing subjective phrases like “powerful tool” and “novel model” to ensure clarity and objectivity.

 

 

The conclusions reiterate previous content but do not clearly summarize the main findings or limitations. A more structured summary (e.g., bullet points or short paragraphs addressing present vs. future performance, variable-specific findings, and model limitations) is recommended.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have restructured the conclusions to clearly summarize key findings, differentiate present vs. future model performance, highlight variable-specific insights, and explicitly state model limitations and future improvement directions. The section is revised as follows:

5. Conclusions

This study proposed a hybrid modeling framework, SWAT-Prophet-ML, that integrates physically based hydrological simulation (SWAT), time-series decomposition (Prophet), and machine learning (multi-output polynomial regression) to forecast monthly water balance and hydrological response variables in the climate-sensitive Cahaba watershed.

Key Findings: Present Climate Scenario (2010–2022)

• Model demonstrated strong predictive performance for water yield (R² = 0.75), • surface runoff (R² = 0.70), evapotranspiration and potential evapotranspiration (RMSE = 15–20 mm)
• Accurate modeling of seasonal trends and smooth climatic behavior was achieved using the Prophet-decomposed features.
• Precipitation and snowmelt showed higher variability and were less accurately predicted (RMSE = 30–50 mm and 5–10 mm, respectively).

Key Findings: Future Climate Scenario (2030–2042)

• Model underperformed, especially for variables like groundwater flow and sediment yield, where the hybrid model failed to capture peak years or sharp shifts.
• The main cause was the model’s reliance on stationary historical patterns, which are not representative of future climate variability.

 

Model Limitations:

• Lack of training on non-stationary or perturbed climate data
• Absence of physical constraints in ML predictions
• Limited capacity to simulate extreme events or abrupt hydrological shifts

Future Work Directions:

• Scenario-based training using synthetic climate inputs from multiple representative concentration pathways
• Data augmentation techniques to simulate rare or extreme meteorological conditions
• Physically informed modeling, integrating hydrological constraints into the ML component
• Model benchmarking using ensemble and deep learning with SHAP-based feature interpretation

          This study provides a modular, semi-automated framework that bridges physical hydrological modeling and data-driven forecasting. While highly effective under historical climate conditions, it also highlights the importance of generalization strategies for adapting predictive models to future climate regimes. The work contributes replicable architecture for modern hydrological forecasting and offers a roadmap for advancing climate-resilient water resource management.

 

 

Given that the model results hinge on customized implementation of Prophet and ML algorithms, code and input data (or at least SWAT outputs) should be made available via a repository (e.g., GitHub or Zenodo) to ensure transparency and reproducibility.

Response: The comments from the reviewer are acknowledged and revised accordingly. We agree with the reviewer and will upload the code and SWAT output data to a public GitHub repository to support transparency and reproducibility.

 

 

 

 

 

2. Comments and Suggestions for Authors – Reviewer 2

Self-citations should be reduced. Include more external ML studies (e.g., Kratzert et al., 2019; Jiang et al., 2021) and deep learning approaches to enrich the literature review.

Response: The comments from the reviewer are acknowledged and revised accordingly. The number of self-citations is reduced in the manuscript. The following references and citations are included in the manuscript.

Lange, Holger, and Sebastian Sippel. "Machine learning applications in hydrology." Forest-water interactions. Cham: Springer International Publishing, 2020. 233-257.

Zounemat-Kermani, Mohammad, et al. "Ensemble machine learning paradigms in hydrology: A review." Journal of Hydrology 598 (2021): 126266.

Xu, Tianfang, and Feng Liang. "Machine learning for hydrologic sciences: An introductory overview." Wiley Interdisciplinary Reviews: Water 8.5 (2021): e1533.

Mosaffa, Hamidreza, et al. "Application of machine learning algorithms in hydrology." Computers in earth and environmental sciences. Elsevier, 2022. 585-591.

Yang, Tao, et al. "Evaluation and machine learning improvement of global hydrological model-based flood simulations." Environmental Research Letters 14.11 (2019): 114027.

Shen, Chaopeng, Xingyuan Chen, and Eric Laloy. "Broadening the use of machine learning in hydrology." Frontiers in Water 3 (2021): 681023.

Kim, Jungho, et al. "Hybrid machine learning framework for hydrological assessment." Journal of hydrology 577 (2019): 123913.

Petty, T. R., and P. Dhingra. "Streamflow hydrology estimate using machine learning (SHEM)." JAWRA Journal of the American Water Resources Association 54.1 (2018): 55-68.

Wang, Shuo, et al. "Analysis of runoff generation driving factors based on hydrological model and interpretable machine learning method." Journal of Hydrology: Regional Studies 42 (2022): 101139.

Rozos, Evangelos, Panayiotis Dimitriadis, and Vasilis Bellos. "Machine learning in assessing the performance of hydrological models." Hydrology 9.1 (2021): 5.

 

 

Add methodological transparency:

Response: The comments from the reviewer are acknowledged and revised accordingly.

We have added detailed descriptions of each modeling step to enhance methodological transparency and ensure reproducibility in section 2 of the manuscript.

 

 

Specify training/testing split,

Response: The comments from the reviewer are acknowledged and revised accordingly. We have specified the training/testing split as 80:20 and clarified its use in model evaluation. The section is as follows:

The machine learning component of the SWAT-Prophet-ML framework was trained and evaluated using a train-test split of 80:20, where 80% of the data was used for training and 20% was held out for testing. The dataset consisted of monthly hydrological values for the period 2010–2022, with Prophet-derived water balance variables (PRECIPmm, PETmm, ETmm, SNOWMELT_PRECIP_ratio) as inputs and the corresponding SWAT-based hydrological response variables (SURQmm, GW_Qmm, SWmm, WYLDmm, SYLDt_ha) as outputs.

To ensure robust model validation, a 5-fold cross-validation was performed on the training set during hyperparameter tuning. Model performance was assessed using RMSE, R², and Nash-Sutcliffe Efficiency (NSE) across both training and testing sets. Predictions were inverse transformed to their original scale using the MinMaxScaler to allow for direct comparison with actual hydrological outputs.

 

 

Clarify cross-validation technique,

Response: The comments from the reviewer are acknowledged and revised accordingly. We have clarified that 5-fold cross-validation was used during training to ensure robust model evaluation. The section is as follows:

To ensure robust model validation, a 5-fold cross-validation was performed on the training set during hyperparameter tuning. Model performance was assessed using RMSE, R², and Nash-Sutcliffe Efficiency (NSE) across both training and testing sets. Predictions were inverse transformed to their original scale using the MinMaxScaler to allow for direct comparison with actual hydrological outputs.

 

 

Mention any hyperparameter tuning strategies. Improve English clarity, especially in technical descriptions and interpretation of results.

Response: We appreciate the reviewer’s suggestion regarding hyperparameter tuning strategies and improved technical clarity. However, detailed hyperparameter optimization and extended interpretation of results are beyond the scope of the current study, which focuses primarily on evaluating the hydrological impacts under different climate scenarios using established model configurations. While basic model calibration was performed to ensure reasonable performance, comprehensive tuning was not the central aim. We agree that such analysis could enhance model precision and interpretability, and we suggest it as a valuable direction for future research. We have also checked and made sure that the clarity in the usage of English language is depicted well throughout the manuscript.

 

 

Discuss uncertainty from the Prophet model, including prediction intervals, and how it affects the model’s practical use.

Response: The comments from the reviewer are acknowledged and revised accordingly. We have added a discussion on Prophet’s prediction intervals, highlighting their role in conveying forecast uncertainty and their impact on the reliability of downstream hydrological predictions under the section “4.3. Uncertainty in Prophet Forecasts and Practical Implications”.

 

 

Future model extensions could include physics-informed ML or climate-driven data augmentation to improve generalization.

Response: The comments from the reviewer are acknowledged and revised accordingly. This is an insightful comment and suggestion. We acknowledge it and will continue to work with these areas in future research for SWAT ML Prophet based modeling. We agree and have outlined future extensions including physics-informed machine learning and climate-driven data augmentation to enhance model generalization under non-stationary conditions in section “4. Discussion and Future Work”.

 

 

 

 

 

3. Comments and Suggestions for Authors – Reviewer 3

This research has certain significance, but there are also the following problems:

1.There is no mapping display in the study area. A mapping display should be made for the Kahaba River Basin and the eight sub-basins it is divided into.

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have added a mapping display showing the Cahaba River Basin and its division into eight sub-basins to enhance spatial context and understanding.

Fig 2: (a) Map showing the Cahaba River and its major tributaries in central Alabama. The map indicates the Cahaba's location in the southeastern United States. Key historical and ecological localities are also highlighted; (b) Location of the Alabama River Basin covering parts of central and southern Alabama. The Cahaba River sub-basin lies within this system and serves as the focus of this study. 

 

2.In the Data and Methods section, tables should be used to introduce the sources of all the data used by the research institute along with the website addresses. Information on the spatio-temporal resolution of the data should also be provided

Response: The comments from the reviewer are acknowledged, revised accordingly, and added Table 1 as follows:

Table 1

Input data used in developing the Soil and Water Assessment Tool (SWAT) model for Cahaba River Basin.

Data	Data sources	Information	Period	Address/Location
Digital elevation models (DEM)	Web GIS	Raster, 30 m	2011	WebGIS - Geographic Information Systems Resource - GIS
Land use land cover	United States Geological Survey, USGS	Raster, 30 m	2011	Annual National Land Cover Database | U.S. Geological Survey
Soil data	United States Department of Agriculture, USDA	Raster, 60 m	2011	Web Soil Survey - Home
Climate data	Climate.gov	Daily	1980-2010 2010-2040	Search | Climate Data Online (CDO) | National Climatic Data Center (NCDC)
Hydrological data	United States Geological Survey, USGS	Monthly	2011-2017	Cahaba River at Trussville, Al. - USGS Water Data for the Nation

 

 

 

3.In the results section of this study, the simulated values of the SWAT model were used as the actual values to compare the results of the fusion model. However, the simulation accuracy of the SWAT model was not demonstrated. A subsection should be added at the very beginning of the result to verify the accuracy of the SWAT model using the collected actual observation data. When the output accuracy of some hydrological components of the SWAT model is relatively high, the credibility of all its components as actual values is relatively high.

Response: The comments from the reviewer are acknowledged and revised accordingly. A subsection 3.1 and Table 3 are included in the manuscript that verifies the accuracy of the SWAT model using the collected actual observation data which enhances the credibility of the existing model and novel models of the study.

3.1. SWAT Model Accuracy and Calibration Settings

SWAT was effectively applied to simulate hydrological responses within the Cahaba watershed under varying land use and climatic conditions. The model was initially configured using the 2011 land use/land cover and 1980–2010 and 2010-2040 climate data across eight sub-basins, 15 land use classes, and 30 soil categories. It was later calibrated using 2011 LULC under the same climate conditions. Hydrologic calibration and validation were conducted at the Trussville station in the upper Cahaba watershed. During calibration, wet conditions prevailed, and major peak flows observed in late March and early April were attributed to late snowmelt and spring runoff. The model demonstrated reasonable accuracy, achieving NSE and R2 values of 0.565 and 0.591 respectively. The coefficient of determination (R²) improved from 0.542 during calibration to 0.591 during validation, potentially due to the dominance of low streamflow events in the validation period, which reduced variability and increased correlation (Table 3). Key parameters such as ESCO (Soil Evaporation Compensation Factor) and CN2 (SCS Curve Number) were identified as highly sensitive to streamflow under wet conditions. Nutrient-related parameters (N_UPDIS, P_UPDIS) and urban erosion indicators (RILL_MULT, C_FACTOR) also showed significant influence on model outputs.

Table 3: SWAT model calibration parameters and model performance evaluation.

Parameter	Description	Calibration Range	Final Calibrated Value	NSE	R2
CN2	SCS curve number	1.00 – 2.00	1.63	0.430	0.456
ESCO	Soil evaporation compensation factor	0.85 – 1.00	0.91	0.465	0.483
P_UPDIS	Phosphorus uptake distribution	20 – 40	31	0.502	0.542
N_UPDIS	Nitrogen uptake distribution	20 – 40	24	0.565	0.591

 

 

 

 

4.It is recommended that the X-axis in Figures 5 and 6 be marked with the actual year and month instead of the months starting from 2010 and 2030

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have updated the x-axis in Figures 5 and 6 to display actual years for clearer interpretation.

 

 

5.I think that when organizing the results section, the water balance prediction and hydrological response prediction under the Present climate should be analyzed first, and then the water balance prediction and hydrological response under the future climate should be analyzed. That is, the organization sequence should be the existing sequence of sections 3.1, 3.3, 3.2, and 3.4.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have reorganized the section “3. Results” to follow the sequence 3.1, 3.3, 3.2, and 3.4, presenting water balance and hydrological response predictions under present climate first, followed by future climate analysis.

 

 

6.How exactly were the model performance evaluation results in Section 3.5 obtained? A detailed introduction should be given. If the simulation results are evaluated using actual observational data, why were the simulation results of the SWAT model taken as the actual values in the previous studies?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the comment. We have added clarification in Section 3.6.1 explaining that due to limited observational data across all variables and sub-basins, the calibrated SWAT model, previously validated with available observations, was used as a reference baseline for evaluating the hybrid model’s performance (“3.6.1 Evaluation Methodology and Use of SWAT as Reference”). A subsection “3.1. SWAT Model Accuracy and Calibration Settings” and Table 3 are also included in the manuscript that verifies the accuracy of the SWAT model using the collected actual observation data which enhances the credibility of the existing model and novel models of the study.

 

 

 

 

4. Comments and Suggestions for Authors – Reviewer 4

The article addresses water resource management by hydrological variable prediction

titled “Predictive Analysis of Hydrological Variables in the Cahaba Watershed: Enhancing

Forecasting Accuracy for Water Resource Management Using Time-Series and Machine

Learning Models.” Having familiarized myself with the manuscript, I have some

suggestions:

Major Comments:

1. In the introduction section, the authors should consider restructuring it to provide

general background information, specific background information, and a

description of the gap in our knowledge that the study was designed to fill.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have restructured the section “1.Introduction” to first present general background, then specific context on SWAT, Prophet, and ML methods, followed by a clear statement of the research gap and the objective our study aims to address.

 

 

2. Section 2.2, how did the authors account for vegetation heterogeneity with varying

spatial variability of soil moisture? How many layers does the SWAT model

consider? And how does it control this behaviour/pattern?

Response: We thank the reviewer for this insightful question. In the revised manuscript, we have expanded Section 2.2 to clarify how the Soil and Water Assessment Tool (SWAT) accounts for vegetation heterogeneity and the spatial variability of soil moisture.

SWAT handles spatial heterogeneity using Hydrologic Response Units (HRUs), which are unique combinations of land use, soil type, and slope within each sub-basin. This allows the model to simulate distinct vegetation and soil interactions within a watershed, thereby capturing the influence of vegetation heterogeneity on soil moisture dynamics. SWAT uses a multi-layer soil profile that typically includes 1 to 10 layers, with the number and depth of layers defined based on the soil input data. Each layer has its own physical properties such as texture, hydraulic conductivity, and available water capacity which influence water movement and storage. In our study, we used the first 4 layers as per the web soil survey-based soil database, ensuring adequate representation of vertical soil heterogeneity. Vegetation influences soil moisture through evapotranspiration and root zone depth. SWAT controls these processes through vegetation indices that vary by land cover type and soil type. Thus, areas with different vegetation types and growth conditions exhibit distinct soil moisture patterns captured by the HRUs of the Cahaba watershed.

 

 

3. Section 2.2, In watershed delineation, what threshold values for the soil, slope and

land use were used to define the hydrological response units/

Response: The comments from the reviewer are acknowledged and revised accordingly. In the revised manuscript, we have expanded Section 2.2 to clarify what threshold values for the soil, slope and land use were used to define the hydrological response units in watershed delineation.

The area threshold for watershed delineation is 1000 ha and the percentage threshold values for the soil, slope and land use that were used to define the hydrological response units are 10%, 10%, and 5% respectively.

 

 

4. Section 2.4: What are the limitations of using the Prophet model in capturing peak

events?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the question. We have addressed this in a new Section “3.2.1. Limitations in Capturing Peak Events” noting that Prophet’s additive model structure tends to smooth short-duration peaks such as heavy rainfall or snowmelt treating them as outliers, which limits its ability to accurately capture extreme hydrological events.

3.2.1. Limitations in Capturing Peak Events

While the Prophet model effectively decomposes time series into trend and seasonal components, it exhibits notable limitations in capturing sharp, short-duration peak events, such as heavy rainfall spikes or sudden snowmelt. This is largely due to Prophet’s underlying additive model structure, which assumes smooth and regular seasonal patterns. As a result, it tends to smooth over localized outliers, treating them as noise rather than meaningful extremes. Furthermore, Prophet assumes piecewise linear or logistic growth for the trend component and may fail to adapt to abrupt shifts or high-frequency variability unless such events are consistently present in the historical data. In the context of hydrological forecasting, this behavior limits the model’s ability to anticipate critical extreme events that significantly influence surface runoff, flash flooding, or sediment transport (Wamg et al., 2022). For instance, as shown in Figure 4(b) and 4(c), Prophet underestimates peak values during storm months, leading to under-propagation of signal amplitude into downstream ML predictions. Future model improvements will consider integrating spike-aware models, such as quantile regression or hybrid Prophet-LSTM structures, and incorporating event-based decomposition techniques to better preserve and forecast peak behaviors.

 

 

5. Section 2.6: Did the authors consider different CO2 concentration for baseline and

future projections? How many models was used for the study?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the question. In Section “2.6. Hydrological Scenarios”, we clarify that future climate projections were based on SWAT simulations driven by outputs under Representative Concentration Pathways (RCPs), which inherently account for varying CO₂ concentrations. One calibrated SWAT model was used, with scenario-based inputs applied for future simulations.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5. Comments and Suggestions for Authors – Reviewer 5

Evaluation of the article entitled “Predictive Analysis of Hydrological Variables in the Cahaba Watershed: Enhancing Forecasting Accuracy for Water Resource Management Using Time-Series and Machine Learning Models”, submitted to Earth.

 

 

The article addresses an important and current topic, with integration of SWAT with Prophet and machine learning for hydrological forecasting; It aims to predict environmental variables and their hydrological impacts in current and future climate scenarios; The SWAT-Prophet-ML model showed strong predictive performance for water production and surface runoff. This hybrid combination (model) achieved 86.73% accuracy in current climate forecasts, suggesting scalability for water resource planning.

 

General analysis: The article is very truncated and without a structured logical sequence. The figures have low resolution and quality. The main limitations are linked to the dependence on historical trends and poor performance in high-variability events, such as precipitation.

Response: The comments from the reviewer are acknowledged and revised accordingly.  Thank you for the feedback. We have revised the manuscript to improve logical flow and structure, replaced low-resolution figures with high-quality versions, and expanded the discussion on model limitations particularly its reliance on historical trends and reduced accuracy during high-variability events like precipitation.

 

 

- What are the objectives of the work? This is not clear in the structure of the article, so are the conclusions bad and need to be redone?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for your observation. We have revised the manuscript to clearly state the study’s objectives at the end of Section “2.1. Methodology Justification for Hybrid Model Design” and restructured the conclusions to directly address these objectives, summarizing key findings, limitations, and implications more clearly.

 

 

- In general, there was too much concern in presenting the models. Still, the understanding/discussion of the seasonality of the hydroclimatic variables is very weak, including associations with the studied river basin.

Response: We thank the reviewer for the valuable comment. We would like to respectfully clarify that the manuscript does include discussion of the seasonality of key hydroclimatic variables (precipitation, snowmelt, evapotranspiration, and potential evapotranspiration) and key hydrological response variables (surface runoff contribution to streamflow, groundwater contribution to streamflow, soil water content, water yield, and sediment yield) for Cahaba watershed. Specifically, seasonal trends were presented in Sections 2.6, 3.1, 3.2, 3.3 and Tables 4 and 5 as follows:

 

“2.6. Hydrological Scenarios 

This timeframe was selected to capture recent historical variability in climate and hydrological responses, including seasonal and interannual fluctuations in rainfall, snowmelt, ET, PET, surface runoff, groundwater flow, soil water content, water yield, and sediment yield.

 

3.1. SWAT vs. SWAT-Prophet-ML for Water Balance Predictions in Present Climate

Historical trends in environmental variables such as ET, PET, precipitation, and snowmelt show distinct seasonal patterns (Figure 3).

 

3.2. SWAT-Prophet-ML based Water Balance Predictions in Future Climate

From 2030 to 2042, projections of the four water balance components show strong seasonal  trends using the SWAT-Prophet-ML model (Figure 4). ET and PET maintain smooth, consistent  annual cycles, indicating high model reliability for temperature-driven processes. PET consistently exceeds ET by 50% - 65%, aligning with theoretical expectations as per the novel model outcomes. In future climate, precipitation ranging between 20 mm and 150 mm and snowmelt ranging between 0 mm and 29 mm exhibit more variability, with sharp peaks suggesting possible extreme weather events. Despite this, both retain regular annual patterns, reflecting the model's strength in capturing seasonality. The SWAT-Prophet-ML modeling results show that the future climate predictions of precipitation (20 mm – 150 mm) is decreasing relative to the present climate estimates of precipitation (40 mm-165 mm) whereas the future climate predictions of snowmelt (0 mm – 29 mm) is increasing relative to the present climate estimates of snowmelt (0 mm – 26 mm). This indicates that rainfall is likely to decrease in future with increases in snow melting and soil water accumulation (Labat et al., 2004). These patterns suggest stable climatic behavior over the projection period, though the seasonal fluctuations highlight areas for model refinement for the SWAT-Prophet-ML model (Tandon et al., 2025).

 

3.3. SWAT vs. SWAT-Prophet-ML for Hydrological Response Predictions in Present Climate

The predictions of surface runoff contribution to streamflow and water yields in both models are in good correlation with coefficient of determination, R2 values of 0.65 and 0.75 respectively. It also highlights significant runoff and water yield patterns in response to extreme precipitation events for the years 2011 and 2019 (Figure 3). These historical patterns have been instrumental in understanding seasonal variations and system responses (Preetha and Al-Hamdan 2020b; Preetha and Joseph 2025). This comparison highlights areas where the prediction model aligns closely with observed data and areas where it deviates, offering critical insights into the model's strengths and limitations.”

 

 

Suggestions for revisions:

- The article does not comply with the standards established in the MDPI Template for citations, references and other textual components.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for pointing this out. We have revised the manuscript to fully align with the MDPI template, ensuring correct formatting of citations, references, and all textual components as per journal guidelines.

 

 

- The figures are presented without previous calls and descriptions in the text; in general, they do not present the variables and units of measurement identified in the respective axes, they confuse “date” with “time”; the titles are not very explanatory.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the valuable feedback. We have revised the manuscript to ensure all figures are properly introduced and described in the text, updated axis labels with appropriate variables and units, corrected terminology (e.g., “date” vs. “time”), and improved figure titles for clarity and interpretability.

 

 

- there is no information, maps, climate, vegetation, soils or details of the physiographic characteristics of the studied river basin

Response: The comments from the reviewer are acknowledged and revised accordingly. We added Table 1 which shows the information about climate, vegetation, soils, and details of the physiographic characteristics of the river basin studied. Maps of the river basin are also included as Figure 1.

Table 1

Input data used in developing the Soil and Water Assessment Tool (SWAT) model for Cahaba River Basin.

Data	Data sources	Information	Period	Address/Location
Digital elevation models (DEM)	Web GIS	Raster, 30 m	2011	WebGIS - Geographic Information Systems Resource - GIS
Land use land cover	United States Geological Survey, USGS	Raster, 30 m	2011	Annual National Land Cover Database | U.S. Geological Survey
Soil data	United States Department of Agriculture, USDA	Raster, 60 m	2011	Web Soil Survey - Home
Climate data	Climate.gov	Daily	1980-2010 2010-2040	Search | Climate Data Online (CDO) | National Climatic Data Center (NCDC)
Hydrological data	United States Geological Survey, USGS	Monthly	2011-2017	Cahaba River at Trussville, Al. - USGS Water Data for the Nation

 

 

 

 

- The use of data only from 2011 for calibration of the SWAT is insufficient, since the model needs data for “warming up” and later, calibration and validation;

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the important observation. We acknowledge this limitation and have updated the methodology to clarify that additional years that were used for model warm-up, and that calibration and validation were performed using a multi-year dataset to ensure model stability and reliability.

Monthly calibration simulations were conducted from November 2013 to October 2017, with the first two years used as a warm-up period (NYSKIP = 2) under a skewed normal rainfall distribution. Streamflow data from the Trussville station, representing the outlet of sub-basin 1 (including HRUs 1–5), was used for calibration and validation.

 

 

- Sensitivity analyses of the SWAT parameters for the river basin were not presented.

Response: The comments from the reviewer are acknowledged and revised accordingly. A section “3.1. SWAT Model Accuracy and Calibration Settings” is added that shows the sensitivity of the SWAT parameters, the calibration ranges and the best fit values for the Cahaba watershed.

3.1. SWAT Model Accuracy and Calibration Settings

SWAT was effectively applied to simulate hydrological responses within the Cahaba watershed under varying land use and climatic conditions. The model was initially configured using the 2011 land use/land cover and 1980–2010 and 2010-2040 climate data across eight sub-basins, 15 land use classes, and 30 soil categories. It was later calibrated using 2011 LULC under the same climate conditions. Hydrologic calibration and validation were conducted at the Trussville station in the upper Cahaba watershed. During calibration, wet conditions prevailed, and major peak flows observed in late March and early April were attributed to late snowmelt and spring runoff. The model demonstrated reasonable accuracy, achieving NSE and R2 values of 0.565 and 0.591 respectively. The coefficient of determination (R²) improved from 0.542 during calibration to 0.591 during validation, potentially due to the dominance of low streamflow events in the validation period, which reduced variability and increased correlation (Table 3). Key parameters such as ESCO (Soil Evaporation Compensation Factor) and CN2 (SCS Curve Number) were identified as highly sensitive to streamflow under wet conditions. Nutrient-related parameters (N_UPDIS, P_UPDIS) and urban erosion indicators (RILL_MULT, C_FACTOR) also showed significant influence on model outputs.

 

Table 3: SWAT model calibration parameters and model performance evaluation.

Parameter	Description	Calibration Range	Final Calibrated Value	NSE	R2
CN2	SCS curve number	1.00 – 2.00	1.63	0.430	0.456
ESCO	Soil evaporation compensation factor	0.85 – 1.00	0.91	0.465	0.483
P_UPDIS	Phosphorus uptake distribution	20 – 40	31	0.502	0.542
N_UPDIS	Nitrogen uptake distribution	20 – 40	24	0.565	0.591

 

 

- In Figure 5, what happened to the GW-Qmm variable in the “current” variation? What does “time” mean on the X-axis of these figures, if the title says it is from 2010-2022?

Response: The comments from the reviewer are acknowledged and revised accordingly.  We appreciate the suggestion and have revised the figures to clearly show the trends of the variable GW-Qmm in the watershed. We have also updated the x-axis in the figure to display actual years for clearer interpretation.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Self-citations should be reduced. Include more external ML studies (e.g., Kratzert et al., 2019; Jiang et al., 2021) and deep learning approaches to enrich the literature review.

Add methodological transparency:

Specify training/testing split,

Clarify cross-validation technique,

Mention any hyperparameter tuning strategies.Improve English clarity, especially in technical descriptions and interpretation of results.

Discuss uncertainty from the Prophet model, including prediction intervals, and how it affects the model’s practical use.

Future model extensions could include physics-informed ML or climate-driven data augmentation to improve generalization.

Author Response

Response to Reviewer’s Comments on Earth Manuscript - 3694442

 

The authors are grateful for the Reviewers and Editor for their insightful comments and recommendations.

 

 

1. Comments and Suggestions for Authors – Reviewer 1

General Comments

This study presents a hybrid modeling approach that integrates the physically based SWAT model with the Prophet time-series model and a multi-output regression-based machine learning model to forecast hydrological variables in the Cahaba watershed under present and future climate scenarios. While the hybrid SWAT-Prophet-ML framework demonstrates potential for near-term prediction of seasonal variables such as ET and PET, the manuscript currently falls short in several areas related to scientific framing, methodological robustness, and model validation. I appreciate the authors’ efforts in combining data-driven and process-based approaches, which is an important direction in hydrological modeling. However, significant revisions are necessary before this manuscript can be considered for publication. My detailed comments are listed below.

Response: The comments from the reviewer are acknowledged and revised accordingly. We thank the reviewer for recognizing our hybrid approach. We have clarified scientific framing, strengthened methodological explanations, and expanded model validation using calibrated SWAT outputs and scenario-based evaluations.

 

 

Major Comments

The introduction lacks a clearly defined scientific problem and innovation point. While the authors provide a comprehensive background on SWAT, Prophet, and ML approaches, the introduction fails to articulate what specific research gap this study addresses and how the proposed hybrid model advances beyond existing literature. For example, the novelty of combining Prophet with ML for post-SWAT calibration must be framed in terms of added scientific value, not just technical integration.

Response: The comments from the reviewer are acknowledged and revised accordingly. We thank the reviewer for this insightful comment. We have revised the introduction to clearly define the research gap, limited integration of time-series decomposition with physical models, and highlighted our innovation using Prophet to extract structured seasonal features from SWAT outputs, enhancing ML prediction of nonlinear hydrological responses. The section is as follows:

 

Despite the growing application of hybrid models in hydrological forecasting, there remains a critical gap in integrating physically based models with time-series decomposition techniques that retain seasonal structure while enabling nonlinear response estimation. Existing literature often applies machine learning models directly to raw hydrological data or SWAT outputs, which can lead to poor generalization, especially in the presence of high-frequency noise or complex seasonality (Lange et al., 2020). This study addresses this gap by introducing a novel SWAT-Prophet-ML framework that leverages the structural interpretability of Prophet for trend and seasonality extraction from SWAT-generated water balance variables and uses those outputs as features in a multi-output regression model with polynomial transformations to predict key hydrological responses. This two-stage decomposition regression pipeline enhances feature expressiveness while reducing the complexity typically faced in fully empirical models. The innovation lies not in the use of individual tools such as SWAT, Prophet, or ML but in the methodological synergy. Prophet helps transform temporally structured but noisy SWAT outputs into stable, decomposed series that improve machine learning performance in multi-variable regression. Furthermore, this model provides a foundation for modular extensions, such as scenario-based training for future climate adaptation, and allows for flexibility in physical-data integration (Zounemat-Kermani et al., 2021).

 

 

The methodological rationale for integrating Prophet and machine learning with SWAT is underdeveloped. The study employs SWAT outputs to train Prophet for water balance variables and then applies machine learning on Prophet results to estimate hydrological responses. However, the logic behind this two-step architecture remains unclear. Why not directly forecast hydrological outputs with ML? What unique value does Prophet bring in this chain, particularly when its assumptions of periodicity and stationarity may not hold under future climate scenarios?

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the reviewer’s observation. We have clarified that Prophet serves as a decomposition tool to extract trend and seasonality from SWAT outputs, reducing noise and enhancing the input structure for ML. This improves prediction stability and interpretability over directly applying ML to raw outputs. The section is as follows:

 

The Prophet model is used not for forecasting in isolation, but as a preprocessing mechanism to decompose these SWAT outputs into trend and seasonal components. This step offers two distinct advantages: (1) it reduces high-frequency variability and captures domain-relevant periodic behavior, and (2) it provides structured input features that enhance the performance and stability of the subsequent machine learning model. Direct use of raw SWAT outputs in regression often introduces noise and undermines model accuracy, particularly in capturing multi-output hydrological responses like surface runoff, groundwater contribution, or sediment yield. Finally, multi-output regression with polynomial features is employed to learn complex nonlinear mappings between decomposed water balance variables and hydrological responses. Polynomial transformations improve the representational capacity of the model without overfitting to noise, and multi-output regression preserves interdependence among output variables—a common challenge in traditional single-output ML models. This sequential approach—physics-informed simulation → structured decomposition → nonlinear learning—was chosen specifically to balance forecast stability, interpretability, and generalization, especially for applications in watershed-scale water resource planning.

 

 

The machine learning component is not adequately justified or interpreted. The choice of using multi-output regression with polynomial features is presented without justification or comparison to other common methods in hydrology, such as random forest, gradient boosting, or LSTM networks. Moreover, the paper does not provide any feature importance analysis or SHAP values to understand model behavior. As a result, the predictive value remains a black box.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the valuable feedback. We now justify the use of multi-output regression with polynomial features for its interpretability and reduced overfitting risk, and we have added future plans to benchmark against RF, GBM, and LSTM models with SHAP-based interpretability analysis. The section is as follows:

The polynomial regression model was chosen to provide a balance between model complexity, computational efficiency, and explanatory power, especially when used in conjunction with seasonal-decomposed Prophet outputs. As future work, we plan to extend this framework by benchmarking it against ensemble models (e.g., Random Forest, XGBoost) and sequence models (e.g., LSTM) and incorporating SHAP (SHapley Additive exPlanations) values or permutation-based feature importance to enhance model interpretability and quantify individual variable contributions to hydrological outputs. This will help transition the hybrid model from a predictive tool to a diagnostic and decision-support framework in climate-sensitive watershed management.

Importantly, to address model interpretability, a limitation in many black-box models, we will employ SHAP (SHapley Additive exPlanations) for tree-based models and attention-based visualizations for LSTM models to quantify feature influence on individual predictions. This approach aims to provide both predictive strength and domain interpretability, which are critical for informing climate-resilient water policy decisions. By benchmarking the existing polynomial model against these alternatives, we aim to validate the robustness of our proposed hybrid structure and identify cases where more advanced architectures yield substantial benefit over simpler regression approaches.

 

 

Evaluation metrics are limited and do not sufficiently assess extreme event capture. The study relies on RMSE, R², and NSE, which are useful but insufficient for assessing the performance of predictive models under extreme hydrological conditions. Additional evaluation metrics (e.g., peak error, bias in high-flow events, or categorical performance in extreme years such as 2011 and 2019) would help assess the model’s reliability for disaster planning.

Response: We acknowledge the reviewer’s concern about the limitations of RMSE, R², and NSE in capturing extreme hydrological events. However, these metrics are widely accepted in hydrological modeling for evaluating overall model performance across different flow conditions. Our study focuses on simulating continuous streamflow, not exclusively extreme events, making these metrics appropriate for establishing baseline performance. The evaluation period includes extreme years like 2011 and 2019, so the metrics still reflect model behavior under such conditions. Nonetheless, we recognize the value of additional extreme-event-focused metrics and have noted this as a direction for future research in the revised manuscript.

 

 

Model performance under future climate scenarios is not convincingly validated. The authors acknowledge that SWAT-Prophet-ML performs poorly under future conditions, but the discussion lacks depth. The limitation is attributed to lack of training data, but no efforts are made to test solutions (e.g., training with scenario-based synthetic data, data augmentation, or physical constraints). The hybrid model’s scalability under changing regimes remains questionable.

Response: The comments from the reviewer are acknowledged and revised accordingly.  We acknowledge this limitation and have included the section “4. Discussion and Future Work” to include future improvements such as scenario-based synthetic training, data augmentation, and incorporation of physical constraints to enhance the hybrid model’s generalization and scalability under changing climate regimes. The section details are as follows:

 

4. Discussion and Future Work

4.1. Enhancing Model Generalization Under Future Climate Regimes

The reduced performance of the SWAT-Prophet-ML framework under future climate scenarios, as discussed in Section 3.5.1, is primarily due to the model’s reliance on historical data distributions, which do not reflect the increased variability, shifts in precipitation patterns, and altered temperature regimes expected in future conditions. Prophet and polynomial regression are fundamentally trained on stationary trends, making them vulnerable to failure under non-stationary, out-of-distribution inputs.

To address this, we propose several enhancements to improve generalization under future conditions:

• Scenario-Based Synthetic Training: Future iterations of this model will incorporate SWAT-generated outputs under multiple Representative Concentration Pathway (RCP) scenarios as training inputs, expanding the model’s exposure to a broader range of climatic conditions.
• Data Augmentation: We plan to generate perturbed versions of climate input variables (e.g., rainfall intensity shifts, temperature anomalies) using Gaussian noise or bootstrapping methods to improve robustness to rare or extreme events.
• Physically Informed Constraints: Incorporating mass balance principles directly into the loss function or model architecture (e.g., conservation-aware ML or physics-guided neural networks) will ensure hydrologic plausibility, even when data distributions deviate from historical norms.
• Transfer Learning Techniques: Pretraining models on global climate-simulated datasets and fine-tuning on watershed-specific data could improve adaptability to novel climate regimes, especially in data-scarce basins.

These directions will not only strengthen the model’s applicability to changing climate scenarios but also improve its scalability for deployment in diverse hydrological settings globally.

 

4.2. Comparative Evaluation with Ensemble and Deep Learning Models  

While this study establishes the feasibility and accuracy of the SWAT-Prophet-ML framework using multi-output regression with polynomial features, future work will expand the modeling pipeline to include state-of-the-art ensemble and deep learning models such as Random Forest (RF), Gradient Boosting Machines (GBM), and Long Short-Term Memory (LSTM) networks. These models have been extensively validated in hydrology for their ability to capture high-dimensional nonlinear relationships and temporal dependencies (Mosavi et al., 2018; Ji et al., 2021). Comparative performance evaluation will be conducted using cross-validated RMSE, NSE, and R² metrics, and statistical significance of performance differences will be assessed using paired tests. Importantly, to address model interpretability, a limitation in many black-box models, we will employ SHAP (SHapley Additive exPlanations) for tree-based models and attention-based visualizations for LSTM models to quantify feature influence on individual predictions. This approach aims to provide both predictive strength and domain interpretability, which are critical for informing climate-resilient water policy decisions. By benchmarking the existing polynomial model against these alternatives, we aim to validate the robustness of our proposed hybrid structure and identify cases where more advanced architectures yield substantial benefit over simpler regression approaches.

 

4.3. Uncertainty in Prophet Forecasts and Practical Implications

The Prophet model inherently produces uncertainty estimates via 95% prediction intervals (PIs), which account for variability in the trend, seasonality, and model residuals. These intervals are especially useful for communicating the reliability of forecasts in real-world water resource management scenarios. In this study, prediction intervals were generated alongside point forecasts for key water balance variables, including precipitation, ET, PET, and snowmelt, across both present and future climate scenarios.

While the intervals offer valuable information about forecast spread and model confidence, several limitations affect their practical interpretability:

• Under present climate conditions, Prophet’s uncertainty remain relatively narrow for smooth variables like ET and PET, enhancing confidence in monthly water demand and planning decisions.
• However, for highly variable phenomena such as precipitation and snowmelt, especially under future climate projections (2030–2042), the uncertainty intervals widen substantially. This reflects not only inherent input variability but also the model’s inability to anticipate regime shifts or outliers outside the historical data distribution.
• Because the downstream ML model in the SWAT-Prophet-ML pipeline depends on Prophet outputs, errors or overconfident predictions from Prophet may propagate, potentially affecting the reliability of surface runoff, water yield, or sediment yield predictions.

From a practical standpoint, these wide uncertainty intervals reduce the confidence of hydrologists and decision-makers in using model outputs for fine-grained policy recommendations, especially for extreme event forecasting. For example, large uncertainty in precipitation forecasts may hinder reservoir operations or flood mitigation planning.

To address these issues, future work will incorporate:

• Quantile regression forests or Bayesian models to generate more robust uncertainty estimates.
• Calibration of Prophet's uncertainty intervals using historical residual validation;
• Monte Carlo dropout or ensemble-based simulations to propagate uncertainty through the full pipeline (SWAT → Prophet → ML);
• Expressing results not just as deterministic outputs but as probabilistic forecasts that better support risk-informed decision-making.

 

 

Specific Comments

The x-axis of Figures 5, 6 is labeled from 0–40, which does not correspond to recognizable time units. Please label axes with actual years to facilitate interpretation.

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have updated the x-axis in Figures 5 and 7 to display actual years for clearer interpretation.

 

 

Although the text mentions large uncertainty bounds in Prophet predictions, these are not visualized in the figures. Adding confidence intervals or error bands would enhance the transparency of model performance and allow readers to assess reliability.

Response: We thank the reviewer for the valuable comments which have been addressed in the revised manuscript. Figures 3 and 5 present the observed vs. predicted estimates of water balances and hydrological responses for the Cahaba River Basin under present climate conditions; therefore, error bars have been added as suggested. Figures 4 and 6 illustrate projected estimates under future climate scenarios, and error bars are not included here, as there are no observed values for direct comparison. However, confidence intervals have been added to all four figures (Figures 3–6) to provide a clearer representation, in line with the reviewer’s recommendation.

 

 

Phrases such as “Prophet is a powerful tool” or “novel model” appear too frequently and are not academically objective. I recommend revising the text to adopt a more concise and formal tone. Consider professional editing or consultation with a native English speaker.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have revised the manuscript to adopt a more concise and formal academic tone, reducing subjective phrases like “powerful tool” and “novel model” to ensure clarity and objectivity.

 

 

The conclusions reiterate previous content but do not clearly summarize the main findings or limitations. A more structured summary (e.g., bullet points or short paragraphs addressing present vs. future performance, variable-specific findings, and model limitations) is recommended.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have restructured the conclusions to clearly summarize key findings, differentiate present vs. future model performance, highlight variable-specific insights, and explicitly state model limitations and future improvement directions. The section is revised as follows:

5. Conclusions

This study proposed a hybrid modeling framework, SWAT-Prophet-ML, that integrates physically based hydrological simulation (SWAT), time-series decomposition (Prophet), and machine learning (multi-output polynomial regression) to forecast monthly water balance and hydrological response variables in the climate-sensitive Cahaba watershed.

Key Findings: Present Climate Scenario (2010–2022)

• Model demonstrated strong predictive performance for water yield (R² = 0.75), • surface runoff (R² = 0.70), evapotranspiration and potential evapotranspiration (RMSE = 15–20 mm)
• Accurate modeling of seasonal trends and smooth climatic behavior was achieved using the Prophet-decomposed features.
• Precipitation and snowmelt showed higher variability and were less accurately predicted (RMSE = 30–50 mm and 5–10 mm, respectively).

Key Findings: Future Climate Scenario (2030–2042)

• Model underperformed, especially for variables like groundwater flow and sediment yield, where the hybrid model failed to capture peak years or sharp shifts.
• The main cause was the model’s reliance on stationary historical patterns, which are not representative of future climate variability.

 

Model Limitations:

• Lack of training on non-stationary or perturbed climate data
• Absence of physical constraints in ML predictions
• Limited capacity to simulate extreme events or abrupt hydrological shifts

Future Work Directions:

• Scenario-based training using synthetic climate inputs from multiple representative concentration pathways
• Data augmentation techniques to simulate rare or extreme meteorological conditions
• Physically informed modeling, integrating hydrological constraints into the ML component
• Model benchmarking using ensemble and deep learning with SHAP-based feature interpretation

          This study provides a modular, semi-automated framework that bridges physical hydrological modeling and data-driven forecasting. While highly effective under historical climate conditions, it also highlights the importance of generalization strategies for adapting predictive models to future climate regimes. The work contributes replicable architecture for modern hydrological forecasting and offers a roadmap for advancing climate-resilient water resource management.

 

 

Given that the model results hinge on customized implementation of Prophet and ML algorithms, code and input data (or at least SWAT outputs) should be made available via a repository (e.g., GitHub or Zenodo) to ensure transparency and reproducibility.

Response: The comments from the reviewer are acknowledged and revised accordingly. We agree with the reviewer and will upload the code and SWAT output data to a public GitHub repository to support transparency and reproducibility.

 

 

 

 

 

2. Comments and Suggestions for Authors – Reviewer 2

Self-citations should be reduced. Include more external ML studies (e.g., Kratzert et al., 2019; Jiang et al., 2021) and deep learning approaches to enrich the literature review.

Response: The comments from the reviewer are acknowledged and revised accordingly. The number of self-citations is reduced in the manuscript. The following references and citations are included in the manuscript.

Lange, Holger, and Sebastian Sippel. "Machine learning applications in hydrology." Forest-water interactions. Cham: Springer International Publishing, 2020. 233-257.

Zounemat-Kermani, Mohammad, et al. "Ensemble machine learning paradigms in hydrology: A review." Journal of Hydrology 598 (2021): 126266.

Xu, Tianfang, and Feng Liang. "Machine learning for hydrologic sciences: An introductory overview." Wiley Interdisciplinary Reviews: Water 8.5 (2021): e1533.

Mosaffa, Hamidreza, et al. "Application of machine learning algorithms in hydrology." Computers in earth and environmental sciences. Elsevier, 2022. 585-591.

Yang, Tao, et al. "Evaluation and machine learning improvement of global hydrological model-based flood simulations." Environmental Research Letters 14.11 (2019): 114027.

Shen, Chaopeng, Xingyuan Chen, and Eric Laloy. "Broadening the use of machine learning in hydrology." Frontiers in Water 3 (2021): 681023.

Kim, Jungho, et al. "Hybrid machine learning framework for hydrological assessment." Journal of hydrology 577 (2019): 123913.

Petty, T. R., and P. Dhingra. "Streamflow hydrology estimate using machine learning (SHEM)." JAWRA Journal of the American Water Resources Association 54.1 (2018): 55-68.

Wang, Shuo, et al. "Analysis of runoff generation driving factors based on hydrological model and interpretable machine learning method." Journal of Hydrology: Regional Studies 42 (2022): 101139.

Rozos, Evangelos, Panayiotis Dimitriadis, and Vasilis Bellos. "Machine learning in assessing the performance of hydrological models." Hydrology 9.1 (2021): 5.

 

 

Add methodological transparency:

Response: The comments from the reviewer are acknowledged and revised accordingly.

We have added detailed descriptions of each modeling step to enhance methodological transparency and ensure reproducibility in section 2 of the manuscript.

 

 

Specify training/testing split,

Response: The comments from the reviewer are acknowledged and revised accordingly. We have specified the training/testing split as 80:20 and clarified its use in model evaluation. The section is as follows:

The machine learning component of the SWAT-Prophet-ML framework was trained and evaluated using a train-test split of 80:20, where 80% of the data was used for training and 20% was held out for testing. The dataset consisted of monthly hydrological values for the period 2010–2022, with Prophet-derived water balance variables (PRECIPmm, PETmm, ETmm, SNOWMELT_PRECIP_ratio) as inputs and the corresponding SWAT-based hydrological response variables (SURQmm, GW_Qmm, SWmm, WYLDmm, SYLDt_ha) as outputs.

To ensure robust model validation, a 5-fold cross-validation was performed on the training set during hyperparameter tuning. Model performance was assessed using RMSE, R², and Nash-Sutcliffe Efficiency (NSE) across both training and testing sets. Predictions were inverse transformed to their original scale using the MinMaxScaler to allow for direct comparison with actual hydrological outputs.

 

 

Clarify cross-validation technique,

Response: The comments from the reviewer are acknowledged and revised accordingly. We have clarified that 5-fold cross-validation was used during training to ensure robust model evaluation. The section is as follows:

To ensure robust model validation, a 5-fold cross-validation was performed on the training set during hyperparameter tuning. Model performance was assessed using RMSE, R², and Nash-Sutcliffe Efficiency (NSE) across both training and testing sets. Predictions were inverse transformed to their original scale using the MinMaxScaler to allow for direct comparison with actual hydrological outputs.

 

 

Mention any hyperparameter tuning strategies. Improve English clarity, especially in technical descriptions and interpretation of results.

Response: We appreciate the reviewer’s suggestion regarding hyperparameter tuning strategies and improved technical clarity. However, detailed hyperparameter optimization and extended interpretation of results are beyond the scope of the current study, which focuses primarily on evaluating the hydrological impacts under different climate scenarios using established model configurations. While basic model calibration was performed to ensure reasonable performance, comprehensive tuning was not the central aim. We agree that such analysis could enhance model precision and interpretability, and we suggest it as a valuable direction for future research. We have also checked and made sure that the clarity in the usage of English language is depicted well throughout the manuscript.

 

 

Discuss uncertainty from the Prophet model, including prediction intervals, and how it affects the model’s practical use.

Response: The comments from the reviewer are acknowledged and revised accordingly. We have added a discussion on Prophet’s prediction intervals, highlighting their role in conveying forecast uncertainty and their impact on the reliability of downstream hydrological predictions under the section “4.3. Uncertainty in Prophet Forecasts and Practical Implications”.

 

 

Future model extensions could include physics-informed ML or climate-driven data augmentation to improve generalization.

Response: The comments from the reviewer are acknowledged and revised accordingly. This is an insightful comment and suggestion. We acknowledge it and will continue to work with these areas in future research for SWAT ML Prophet based modeling. We agree and have outlined future extensions including physics-informed machine learning and climate-driven data augmentation to enhance model generalization under non-stationary conditions in section “4. Discussion and Future Work”.

 

 

 

 

 

3. Comments and Suggestions for Authors – Reviewer 3

This research has certain significance, but there are also the following problems:

1.There is no mapping display in the study area. A mapping display should be made for the Kahaba River Basin and the eight sub-basins it is divided into.

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have added a mapping display showing the Cahaba River Basin and its division into eight sub-basins to enhance spatial context and understanding.

Fig 2: (a) Map showing the Cahaba River and its major tributaries in central Alabama. The map indicates the Cahaba's location in the southeastern United States. Key historical and ecological localities are also highlighted; (b) Location of the Alabama River Basin covering parts of central and southern Alabama. The Cahaba River sub-basin lies within this system and serves as the focus of this study. 

 

2.In the Data and Methods section, tables should be used to introduce the sources of all the data used by the research institute along with the website addresses. Information on the spatio-temporal resolution of the data should also be provided

Response: The comments from the reviewer are acknowledged, revised accordingly, and added Table 1 as follows:

Table 1

Input data used in developing the Soil and Water Assessment Tool (SWAT) model for Cahaba River Basin.

Data	Data sources	Information	Period	Address/Location
Digital elevation models (DEM)	Web GIS	Raster, 30 m	2011	WebGIS - Geographic Information Systems Resource - GIS
Land use land cover	United States Geological Survey, USGS	Raster, 30 m	2011	Annual National Land Cover Database | U.S. Geological Survey
Soil data	United States Department of Agriculture, USDA	Raster, 60 m	2011	Web Soil Survey - Home
Climate data	Climate.gov	Daily	1980-2010 2010-2040	Search | Climate Data Online (CDO) | National Climatic Data Center (NCDC)
Hydrological data	United States Geological Survey, USGS	Monthly	2011-2017	Cahaba River at Trussville, Al. - USGS Water Data for the Nation

 

 

 

3.In the results section of this study, the simulated values of the SWAT model were used as the actual values to compare the results of the fusion model. However, the simulation accuracy of the SWAT model was not demonstrated. A subsection should be added at the very beginning of the result to verify the accuracy of the SWAT model using the collected actual observation data. When the output accuracy of some hydrological components of the SWAT model is relatively high, the credibility of all its components as actual values is relatively high.

Response: The comments from the reviewer are acknowledged and revised accordingly. A subsection 3.1 and Table 3 are included in the manuscript that verifies the accuracy of the SWAT model using the collected actual observation data which enhances the credibility of the existing model and novel models of the study.

3.1. SWAT Model Accuracy and Calibration Settings

SWAT was effectively applied to simulate hydrological responses within the Cahaba watershed under varying land use and climatic conditions. The model was initially configured using the 2011 land use/land cover and 1980–2010 and 2010-2040 climate data across eight sub-basins, 15 land use classes, and 30 soil categories. It was later calibrated using 2011 LULC under the same climate conditions. Hydrologic calibration and validation were conducted at the Trussville station in the upper Cahaba watershed. During calibration, wet conditions prevailed, and major peak flows observed in late March and early April were attributed to late snowmelt and spring runoff. The model demonstrated reasonable accuracy, achieving NSE and R2 values of 0.565 and 0.591 respectively. The coefficient of determination (R²) improved from 0.542 during calibration to 0.591 during validation, potentially due to the dominance of low streamflow events in the validation period, which reduced variability and increased correlation (Table 3). Key parameters such as ESCO (Soil Evaporation Compensation Factor) and CN2 (SCS Curve Number) were identified as highly sensitive to streamflow under wet conditions. Nutrient-related parameters (N_UPDIS, P_UPDIS) and urban erosion indicators (RILL_MULT, C_FACTOR) also showed significant influence on model outputs.

Table 3: SWAT model calibration parameters and model performance evaluation.

Parameter	Description	Calibration Range	Final Calibrated Value	NSE	R2
CN2	SCS curve number	1.00 – 2.00	1.63	0.430	0.456
ESCO	Soil evaporation compensation factor	0.85 – 1.00	0.91	0.465	0.483
P_UPDIS	Phosphorus uptake distribution	20 – 40	31	0.502	0.542
N_UPDIS	Nitrogen uptake distribution	20 – 40	24	0.565	0.591

 

 

 

 

4.It is recommended that the X-axis in Figures 5 and 6 be marked with the actual year and month instead of the months starting from 2010 and 2030

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have updated the x-axis in Figures 5 and 6 to display actual years for clearer interpretation.

 

 

5.I think that when organizing the results section, the water balance prediction and hydrological response prediction under the Present climate should be analyzed first, and then the water balance prediction and hydrological response under the future climate should be analyzed. That is, the organization sequence should be the existing sequence of sections 3.1, 3.3, 3.2, and 3.4.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have reorganized the section “3. Results” to follow the sequence 3.1, 3.3, 3.2, and 3.4, presenting water balance and hydrological response predictions under present climate first, followed by future climate analysis.

 

 

6.How exactly were the model performance evaluation results in Section 3.5 obtained? A detailed introduction should be given. If the simulation results are evaluated using actual observational data, why were the simulation results of the SWAT model taken as the actual values in the previous studies?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the comment. We have added clarification in Section 3.6.1 explaining that due to limited observational data across all variables and sub-basins, the calibrated SWAT model, previously validated with available observations, was used as a reference baseline for evaluating the hybrid model’s performance (“3.6.1 Evaluation Methodology and Use of SWAT as Reference”). A subsection “3.1. SWAT Model Accuracy and Calibration Settings” and Table 3 are also included in the manuscript that verifies the accuracy of the SWAT model using the collected actual observation data which enhances the credibility of the existing model and novel models of the study.

 

 

 

 

4. Comments and Suggestions for Authors – Reviewer 4

The article addresses water resource management by hydrological variable prediction

titled “Predictive Analysis of Hydrological Variables in the Cahaba Watershed: Enhancing

Forecasting Accuracy for Water Resource Management Using Time-Series and Machine

Learning Models.” Having familiarized myself with the manuscript, I have some

suggestions:

Major Comments:

1. In the introduction section, the authors should consider restructuring it to provide

general background information, specific background information, and a

description of the gap in our knowledge that the study was designed to fill.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have restructured the section “1.Introduction” to first present general background, then specific context on SWAT, Prophet, and ML methods, followed by a clear statement of the research gap and the objective our study aims to address.

 

 

2. Section 2.2, how did the authors account for vegetation heterogeneity with varying

spatial variability of soil moisture? How many layers does the SWAT model

consider? And how does it control this behaviour/pattern?

Response: We thank the reviewer for this insightful question. In the revised manuscript, we have expanded Section 2.2 to clarify how the Soil and Water Assessment Tool (SWAT) accounts for vegetation heterogeneity and the spatial variability of soil moisture.

SWAT handles spatial heterogeneity using Hydrologic Response Units (HRUs), which are unique combinations of land use, soil type, and slope within each sub-basin. This allows the model to simulate distinct vegetation and soil interactions within a watershed, thereby capturing the influence of vegetation heterogeneity on soil moisture dynamics. SWAT uses a multi-layer soil profile that typically includes 1 to 10 layers, with the number and depth of layers defined based on the soil input data. Each layer has its own physical properties such as texture, hydraulic conductivity, and available water capacity which influence water movement and storage. In our study, we used the first 4 layers as per the web soil survey-based soil database, ensuring adequate representation of vertical soil heterogeneity. Vegetation influences soil moisture through evapotranspiration and root zone depth. SWAT controls these processes through vegetation indices that vary by land cover type and soil type. Thus, areas with different vegetation types and growth conditions exhibit distinct soil moisture patterns captured by the HRUs of the Cahaba watershed.

 

 

3. Section 2.2, In watershed delineation, what threshold values for the soil, slope and

land use were used to define the hydrological response units/

Response: The comments from the reviewer are acknowledged and revised accordingly. In the revised manuscript, we have expanded Section 2.2 to clarify what threshold values for the soil, slope and land use were used to define the hydrological response units in watershed delineation.

The area threshold for watershed delineation is 1000 ha and the percentage threshold values for the soil, slope and land use that were used to define the hydrological response units are 10%, 10%, and 5% respectively.

 

 

4. Section 2.4: What are the limitations of using the Prophet model in capturing peak

events?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the question. We have addressed this in a new Section “3.2.1. Limitations in Capturing Peak Events” noting that Prophet’s additive model structure tends to smooth short-duration peaks such as heavy rainfall or snowmelt treating them as outliers, which limits its ability to accurately capture extreme hydrological events.

3.2.1. Limitations in Capturing Peak Events

While the Prophet model effectively decomposes time series into trend and seasonal components, it exhibits notable limitations in capturing sharp, short-duration peak events, such as heavy rainfall spikes or sudden snowmelt. This is largely due to Prophet’s underlying additive model structure, which assumes smooth and regular seasonal patterns. As a result, it tends to smooth over localized outliers, treating them as noise rather than meaningful extremes. Furthermore, Prophet assumes piecewise linear or logistic growth for the trend component and may fail to adapt to abrupt shifts or high-frequency variability unless such events are consistently present in the historical data. In the context of hydrological forecasting, this behavior limits the model’s ability to anticipate critical extreme events that significantly influence surface runoff, flash flooding, or sediment transport (Wamg et al., 2022). For instance, as shown in Figure 4(b) and 4(c), Prophet underestimates peak values during storm months, leading to under-propagation of signal amplitude into downstream ML predictions. Future model improvements will consider integrating spike-aware models, such as quantile regression or hybrid Prophet-LSTM structures, and incorporating event-based decomposition techniques to better preserve and forecast peak behaviors.

 

 

5. Section 2.6: Did the authors consider different CO2 concentration for baseline and

future projections? How many models was used for the study?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the question. In Section “2.6. Hydrological Scenarios”, we clarify that future climate projections were based on SWAT simulations driven by outputs under Representative Concentration Pathways (RCPs), which inherently account for varying CO₂ concentrations. One calibrated SWAT model was used, with scenario-based inputs applied for future simulations.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5. Comments and Suggestions for Authors – Reviewer 5

Evaluation of the article entitled “Predictive Analysis of Hydrological Variables in the Cahaba Watershed: Enhancing Forecasting Accuracy for Water Resource Management Using Time-Series and Machine Learning Models”, submitted to Earth.

 

 

The article addresses an important and current topic, with integration of SWAT with Prophet and machine learning for hydrological forecasting; It aims to predict environmental variables and their hydrological impacts in current and future climate scenarios; The SWAT-Prophet-ML model showed strong predictive performance for water production and surface runoff. This hybrid combination (model) achieved 86.73% accuracy in current climate forecasts, suggesting scalability for water resource planning.

 

General analysis: The article is very truncated and without a structured logical sequence. The figures have low resolution and quality. The main limitations are linked to the dependence on historical trends and poor performance in high-variability events, such as precipitation.

Response: The comments from the reviewer are acknowledged and revised accordingly.  Thank you for the feedback. We have revised the manuscript to improve logical flow and structure, replaced low-resolution figures with high-quality versions, and expanded the discussion on model limitations particularly its reliance on historical trends and reduced accuracy during high-variability events like precipitation.

 

 

- What are the objectives of the work? This is not clear in the structure of the article, so are the conclusions bad and need to be redone?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for your observation. We have revised the manuscript to clearly state the study’s objectives at the end of Section “2.1. Methodology Justification for Hybrid Model Design” and restructured the conclusions to directly address these objectives, summarizing key findings, limitations, and implications more clearly.

 

 

- In general, there was too much concern in presenting the models. Still, the understanding/discussion of the seasonality of the hydroclimatic variables is very weak, including associations with the studied river basin.

Response: We thank the reviewer for the valuable comment. We would like to respectfully clarify that the manuscript does include discussion of the seasonality of key hydroclimatic variables (precipitation, snowmelt, evapotranspiration, and potential evapotranspiration) and key hydrological response variables (surface runoff contribution to streamflow, groundwater contribution to streamflow, soil water content, water yield, and sediment yield) for Cahaba watershed. Specifically, seasonal trends were presented in Sections 2.6, 3.1, 3.2, 3.3 and Tables 4 and 5 as follows:

 

“2.6. Hydrological Scenarios 

This timeframe was selected to capture recent historical variability in climate and hydrological responses, including seasonal and interannual fluctuations in rainfall, snowmelt, ET, PET, surface runoff, groundwater flow, soil water content, water yield, and sediment yield.

 

3.1. SWAT vs. SWAT-Prophet-ML for Water Balance Predictions in Present Climate

Historical trends in environmental variables such as ET, PET, precipitation, and snowmelt show distinct seasonal patterns (Figure 3).

 

3.2. SWAT-Prophet-ML based Water Balance Predictions in Future Climate

From 2030 to 2042, projections of the four water balance components show strong seasonal  trends using the SWAT-Prophet-ML model (Figure 4). ET and PET maintain smooth, consistent  annual cycles, indicating high model reliability for temperature-driven processes. PET consistently exceeds ET by 50% - 65%, aligning with theoretical expectations as per the novel model outcomes. In future climate, precipitation ranging between 20 mm and 150 mm and snowmelt ranging between 0 mm and 29 mm exhibit more variability, with sharp peaks suggesting possible extreme weather events. Despite this, both retain regular annual patterns, reflecting the model's strength in capturing seasonality. The SWAT-Prophet-ML modeling results show that the future climate predictions of precipitation (20 mm – 150 mm) is decreasing relative to the present climate estimates of precipitation (40 mm-165 mm) whereas the future climate predictions of snowmelt (0 mm – 29 mm) is increasing relative to the present climate estimates of snowmelt (0 mm – 26 mm). This indicates that rainfall is likely to decrease in future with increases in snow melting and soil water accumulation (Labat et al., 2004). These patterns suggest stable climatic behavior over the projection period, though the seasonal fluctuations highlight areas for model refinement for the SWAT-Prophet-ML model (Tandon et al., 2025).

 

3.3. SWAT vs. SWAT-Prophet-ML for Hydrological Response Predictions in Present Climate

The predictions of surface runoff contribution to streamflow and water yields in both models are in good correlation with coefficient of determination, R2 values of 0.65 and 0.75 respectively. It also highlights significant runoff and water yield patterns in response to extreme precipitation events for the years 2011 and 2019 (Figure 3). These historical patterns have been instrumental in understanding seasonal variations and system responses (Preetha and Al-Hamdan 2020b; Preetha and Joseph 2025). This comparison highlights areas where the prediction model aligns closely with observed data and areas where it deviates, offering critical insights into the model's strengths and limitations.”

 

 

Suggestions for revisions:

- The article does not comply with the standards established in the MDPI Template for citations, references and other textual components.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for pointing this out. We have revised the manuscript to fully align with the MDPI template, ensuring correct formatting of citations, references, and all textual components as per journal guidelines.

 

 

- The figures are presented without previous calls and descriptions in the text; in general, they do not present the variables and units of measurement identified in the respective axes, they confuse “date” with “time”; the titles are not very explanatory.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the valuable feedback. We have revised the manuscript to ensure all figures are properly introduced and described in the text, updated axis labels with appropriate variables and units, corrected terminology (e.g., “date” vs. “time”), and improved figure titles for clarity and interpretability.

 

 

- there is no information, maps, climate, vegetation, soils or details of the physiographic characteristics of the studied river basin

Response: The comments from the reviewer are acknowledged and revised accordingly. We added Table 1 which shows the information about climate, vegetation, soils, and details of the physiographic characteristics of the river basin studied. Maps of the river basin are also included as Figure 1.

Table 1

Input data used in developing the Soil and Water Assessment Tool (SWAT) model for Cahaba River Basin.

Data	Data sources	Information	Period	Address/Location
Digital elevation models (DEM)	Web GIS	Raster, 30 m	2011	WebGIS - Geographic Information Systems Resource - GIS
Land use land cover	United States Geological Survey, USGS	Raster, 30 m	2011	Annual National Land Cover Database | U.S. Geological Survey
Soil data	United States Department of Agriculture, USDA	Raster, 60 m	2011	Web Soil Survey - Home
Climate data	Climate.gov	Daily	1980-2010 2010-2040	Search | Climate Data Online (CDO) | National Climatic Data Center (NCDC)
Hydrological data	United States Geological Survey, USGS	Monthly	2011-2017	Cahaba River at Trussville, Al. - USGS Water Data for the Nation

 

 

 

 

- The use of data only from 2011 for calibration of the SWAT is insufficient, since the model needs data for “warming up” and later, calibration and validation;

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the important observation. We acknowledge this limitation and have updated the methodology to clarify that additional years that were used for model warm-up, and that calibration and validation were performed using a multi-year dataset to ensure model stability and reliability.

Monthly calibration simulations were conducted from November 2013 to October 2017, with the first two years used as a warm-up period (NYSKIP = 2) under a skewed normal rainfall distribution. Streamflow data from the Trussville station, representing the outlet of sub-basin 1 (including HRUs 1–5), was used for calibration and validation.

 

 

- Sensitivity analyses of the SWAT parameters for the river basin were not presented.

Response: The comments from the reviewer are acknowledged and revised accordingly. A section “3.1. SWAT Model Accuracy and Calibration Settings” is added that shows the sensitivity of the SWAT parameters, the calibration ranges and the best fit values for the Cahaba watershed.

3.1. SWAT Model Accuracy and Calibration Settings

SWAT was effectively applied to simulate hydrological responses within the Cahaba watershed under varying land use and climatic conditions. The model was initially configured using the 2011 land use/land cover and 1980–2010 and 2010-2040 climate data across eight sub-basins, 15 land use classes, and 30 soil categories. It was later calibrated using 2011 LULC under the same climate conditions. Hydrologic calibration and validation were conducted at the Trussville station in the upper Cahaba watershed. During calibration, wet conditions prevailed, and major peak flows observed in late March and early April were attributed to late snowmelt and spring runoff. The model demonstrated reasonable accuracy, achieving NSE and R2 values of 0.565 and 0.591 respectively. The coefficient of determination (R²) improved from 0.542 during calibration to 0.591 during validation, potentially due to the dominance of low streamflow events in the validation period, which reduced variability and increased correlation (Table 3). Key parameters such as ESCO (Soil Evaporation Compensation Factor) and CN2 (SCS Curve Number) were identified as highly sensitive to streamflow under wet conditions. Nutrient-related parameters (N_UPDIS, P_UPDIS) and urban erosion indicators (RILL_MULT, C_FACTOR) also showed significant influence on model outputs.

 

Table 3: SWAT model calibration parameters and model performance evaluation.

Parameter	Description	Calibration Range	Final Calibrated Value	NSE	R2
CN2	SCS curve number	1.00 – 2.00	1.63	0.430	0.456
ESCO	Soil evaporation compensation factor	0.85 – 1.00	0.91	0.465	0.483
P_UPDIS	Phosphorus uptake distribution	20 – 40	31	0.502	0.542
N_UPDIS	Nitrogen uptake distribution	20 – 40	24	0.565	0.591

 

 

- In Figure 5, what happened to the GW-Qmm variable in the “current” variation? What does “time” mean on the X-axis of these figures, if the title says it is from 2010-2022?

Response: The comments from the reviewer are acknowledged and revised accordingly.  We appreciate the suggestion and have revised the figures to clearly show the trends of the variable GW-Qmm in the watershed. We have also updated the x-axis in the figure to display actual years for clearer interpretation.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This research has certain significance, but there are also the following problems:

1.There is no mapping display in the study area. A mapping display should be made for the Kahaba River Basin and the eight sub-basins it is divided into.

2.In the Data and Methods section, tables should be used to introduce the sources of all the data used by the research institute along with the website addresses. Information on the spatio-temporal resolution of the data should also be provided.

3.In the results section of this study, the simulated values of the SWAT model were used as the actual values to compare the results of the fusion model. However, the simulation accuracy of the SWAT model was not demonstrated. A subsection should be added at the very beginning of the result to verify the accuracy of the SWAT model using the collected actual observation data. When the output accuracy of some hydrological components of the SWAT model is relatively high, the credibility of all its components as actual values is relatively high.

4.It is recommended that the X-axis in Figures 5 and 6 be marked with the actual year and month instead of the months starting from 2010 and 2030

5.I think that when organizing the results section, the water balance prediction and hydrological response prediction under the Present climate should be analyzed first, and then the water balance prediction and hydrological response under the future climate should be analyzed. That is, the organization sequence should be the existing sequence of sections 3.1, 3.3, 3.2, and 3.4.

6.How exactly were the model performance evaluation results in Section 3.5 obtained? A detailed introduction should be given. If the simulation results are evaluated using actual observational data, why were the simulation results of the SWAT model taken as the actual values in the previous studies?

Author Response

Response to Reviewer’s Comments on Earth Manuscript - 3694442

 

The authors are grateful for the Reviewers and Editor for their insightful comments and recommendations.

 

 

1. Comments and Suggestions for Authors – Reviewer 1

General Comments

This study presents a hybrid modeling approach that integrates the physically based SWAT model with the Prophet time-series model and a multi-output regression-based machine learning model to forecast hydrological variables in the Cahaba watershed under present and future climate scenarios. While the hybrid SWAT-Prophet-ML framework demonstrates potential for near-term prediction of seasonal variables such as ET and PET, the manuscript currently falls short in several areas related to scientific framing, methodological robustness, and model validation. I appreciate the authors’ efforts in combining data-driven and process-based approaches, which is an important direction in hydrological modeling. However, significant revisions are necessary before this manuscript can be considered for publication. My detailed comments are listed below.

Response: The comments from the reviewer are acknowledged and revised accordingly. We thank the reviewer for recognizing our hybrid approach. We have clarified scientific framing, strengthened methodological explanations, and expanded model validation using calibrated SWAT outputs and scenario-based evaluations.

 

 

Major Comments

The introduction lacks a clearly defined scientific problem and innovation point. While the authors provide a comprehensive background on SWAT, Prophet, and ML approaches, the introduction fails to articulate what specific research gap this study addresses and how the proposed hybrid model advances beyond existing literature. For example, the novelty of combining Prophet with ML for post-SWAT calibration must be framed in terms of added scientific value, not just technical integration.

Response: The comments from the reviewer are acknowledged and revised accordingly. We thank the reviewer for this insightful comment. We have revised the introduction to clearly define the research gap, limited integration of time-series decomposition with physical models, and highlighted our innovation using Prophet to extract structured seasonal features from SWAT outputs, enhancing ML prediction of nonlinear hydrological responses. The section is as follows:

 

Despite the growing application of hybrid models in hydrological forecasting, there remains a critical gap in integrating physically based models with time-series decomposition techniques that retain seasonal structure while enabling nonlinear response estimation. Existing literature often applies machine learning models directly to raw hydrological data or SWAT outputs, which can lead to poor generalization, especially in the presence of high-frequency noise or complex seasonality (Lange et al., 2020). This study addresses this gap by introducing a novel SWAT-Prophet-ML framework that leverages the structural interpretability of Prophet for trend and seasonality extraction from SWAT-generated water balance variables and uses those outputs as features in a multi-output regression model with polynomial transformations to predict key hydrological responses. This two-stage decomposition regression pipeline enhances feature expressiveness while reducing the complexity typically faced in fully empirical models. The innovation lies not in the use of individual tools such as SWAT, Prophet, or ML but in the methodological synergy. Prophet helps transform temporally structured but noisy SWAT outputs into stable, decomposed series that improve machine learning performance in multi-variable regression. Furthermore, this model provides a foundation for modular extensions, such as scenario-based training for future climate adaptation, and allows for flexibility in physical-data integration (Zounemat-Kermani et al., 2021).

 

 

The methodological rationale for integrating Prophet and machine learning with SWAT is underdeveloped. The study employs SWAT outputs to train Prophet for water balance variables and then applies machine learning on Prophet results to estimate hydrological responses. However, the logic behind this two-step architecture remains unclear. Why not directly forecast hydrological outputs with ML? What unique value does Prophet bring in this chain, particularly when its assumptions of periodicity and stationarity may not hold under future climate scenarios?

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the reviewer’s observation. We have clarified that Prophet serves as a decomposition tool to extract trend and seasonality from SWAT outputs, reducing noise and enhancing the input structure for ML. This improves prediction stability and interpretability over directly applying ML to raw outputs. The section is as follows:

 

The Prophet model is used not for forecasting in isolation, but as a preprocessing mechanism to decompose these SWAT outputs into trend and seasonal components. This step offers two distinct advantages: (1) it reduces high-frequency variability and captures domain-relevant periodic behavior, and (2) it provides structured input features that enhance the performance and stability of the subsequent machine learning model. Direct use of raw SWAT outputs in regression often introduces noise and undermines model accuracy, particularly in capturing multi-output hydrological responses like surface runoff, groundwater contribution, or sediment yield. Finally, multi-output regression with polynomial features is employed to learn complex nonlinear mappings between decomposed water balance variables and hydrological responses. Polynomial transformations improve the representational capacity of the model without overfitting to noise, and multi-output regression preserves interdependence among output variables—a common challenge in traditional single-output ML models. This sequential approach—physics-informed simulation → structured decomposition → nonlinear learning—was chosen specifically to balance forecast stability, interpretability, and generalization, especially for applications in watershed-scale water resource planning.

 

 

The machine learning component is not adequately justified or interpreted. The choice of using multi-output regression with polynomial features is presented without justification or comparison to other common methods in hydrology, such as random forest, gradient boosting, or LSTM networks. Moreover, the paper does not provide any feature importance analysis or SHAP values to understand model behavior. As a result, the predictive value remains a black box.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the valuable feedback. We now justify the use of multi-output regression with polynomial features for its interpretability and reduced overfitting risk, and we have added future plans to benchmark against RF, GBM, and LSTM models with SHAP-based interpretability analysis. The section is as follows:

The polynomial regression model was chosen to provide a balance between model complexity, computational efficiency, and explanatory power, especially when used in conjunction with seasonal-decomposed Prophet outputs. As future work, we plan to extend this framework by benchmarking it against ensemble models (e.g., Random Forest, XGBoost) and sequence models (e.g., LSTM) and incorporating SHAP (SHapley Additive exPlanations) values or permutation-based feature importance to enhance model interpretability and quantify individual variable contributions to hydrological outputs. This will help transition the hybrid model from a predictive tool to a diagnostic and decision-support framework in climate-sensitive watershed management.

Importantly, to address model interpretability, a limitation in many black-box models, we will employ SHAP (SHapley Additive exPlanations) for tree-based models and attention-based visualizations for LSTM models to quantify feature influence on individual predictions. This approach aims to provide both predictive strength and domain interpretability, which are critical for informing climate-resilient water policy decisions. By benchmarking the existing polynomial model against these alternatives, we aim to validate the robustness of our proposed hybrid structure and identify cases where more advanced architectures yield substantial benefit over simpler regression approaches.

 

 

Evaluation metrics are limited and do not sufficiently assess extreme event capture. The study relies on RMSE, R², and NSE, which are useful but insufficient for assessing the performance of predictive models under extreme hydrological conditions. Additional evaluation metrics (e.g., peak error, bias in high-flow events, or categorical performance in extreme years such as 2011 and 2019) would help assess the model’s reliability for disaster planning.

Response: We acknowledge the reviewer’s concern about the limitations of RMSE, R², and NSE in capturing extreme hydrological events. However, these metrics are widely accepted in hydrological modeling for evaluating overall model performance across different flow conditions. Our study focuses on simulating continuous streamflow, not exclusively extreme events, making these metrics appropriate for establishing baseline performance. The evaluation period includes extreme years like 2011 and 2019, so the metrics still reflect model behavior under such conditions. Nonetheless, we recognize the value of additional extreme-event-focused metrics and have noted this as a direction for future research in the revised manuscript.

 

 

Model performance under future climate scenarios is not convincingly validated. The authors acknowledge that SWAT-Prophet-ML performs poorly under future conditions, but the discussion lacks depth. The limitation is attributed to lack of training data, but no efforts are made to test solutions (e.g., training with scenario-based synthetic data, data augmentation, or physical constraints). The hybrid model’s scalability under changing regimes remains questionable.

Response: The comments from the reviewer are acknowledged and revised accordingly.  We acknowledge this limitation and have included the section “4. Discussion and Future Work” to include future improvements such as scenario-based synthetic training, data augmentation, and incorporation of physical constraints to enhance the hybrid model’s generalization and scalability under changing climate regimes. The section details are as follows:

 

4. Discussion and Future Work

4.1. Enhancing Model Generalization Under Future Climate Regimes

The reduced performance of the SWAT-Prophet-ML framework under future climate scenarios, as discussed in Section 3.5.1, is primarily due to the model’s reliance on historical data distributions, which do not reflect the increased variability, shifts in precipitation patterns, and altered temperature regimes expected in future conditions. Prophet and polynomial regression are fundamentally trained on stationary trends, making them vulnerable to failure under non-stationary, out-of-distribution inputs.

To address this, we propose several enhancements to improve generalization under future conditions:

• Scenario-Based Synthetic Training: Future iterations of this model will incorporate SWAT-generated outputs under multiple Representative Concentration Pathway (RCP) scenarios as training inputs, expanding the model’s exposure to a broader range of climatic conditions.
• Data Augmentation: We plan to generate perturbed versions of climate input variables (e.g., rainfall intensity shifts, temperature anomalies) using Gaussian noise or bootstrapping methods to improve robustness to rare or extreme events.
• Physically Informed Constraints: Incorporating mass balance principles directly into the loss function or model architecture (e.g., conservation-aware ML or physics-guided neural networks) will ensure hydrologic plausibility, even when data distributions deviate from historical norms.
• Transfer Learning Techniques: Pretraining models on global climate-simulated datasets and fine-tuning on watershed-specific data could improve adaptability to novel climate regimes, especially in data-scarce basins.

These directions will not only strengthen the model’s applicability to changing climate scenarios but also improve its scalability for deployment in diverse hydrological settings globally.

 

4.2. Comparative Evaluation with Ensemble and Deep Learning Models  

While this study establishes the feasibility and accuracy of the SWAT-Prophet-ML framework using multi-output regression with polynomial features, future work will expand the modeling pipeline to include state-of-the-art ensemble and deep learning models such as Random Forest (RF), Gradient Boosting Machines (GBM), and Long Short-Term Memory (LSTM) networks. These models have been extensively validated in hydrology for their ability to capture high-dimensional nonlinear relationships and temporal dependencies (Mosavi et al., 2018; Ji et al., 2021). Comparative performance evaluation will be conducted using cross-validated RMSE, NSE, and R² metrics, and statistical significance of performance differences will be assessed using paired tests. Importantly, to address model interpretability, a limitation in many black-box models, we will employ SHAP (SHapley Additive exPlanations) for tree-based models and attention-based visualizations for LSTM models to quantify feature influence on individual predictions. This approach aims to provide both predictive strength and domain interpretability, which are critical for informing climate-resilient water policy decisions. By benchmarking the existing polynomial model against these alternatives, we aim to validate the robustness of our proposed hybrid structure and identify cases where more advanced architectures yield substantial benefit over simpler regression approaches.

 

4.3. Uncertainty in Prophet Forecasts and Practical Implications

The Prophet model inherently produces uncertainty estimates via 95% prediction intervals (PIs), which account for variability in the trend, seasonality, and model residuals. These intervals are especially useful for communicating the reliability of forecasts in real-world water resource management scenarios. In this study, prediction intervals were generated alongside point forecasts for key water balance variables, including precipitation, ET, PET, and snowmelt, across both present and future climate scenarios.

While the intervals offer valuable information about forecast spread and model confidence, several limitations affect their practical interpretability:

• Under present climate conditions, Prophet’s uncertainty remain relatively narrow for smooth variables like ET and PET, enhancing confidence in monthly water demand and planning decisions.
• However, for highly variable phenomena such as precipitation and snowmelt, especially under future climate projections (2030–2042), the uncertainty intervals widen substantially. This reflects not only inherent input variability but also the model’s inability to anticipate regime shifts or outliers outside the historical data distribution.
• Because the downstream ML model in the SWAT-Prophet-ML pipeline depends on Prophet outputs, errors or overconfident predictions from Prophet may propagate, potentially affecting the reliability of surface runoff, water yield, or sediment yield predictions.

From a practical standpoint, these wide uncertainty intervals reduce the confidence of hydrologists and decision-makers in using model outputs for fine-grained policy recommendations, especially for extreme event forecasting. For example, large uncertainty in precipitation forecasts may hinder reservoir operations or flood mitigation planning.

To address these issues, future work will incorporate:

• Quantile regression forests or Bayesian models to generate more robust uncertainty estimates.
• Calibration of Prophet's uncertainty intervals using historical residual validation;
• Monte Carlo dropout or ensemble-based simulations to propagate uncertainty through the full pipeline (SWAT → Prophet → ML);
• Expressing results not just as deterministic outputs but as probabilistic forecasts that better support risk-informed decision-making.

 

 

Specific Comments

The x-axis of Figures 5, 6 is labeled from 0–40, which does not correspond to recognizable time units. Please label axes with actual years to facilitate interpretation.

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have updated the x-axis in Figures 5 and 7 to display actual years for clearer interpretation.

 

 

Although the text mentions large uncertainty bounds in Prophet predictions, these are not visualized in the figures. Adding confidence intervals or error bands would enhance the transparency of model performance and allow readers to assess reliability.

Response: We thank the reviewer for the valuable comments which have been addressed in the revised manuscript. Figures 3 and 5 present the observed vs. predicted estimates of water balances and hydrological responses for the Cahaba River Basin under present climate conditions; therefore, error bars have been added as suggested. Figures 4 and 6 illustrate projected estimates under future climate scenarios, and error bars are not included here, as there are no observed values for direct comparison. However, confidence intervals have been added to all four figures (Figures 3–6) to provide a clearer representation, in line with the reviewer’s recommendation.

 

 

Phrases such as “Prophet is a powerful tool” or “novel model” appear too frequently and are not academically objective. I recommend revising the text to adopt a more concise and formal tone. Consider professional editing or consultation with a native English speaker.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have revised the manuscript to adopt a more concise and formal academic tone, reducing subjective phrases like “powerful tool” and “novel model” to ensure clarity and objectivity.

 

 

The conclusions reiterate previous content but do not clearly summarize the main findings or limitations. A more structured summary (e.g., bullet points or short paragraphs addressing present vs. future performance, variable-specific findings, and model limitations) is recommended.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have restructured the conclusions to clearly summarize key findings, differentiate present vs. future model performance, highlight variable-specific insights, and explicitly state model limitations and future improvement directions. The section is revised as follows:

5. Conclusions

This study proposed a hybrid modeling framework, SWAT-Prophet-ML, that integrates physically based hydrological simulation (SWAT), time-series decomposition (Prophet), and machine learning (multi-output polynomial regression) to forecast monthly water balance and hydrological response variables in the climate-sensitive Cahaba watershed.

Key Findings: Present Climate Scenario (2010–2022)

• Model demonstrated strong predictive performance for water yield (R² = 0.75), • surface runoff (R² = 0.70), evapotranspiration and potential evapotranspiration (RMSE = 15–20 mm)
• Accurate modeling of seasonal trends and smooth climatic behavior was achieved using the Prophet-decomposed features.
• Precipitation and snowmelt showed higher variability and were less accurately predicted (RMSE = 30–50 mm and 5–10 mm, respectively).

Key Findings: Future Climate Scenario (2030–2042)

• Model underperformed, especially for variables like groundwater flow and sediment yield, where the hybrid model failed to capture peak years or sharp shifts.
• The main cause was the model’s reliance on stationary historical patterns, which are not representative of future climate variability.

 

Model Limitations:

• Lack of training on non-stationary or perturbed climate data
• Absence of physical constraints in ML predictions
• Limited capacity to simulate extreme events or abrupt hydrological shifts

Future Work Directions:

• Scenario-based training using synthetic climate inputs from multiple representative concentration pathways
• Data augmentation techniques to simulate rare or extreme meteorological conditions
• Physically informed modeling, integrating hydrological constraints into the ML component
• Model benchmarking using ensemble and deep learning with SHAP-based feature interpretation

          This study provides a modular, semi-automated framework that bridges physical hydrological modeling and data-driven forecasting. While highly effective under historical climate conditions, it also highlights the importance of generalization strategies for adapting predictive models to future climate regimes. The work contributes replicable architecture for modern hydrological forecasting and offers a roadmap for advancing climate-resilient water resource management.

 

 

Given that the model results hinge on customized implementation of Prophet and ML algorithms, code and input data (or at least SWAT outputs) should be made available via a repository (e.g., GitHub or Zenodo) to ensure transparency and reproducibility.

Response: The comments from the reviewer are acknowledged and revised accordingly. We agree with the reviewer and will upload the code and SWAT output data to a public GitHub repository to support transparency and reproducibility.

 

 

 

 

 

2. Comments and Suggestions for Authors – Reviewer 2

Self-citations should be reduced. Include more external ML studies (e.g., Kratzert et al., 2019; Jiang et al., 2021) and deep learning approaches to enrich the literature review.

Response: The comments from the reviewer are acknowledged and revised accordingly. The number of self-citations is reduced in the manuscript. The following references and citations are included in the manuscript.

Lange, Holger, and Sebastian Sippel. "Machine learning applications in hydrology." Forest-water interactions. Cham: Springer International Publishing, 2020. 233-257.

Zounemat-Kermani, Mohammad, et al. "Ensemble machine learning paradigms in hydrology: A review." Journal of Hydrology 598 (2021): 126266.

Xu, Tianfang, and Feng Liang. "Machine learning for hydrologic sciences: An introductory overview." Wiley Interdisciplinary Reviews: Water 8.5 (2021): e1533.

Mosaffa, Hamidreza, et al. "Application of machine learning algorithms in hydrology." Computers in earth and environmental sciences. Elsevier, 2022. 585-591.

Yang, Tao, et al. "Evaluation and machine learning improvement of global hydrological model-based flood simulations." Environmental Research Letters 14.11 (2019): 114027.

Shen, Chaopeng, Xingyuan Chen, and Eric Laloy. "Broadening the use of machine learning in hydrology." Frontiers in Water 3 (2021): 681023.

Kim, Jungho, et al. "Hybrid machine learning framework for hydrological assessment." Journal of hydrology 577 (2019): 123913.

Petty, T. R., and P. Dhingra. "Streamflow hydrology estimate using machine learning (SHEM)." JAWRA Journal of the American Water Resources Association 54.1 (2018): 55-68.

Wang, Shuo, et al. "Analysis of runoff generation driving factors based on hydrological model and interpretable machine learning method." Journal of Hydrology: Regional Studies 42 (2022): 101139.

Rozos, Evangelos, Panayiotis Dimitriadis, and Vasilis Bellos. "Machine learning in assessing the performance of hydrological models." Hydrology 9.1 (2021): 5.

 

 

Add methodological transparency:

Response: The comments from the reviewer are acknowledged and revised accordingly.

We have added detailed descriptions of each modeling step to enhance methodological transparency and ensure reproducibility in section 2 of the manuscript.

 

 

Specify training/testing split,

Response: The comments from the reviewer are acknowledged and revised accordingly. We have specified the training/testing split as 80:20 and clarified its use in model evaluation. The section is as follows:

The machine learning component of the SWAT-Prophet-ML framework was trained and evaluated using a train-test split of 80:20, where 80% of the data was used for training and 20% was held out for testing. The dataset consisted of monthly hydrological values for the period 2010–2022, with Prophet-derived water balance variables (PRECIPmm, PETmm, ETmm, SNOWMELT_PRECIP_ratio) as inputs and the corresponding SWAT-based hydrological response variables (SURQmm, GW_Qmm, SWmm, WYLDmm, SYLDt_ha) as outputs.

To ensure robust model validation, a 5-fold cross-validation was performed on the training set during hyperparameter tuning. Model performance was assessed using RMSE, R², and Nash-Sutcliffe Efficiency (NSE) across both training and testing sets. Predictions were inverse transformed to their original scale using the MinMaxScaler to allow for direct comparison with actual hydrological outputs.

 

 

Clarify cross-validation technique,

Response: The comments from the reviewer are acknowledged and revised accordingly. We have clarified that 5-fold cross-validation was used during training to ensure robust model evaluation. The section is as follows:

To ensure robust model validation, a 5-fold cross-validation was performed on the training set during hyperparameter tuning. Model performance was assessed using RMSE, R², and Nash-Sutcliffe Efficiency (NSE) across both training and testing sets. Predictions were inverse transformed to their original scale using the MinMaxScaler to allow for direct comparison with actual hydrological outputs.

 

 

Mention any hyperparameter tuning strategies. Improve English clarity, especially in technical descriptions and interpretation of results.

Response: We appreciate the reviewer’s suggestion regarding hyperparameter tuning strategies and improved technical clarity. However, detailed hyperparameter optimization and extended interpretation of results are beyond the scope of the current study, which focuses primarily on evaluating the hydrological impacts under different climate scenarios using established model configurations. While basic model calibration was performed to ensure reasonable performance, comprehensive tuning was not the central aim. We agree that such analysis could enhance model precision and interpretability, and we suggest it as a valuable direction for future research. We have also checked and made sure that the clarity in the usage of English language is depicted well throughout the manuscript.

 

 

Discuss uncertainty from the Prophet model, including prediction intervals, and how it affects the model’s practical use.

Response: The comments from the reviewer are acknowledged and revised accordingly. We have added a discussion on Prophet’s prediction intervals, highlighting their role in conveying forecast uncertainty and their impact on the reliability of downstream hydrological predictions under the section “4.3. Uncertainty in Prophet Forecasts and Practical Implications”.

 

 

Future model extensions could include physics-informed ML or climate-driven data augmentation to improve generalization.

Response: The comments from the reviewer are acknowledged and revised accordingly. This is an insightful comment and suggestion. We acknowledge it and will continue to work with these areas in future research for SWAT ML Prophet based modeling. We agree and have outlined future extensions including physics-informed machine learning and climate-driven data augmentation to enhance model generalization under non-stationary conditions in section “4. Discussion and Future Work”.

 

 

 

 

 

3. Comments and Suggestions for Authors – Reviewer 3

This research has certain significance, but there are also the following problems:

1.There is no mapping display in the study area. A mapping display should be made for the Kahaba River Basin and the eight sub-basins it is divided into.

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have added a mapping display showing the Cahaba River Basin and its division into eight sub-basins to enhance spatial context and understanding.

Fig 2: (a) Map showing the Cahaba River and its major tributaries in central Alabama. The map indicates the Cahaba's location in the southeastern United States. Key historical and ecological localities are also highlighted; (b) Location of the Alabama River Basin covering parts of central and southern Alabama. The Cahaba River sub-basin lies within this system and serves as the focus of this study. 

 

2.In the Data and Methods section, tables should be used to introduce the sources of all the data used by the research institute along with the website addresses. Information on the spatio-temporal resolution of the data should also be provided

Response: The comments from the reviewer are acknowledged, revised accordingly, and added Table 1 as follows:

Table 1

Input data used in developing the Soil and Water Assessment Tool (SWAT) model for Cahaba River Basin.

Data	Data sources	Information	Period	Address/Location
Digital elevation models (DEM)	Web GIS	Raster, 30 m	2011	WebGIS - Geographic Information Systems Resource - GIS
Land use land cover	United States Geological Survey, USGS	Raster, 30 m	2011	Annual National Land Cover Database | U.S. Geological Survey
Soil data	United States Department of Agriculture, USDA	Raster, 60 m	2011	Web Soil Survey - Home
Climate data	Climate.gov	Daily	1980-2010 2010-2040	Search | Climate Data Online (CDO) | National Climatic Data Center (NCDC)
Hydrological data	United States Geological Survey, USGS	Monthly	2011-2017	Cahaba River at Trussville, Al. - USGS Water Data for the Nation

 

 

 

3.In the results section of this study, the simulated values of the SWAT model were used as the actual values to compare the results of the fusion model. However, the simulation accuracy of the SWAT model was not demonstrated. A subsection should be added at the very beginning of the result to verify the accuracy of the SWAT model using the collected actual observation data. When the output accuracy of some hydrological components of the SWAT model is relatively high, the credibility of all its components as actual values is relatively high.

Response: The comments from the reviewer are acknowledged and revised accordingly. A subsection 3.1 and Table 3 are included in the manuscript that verifies the accuracy of the SWAT model using the collected actual observation data which enhances the credibility of the existing model and novel models of the study.

3.1. SWAT Model Accuracy and Calibration Settings

SWAT was effectively applied to simulate hydrological responses within the Cahaba watershed under varying land use and climatic conditions. The model was initially configured using the 2011 land use/land cover and 1980–2010 and 2010-2040 climate data across eight sub-basins, 15 land use classes, and 30 soil categories. It was later calibrated using 2011 LULC under the same climate conditions. Hydrologic calibration and validation were conducted at the Trussville station in the upper Cahaba watershed. During calibration, wet conditions prevailed, and major peak flows observed in late March and early April were attributed to late snowmelt and spring runoff. The model demonstrated reasonable accuracy, achieving NSE and R2 values of 0.565 and 0.591 respectively. The coefficient of determination (R²) improved from 0.542 during calibration to 0.591 during validation, potentially due to the dominance of low streamflow events in the validation period, which reduced variability and increased correlation (Table 3). Key parameters such as ESCO (Soil Evaporation Compensation Factor) and CN2 (SCS Curve Number) were identified as highly sensitive to streamflow under wet conditions. Nutrient-related parameters (N_UPDIS, P_UPDIS) and urban erosion indicators (RILL_MULT, C_FACTOR) also showed significant influence on model outputs.

Table 3: SWAT model calibration parameters and model performance evaluation.

Parameter	Description	Calibration Range	Final Calibrated Value	NSE	R2
CN2	SCS curve number	1.00 – 2.00	1.63	0.430	0.456
ESCO	Soil evaporation compensation factor	0.85 – 1.00	0.91	0.465	0.483
P_UPDIS	Phosphorus uptake distribution	20 – 40	31	0.502	0.542
N_UPDIS	Nitrogen uptake distribution	20 – 40	24	0.565	0.591

 

 

 

 

4.It is recommended that the X-axis in Figures 5 and 6 be marked with the actual year and month instead of the months starting from 2010 and 2030

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have updated the x-axis in Figures 5 and 6 to display actual years for clearer interpretation.

 

 

5.I think that when organizing the results section, the water balance prediction and hydrological response prediction under the Present climate should be analyzed first, and then the water balance prediction and hydrological response under the future climate should be analyzed. That is, the organization sequence should be the existing sequence of sections 3.1, 3.3, 3.2, and 3.4.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have reorganized the section “3. Results” to follow the sequence 3.1, 3.3, 3.2, and 3.4, presenting water balance and hydrological response predictions under present climate first, followed by future climate analysis.

 

 

6.How exactly were the model performance evaluation results in Section 3.5 obtained? A detailed introduction should be given. If the simulation results are evaluated using actual observational data, why were the simulation results of the SWAT model taken as the actual values in the previous studies?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the comment. We have added clarification in Section 3.6.1 explaining that due to limited observational data across all variables and sub-basins, the calibrated SWAT model, previously validated with available observations, was used as a reference baseline for evaluating the hybrid model’s performance (“3.6.1 Evaluation Methodology and Use of SWAT as Reference”). A subsection “3.1. SWAT Model Accuracy and Calibration Settings” and Table 3 are also included in the manuscript that verifies the accuracy of the SWAT model using the collected actual observation data which enhances the credibility of the existing model and novel models of the study.

 

 

 

 

4. Comments and Suggestions for Authors – Reviewer 4

The article addresses water resource management by hydrological variable prediction

titled “Predictive Analysis of Hydrological Variables in the Cahaba Watershed: Enhancing

Forecasting Accuracy for Water Resource Management Using Time-Series and Machine

Learning Models.” Having familiarized myself with the manuscript, I have some

suggestions:

Major Comments:

1. In the introduction section, the authors should consider restructuring it to provide

general background information, specific background information, and a

description of the gap in our knowledge that the study was designed to fill.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have restructured the section “1.Introduction” to first present general background, then specific context on SWAT, Prophet, and ML methods, followed by a clear statement of the research gap and the objective our study aims to address.

 

 

2. Section 2.2, how did the authors account for vegetation heterogeneity with varying

spatial variability of soil moisture? How many layers does the SWAT model

consider? And how does it control this behaviour/pattern?

Response: We thank the reviewer for this insightful question. In the revised manuscript, we have expanded Section 2.2 to clarify how the Soil and Water Assessment Tool (SWAT) accounts for vegetation heterogeneity and the spatial variability of soil moisture.

SWAT handles spatial heterogeneity using Hydrologic Response Units (HRUs), which are unique combinations of land use, soil type, and slope within each sub-basin. This allows the model to simulate distinct vegetation and soil interactions within a watershed, thereby capturing the influence of vegetation heterogeneity on soil moisture dynamics. SWAT uses a multi-layer soil profile that typically includes 1 to 10 layers, with the number and depth of layers defined based on the soil input data. Each layer has its own physical properties such as texture, hydraulic conductivity, and available water capacity which influence water movement and storage. In our study, we used the first 4 layers as per the web soil survey-based soil database, ensuring adequate representation of vertical soil heterogeneity. Vegetation influences soil moisture through evapotranspiration and root zone depth. SWAT controls these processes through vegetation indices that vary by land cover type and soil type. Thus, areas with different vegetation types and growth conditions exhibit distinct soil moisture patterns captured by the HRUs of the Cahaba watershed.

 

 

3. Section 2.2, In watershed delineation, what threshold values for the soil, slope and

land use were used to define the hydrological response units/

Response: The comments from the reviewer are acknowledged and revised accordingly. In the revised manuscript, we have expanded Section 2.2 to clarify what threshold values for the soil, slope and land use were used to define the hydrological response units in watershed delineation.

The area threshold for watershed delineation is 1000 ha and the percentage threshold values for the soil, slope and land use that were used to define the hydrological response units are 10%, 10%, and 5% respectively.

 

 

4. Section 2.4: What are the limitations of using the Prophet model in capturing peak

events?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the question. We have addressed this in a new Section “3.2.1. Limitations in Capturing Peak Events” noting that Prophet’s additive model structure tends to smooth short-duration peaks such as heavy rainfall or snowmelt treating them as outliers, which limits its ability to accurately capture extreme hydrological events.

3.2.1. Limitations in Capturing Peak Events

While the Prophet model effectively decomposes time series into trend and seasonal components, it exhibits notable limitations in capturing sharp, short-duration peak events, such as heavy rainfall spikes or sudden snowmelt. This is largely due to Prophet’s underlying additive model structure, which assumes smooth and regular seasonal patterns. As a result, it tends to smooth over localized outliers, treating them as noise rather than meaningful extremes. Furthermore, Prophet assumes piecewise linear or logistic growth for the trend component and may fail to adapt to abrupt shifts or high-frequency variability unless such events are consistently present in the historical data. In the context of hydrological forecasting, this behavior limits the model’s ability to anticipate critical extreme events that significantly influence surface runoff, flash flooding, or sediment transport (Wamg et al., 2022). For instance, as shown in Figure 4(b) and 4(c), Prophet underestimates peak values during storm months, leading to under-propagation of signal amplitude into downstream ML predictions. Future model improvements will consider integrating spike-aware models, such as quantile regression or hybrid Prophet-LSTM structures, and incorporating event-based decomposition techniques to better preserve and forecast peak behaviors.

 

 

5. Section 2.6: Did the authors consider different CO2 concentration for baseline and

future projections? How many models was used for the study?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the question. In Section “2.6. Hydrological Scenarios”, we clarify that future climate projections were based on SWAT simulations driven by outputs under Representative Concentration Pathways (RCPs), which inherently account for varying CO₂ concentrations. One calibrated SWAT model was used, with scenario-based inputs applied for future simulations.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5. Comments and Suggestions for Authors – Reviewer 5

Evaluation of the article entitled “Predictive Analysis of Hydrological Variables in the Cahaba Watershed: Enhancing Forecasting Accuracy for Water Resource Management Using Time-Series and Machine Learning Models”, submitted to Earth.

 

 

The article addresses an important and current topic, with integration of SWAT with Prophet and machine learning for hydrological forecasting; It aims to predict environmental variables and their hydrological impacts in current and future climate scenarios; The SWAT-Prophet-ML model showed strong predictive performance for water production and surface runoff. This hybrid combination (model) achieved 86.73% accuracy in current climate forecasts, suggesting scalability for water resource planning.

 

General analysis: The article is very truncated and without a structured logical sequence. The figures have low resolution and quality. The main limitations are linked to the dependence on historical trends and poor performance in high-variability events, such as precipitation.

Response: The comments from the reviewer are acknowledged and revised accordingly.  Thank you for the feedback. We have revised the manuscript to improve logical flow and structure, replaced low-resolution figures with high-quality versions, and expanded the discussion on model limitations particularly its reliance on historical trends and reduced accuracy during high-variability events like precipitation.

 

 

- What are the objectives of the work? This is not clear in the structure of the article, so are the conclusions bad and need to be redone?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for your observation. We have revised the manuscript to clearly state the study’s objectives at the end of Section “2.1. Methodology Justification for Hybrid Model Design” and restructured the conclusions to directly address these objectives, summarizing key findings, limitations, and implications more clearly.

 

 

- In general, there was too much concern in presenting the models. Still, the understanding/discussion of the seasonality of the hydroclimatic variables is very weak, including associations with the studied river basin.

Response: We thank the reviewer for the valuable comment. We would like to respectfully clarify that the manuscript does include discussion of the seasonality of key hydroclimatic variables (precipitation, snowmelt, evapotranspiration, and potential evapotranspiration) and key hydrological response variables (surface runoff contribution to streamflow, groundwater contribution to streamflow, soil water content, water yield, and sediment yield) for Cahaba watershed. Specifically, seasonal trends were presented in Sections 2.6, 3.1, 3.2, 3.3 and Tables 4 and 5 as follows:

 

“2.6. Hydrological Scenarios 

This timeframe was selected to capture recent historical variability in climate and hydrological responses, including seasonal and interannual fluctuations in rainfall, snowmelt, ET, PET, surface runoff, groundwater flow, soil water content, water yield, and sediment yield.

 

3.1. SWAT vs. SWAT-Prophet-ML for Water Balance Predictions in Present Climate

Historical trends in environmental variables such as ET, PET, precipitation, and snowmelt show distinct seasonal patterns (Figure 3).

 

3.2. SWAT-Prophet-ML based Water Balance Predictions in Future Climate

From 2030 to 2042, projections of the four water balance components show strong seasonal  trends using the SWAT-Prophet-ML model (Figure 4). ET and PET maintain smooth, consistent  annual cycles, indicating high model reliability for temperature-driven processes. PET consistently exceeds ET by 50% - 65%, aligning with theoretical expectations as per the novel model outcomes. In future climate, precipitation ranging between 20 mm and 150 mm and snowmelt ranging between 0 mm and 29 mm exhibit more variability, with sharp peaks suggesting possible extreme weather events. Despite this, both retain regular annual patterns, reflecting the model's strength in capturing seasonality. The SWAT-Prophet-ML modeling results show that the future climate predictions of precipitation (20 mm – 150 mm) is decreasing relative to the present climate estimates of precipitation (40 mm-165 mm) whereas the future climate predictions of snowmelt (0 mm – 29 mm) is increasing relative to the present climate estimates of snowmelt (0 mm – 26 mm). This indicates that rainfall is likely to decrease in future with increases in snow melting and soil water accumulation (Labat et al., 2004). These patterns suggest stable climatic behavior over the projection period, though the seasonal fluctuations highlight areas for model refinement for the SWAT-Prophet-ML model (Tandon et al., 2025).

 

3.3. SWAT vs. SWAT-Prophet-ML for Hydrological Response Predictions in Present Climate

The predictions of surface runoff contribution to streamflow and water yields in both models are in good correlation with coefficient of determination, R2 values of 0.65 and 0.75 respectively. It also highlights significant runoff and water yield patterns in response to extreme precipitation events for the years 2011 and 2019 (Figure 3). These historical patterns have been instrumental in understanding seasonal variations and system responses (Preetha and Al-Hamdan 2020b; Preetha and Joseph 2025). This comparison highlights areas where the prediction model aligns closely with observed data and areas where it deviates, offering critical insights into the model's strengths and limitations.”

 

 

Suggestions for revisions:

- The article does not comply with the standards established in the MDPI Template for citations, references and other textual components.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for pointing this out. We have revised the manuscript to fully align with the MDPI template, ensuring correct formatting of citations, references, and all textual components as per journal guidelines.

 

 

- The figures are presented without previous calls and descriptions in the text; in general, they do not present the variables and units of measurement identified in the respective axes, they confuse “date” with “time”; the titles are not very explanatory.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the valuable feedback. We have revised the manuscript to ensure all figures are properly introduced and described in the text, updated axis labels with appropriate variables and units, corrected terminology (e.g., “date” vs. “time”), and improved figure titles for clarity and interpretability.

 

 

- there is no information, maps, climate, vegetation, soils or details of the physiographic characteristics of the studied river basin

Response: The comments from the reviewer are acknowledged and revised accordingly. We added Table 1 which shows the information about climate, vegetation, soils, and details of the physiographic characteristics of the river basin studied. Maps of the river basin are also included as Figure 1.

Table 1

Input data used in developing the Soil and Water Assessment Tool (SWAT) model for Cahaba River Basin.

Data	Data sources	Information	Period	Address/Location
Digital elevation models (DEM)	Web GIS	Raster, 30 m	2011	WebGIS - Geographic Information Systems Resource - GIS
Land use land cover	United States Geological Survey, USGS	Raster, 30 m	2011	Annual National Land Cover Database | U.S. Geological Survey
Soil data	United States Department of Agriculture, USDA	Raster, 60 m	2011	Web Soil Survey - Home
Climate data	Climate.gov	Daily	1980-2010 2010-2040	Search | Climate Data Online (CDO) | National Climatic Data Center (NCDC)
Hydrological data	United States Geological Survey, USGS	Monthly	2011-2017	Cahaba River at Trussville, Al. - USGS Water Data for the Nation

 

 

 

 

- The use of data only from 2011 for calibration of the SWAT is insufficient, since the model needs data for “warming up” and later, calibration and validation;

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the important observation. We acknowledge this limitation and have updated the methodology to clarify that additional years that were used for model warm-up, and that calibration and validation were performed using a multi-year dataset to ensure model stability and reliability.

Monthly calibration simulations were conducted from November 2013 to October 2017, with the first two years used as a warm-up period (NYSKIP = 2) under a skewed normal rainfall distribution. Streamflow data from the Trussville station, representing the outlet of sub-basin 1 (including HRUs 1–5), was used for calibration and validation.

 

 

- Sensitivity analyses of the SWAT parameters for the river basin were not presented.

Response: The comments from the reviewer are acknowledged and revised accordingly. A section “3.1. SWAT Model Accuracy and Calibration Settings” is added that shows the sensitivity of the SWAT parameters, the calibration ranges and the best fit values for the Cahaba watershed.

3.1. SWAT Model Accuracy and Calibration Settings

SWAT was effectively applied to simulate hydrological responses within the Cahaba watershed under varying land use and climatic conditions. The model was initially configured using the 2011 land use/land cover and 1980–2010 and 2010-2040 climate data across eight sub-basins, 15 land use classes, and 30 soil categories. It was later calibrated using 2011 LULC under the same climate conditions. Hydrologic calibration and validation were conducted at the Trussville station in the upper Cahaba watershed. During calibration, wet conditions prevailed, and major peak flows observed in late March and early April were attributed to late snowmelt and spring runoff. The model demonstrated reasonable accuracy, achieving NSE and R2 values of 0.565 and 0.591 respectively. The coefficient of determination (R²) improved from 0.542 during calibration to 0.591 during validation, potentially due to the dominance of low streamflow events in the validation period, which reduced variability and increased correlation (Table 3). Key parameters such as ESCO (Soil Evaporation Compensation Factor) and CN2 (SCS Curve Number) were identified as highly sensitive to streamflow under wet conditions. Nutrient-related parameters (N_UPDIS, P_UPDIS) and urban erosion indicators (RILL_MULT, C_FACTOR) also showed significant influence on model outputs.

 

Table 3: SWAT model calibration parameters and model performance evaluation.

Parameter	Description	Calibration Range	Final Calibrated Value	NSE	R2
CN2	SCS curve number	1.00 – 2.00	1.63	0.430	0.456
ESCO	Soil evaporation compensation factor	0.85 – 1.00	0.91	0.465	0.483
P_UPDIS	Phosphorus uptake distribution	20 – 40	31	0.502	0.542
N_UPDIS	Nitrogen uptake distribution	20 – 40	24	0.565	0.591

 

 

- In Figure 5, what happened to the GW-Qmm variable in the “current” variation? What does “time” mean on the X-axis of these figures, if the title says it is from 2010-2022?

Response: The comments from the reviewer are acknowledged and revised accordingly.  We appreciate the suggestion and have revised the figures to clearly show the trends of the variable GW-Qmm in the watershed. We have also updated the x-axis in the figure to display actual years for clearer interpretation.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

See the attachment

Comments for author File: Comments.pdf

Author Response

Response to Reviewer’s Comments on Earth Manuscript - 3694442

 

The authors are grateful for the Reviewers and Editor for their insightful comments and recommendations.

 

 

1. Comments and Suggestions for Authors – Reviewer 1

General Comments

This study presents a hybrid modeling approach that integrates the physically based SWAT model with the Prophet time-series model and a multi-output regression-based machine learning model to forecast hydrological variables in the Cahaba watershed under present and future climate scenarios. While the hybrid SWAT-Prophet-ML framework demonstrates potential for near-term prediction of seasonal variables such as ET and PET, the manuscript currently falls short in several areas related to scientific framing, methodological robustness, and model validation. I appreciate the authors’ efforts in combining data-driven and process-based approaches, which is an important direction in hydrological modeling. However, significant revisions are necessary before this manuscript can be considered for publication. My detailed comments are listed below.

Response: The comments from the reviewer are acknowledged and revised accordingly. We thank the reviewer for recognizing our hybrid approach. We have clarified scientific framing, strengthened methodological explanations, and expanded model validation using calibrated SWAT outputs and scenario-based evaluations.

 

 

Major Comments

The introduction lacks a clearly defined scientific problem and innovation point. While the authors provide a comprehensive background on SWAT, Prophet, and ML approaches, the introduction fails to articulate what specific research gap this study addresses and how the proposed hybrid model advances beyond existing literature. For example, the novelty of combining Prophet with ML for post-SWAT calibration must be framed in terms of added scientific value, not just technical integration.

Response: The comments from the reviewer are acknowledged and revised accordingly. We thank the reviewer for this insightful comment. We have revised the introduction to clearly define the research gap, limited integration of time-series decomposition with physical models, and highlighted our innovation using Prophet to extract structured seasonal features from SWAT outputs, enhancing ML prediction of nonlinear hydrological responses. The section is as follows:

 

Despite the growing application of hybrid models in hydrological forecasting, there remains a critical gap in integrating physically based models with time-series decomposition techniques that retain seasonal structure while enabling nonlinear response estimation. Existing literature often applies machine learning models directly to raw hydrological data or SWAT outputs, which can lead to poor generalization, especially in the presence of high-frequency noise or complex seasonality (Lange et al., 2020). This study addresses this gap by introducing a novel SWAT-Prophet-ML framework that leverages the structural interpretability of Prophet for trend and seasonality extraction from SWAT-generated water balance variables and uses those outputs as features in a multi-output regression model with polynomial transformations to predict key hydrological responses. This two-stage decomposition regression pipeline enhances feature expressiveness while reducing the complexity typically faced in fully empirical models. The innovation lies not in the use of individual tools such as SWAT, Prophet, or ML but in the methodological synergy. Prophet helps transform temporally structured but noisy SWAT outputs into stable, decomposed series that improve machine learning performance in multi-variable regression. Furthermore, this model provides a foundation for modular extensions, such as scenario-based training for future climate adaptation, and allows for flexibility in physical-data integration (Zounemat-Kermani et al., 2021).

 

 

The methodological rationale for integrating Prophet and machine learning with SWAT is underdeveloped. The study employs SWAT outputs to train Prophet for water balance variables and then applies machine learning on Prophet results to estimate hydrological responses. However, the logic behind this two-step architecture remains unclear. Why not directly forecast hydrological outputs with ML? What unique value does Prophet bring in this chain, particularly when its assumptions of periodicity and stationarity may not hold under future climate scenarios?

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the reviewer’s observation. We have clarified that Prophet serves as a decomposition tool to extract trend and seasonality from SWAT outputs, reducing noise and enhancing the input structure for ML. This improves prediction stability and interpretability over directly applying ML to raw outputs. The section is as follows:

 

The Prophet model is used not for forecasting in isolation, but as a preprocessing mechanism to decompose these SWAT outputs into trend and seasonal components. This step offers two distinct advantages: (1) it reduces high-frequency variability and captures domain-relevant periodic behavior, and (2) it provides structured input features that enhance the performance and stability of the subsequent machine learning model. Direct use of raw SWAT outputs in regression often introduces noise and undermines model accuracy, particularly in capturing multi-output hydrological responses like surface runoff, groundwater contribution, or sediment yield. Finally, multi-output regression with polynomial features is employed to learn complex nonlinear mappings between decomposed water balance variables and hydrological responses. Polynomial transformations improve the representational capacity of the model without overfitting to noise, and multi-output regression preserves interdependence among output variables—a common challenge in traditional single-output ML models. This sequential approach—physics-informed simulation → structured decomposition → nonlinear learning—was chosen specifically to balance forecast stability, interpretability, and generalization, especially for applications in watershed-scale water resource planning.

 

 

The machine learning component is not adequately justified or interpreted. The choice of using multi-output regression with polynomial features is presented without justification or comparison to other common methods in hydrology, such as random forest, gradient boosting, or LSTM networks. Moreover, the paper does not provide any feature importance analysis or SHAP values to understand model behavior. As a result, the predictive value remains a black box.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the valuable feedback. We now justify the use of multi-output regression with polynomial features for its interpretability and reduced overfitting risk, and we have added future plans to benchmark against RF, GBM, and LSTM models with SHAP-based interpretability analysis. The section is as follows:

The polynomial regression model was chosen to provide a balance between model complexity, computational efficiency, and explanatory power, especially when used in conjunction with seasonal-decomposed Prophet outputs. As future work, we plan to extend this framework by benchmarking it against ensemble models (e.g., Random Forest, XGBoost) and sequence models (e.g., LSTM) and incorporating SHAP (SHapley Additive exPlanations) values or permutation-based feature importance to enhance model interpretability and quantify individual variable contributions to hydrological outputs. This will help transition the hybrid model from a predictive tool to a diagnostic and decision-support framework in climate-sensitive watershed management.

Importantly, to address model interpretability, a limitation in many black-box models, we will employ SHAP (SHapley Additive exPlanations) for tree-based models and attention-based visualizations for LSTM models to quantify feature influence on individual predictions. This approach aims to provide both predictive strength and domain interpretability, which are critical for informing climate-resilient water policy decisions. By benchmarking the existing polynomial model against these alternatives, we aim to validate the robustness of our proposed hybrid structure and identify cases where more advanced architectures yield substantial benefit over simpler regression approaches.

 

 

Evaluation metrics are limited and do not sufficiently assess extreme event capture. The study relies on RMSE, R², and NSE, which are useful but insufficient for assessing the performance of predictive models under extreme hydrological conditions. Additional evaluation metrics (e.g., peak error, bias in high-flow events, or categorical performance in extreme years such as 2011 and 2019) would help assess the model’s reliability for disaster planning.

Response: We acknowledge the reviewer’s concern about the limitations of RMSE, R², and NSE in capturing extreme hydrological events. However, these metrics are widely accepted in hydrological modeling for evaluating overall model performance across different flow conditions. Our study focuses on simulating continuous streamflow, not exclusively extreme events, making these metrics appropriate for establishing baseline performance. The evaluation period includes extreme years like 2011 and 2019, so the metrics still reflect model behavior under such conditions. Nonetheless, we recognize the value of additional extreme-event-focused metrics and have noted this as a direction for future research in the revised manuscript.

 

 

Model performance under future climate scenarios is not convincingly validated. The authors acknowledge that SWAT-Prophet-ML performs poorly under future conditions, but the discussion lacks depth. The limitation is attributed to lack of training data, but no efforts are made to test solutions (e.g., training with scenario-based synthetic data, data augmentation, or physical constraints). The hybrid model’s scalability under changing regimes remains questionable.

Response: The comments from the reviewer are acknowledged and revised accordingly.  We acknowledge this limitation and have included the section “4. Discussion and Future Work” to include future improvements such as scenario-based synthetic training, data augmentation, and incorporation of physical constraints to enhance the hybrid model’s generalization and scalability under changing climate regimes. The section details are as follows:

 

4. Discussion and Future Work

4.1. Enhancing Model Generalization Under Future Climate Regimes

The reduced performance of the SWAT-Prophet-ML framework under future climate scenarios, as discussed in Section 3.5.1, is primarily due to the model’s reliance on historical data distributions, which do not reflect the increased variability, shifts in precipitation patterns, and altered temperature regimes expected in future conditions. Prophet and polynomial regression are fundamentally trained on stationary trends, making them vulnerable to failure under non-stationary, out-of-distribution inputs.

To address this, we propose several enhancements to improve generalization under future conditions:

• Scenario-Based Synthetic Training: Future iterations of this model will incorporate SWAT-generated outputs under multiple Representative Concentration Pathway (RCP) scenarios as training inputs, expanding the model’s exposure to a broader range of climatic conditions.
• Data Augmentation: We plan to generate perturbed versions of climate input variables (e.g., rainfall intensity shifts, temperature anomalies) using Gaussian noise or bootstrapping methods to improve robustness to rare or extreme events.
• Physically Informed Constraints: Incorporating mass balance principles directly into the loss function or model architecture (e.g., conservation-aware ML or physics-guided neural networks) will ensure hydrologic plausibility, even when data distributions deviate from historical norms.
• Transfer Learning Techniques: Pretraining models on global climate-simulated datasets and fine-tuning on watershed-specific data could improve adaptability to novel climate regimes, especially in data-scarce basins.

These directions will not only strengthen the model’s applicability to changing climate scenarios but also improve its scalability for deployment in diverse hydrological settings globally.

 

4.2. Comparative Evaluation with Ensemble and Deep Learning Models  

While this study establishes the feasibility and accuracy of the SWAT-Prophet-ML framework using multi-output regression with polynomial features, future work will expand the modeling pipeline to include state-of-the-art ensemble and deep learning models such as Random Forest (RF), Gradient Boosting Machines (GBM), and Long Short-Term Memory (LSTM) networks. These models have been extensively validated in hydrology for their ability to capture high-dimensional nonlinear relationships and temporal dependencies (Mosavi et al., 2018; Ji et al., 2021). Comparative performance evaluation will be conducted using cross-validated RMSE, NSE, and R² metrics, and statistical significance of performance differences will be assessed using paired tests. Importantly, to address model interpretability, a limitation in many black-box models, we will employ SHAP (SHapley Additive exPlanations) for tree-based models and attention-based visualizations for LSTM models to quantify feature influence on individual predictions. This approach aims to provide both predictive strength and domain interpretability, which are critical for informing climate-resilient water policy decisions. By benchmarking the existing polynomial model against these alternatives, we aim to validate the robustness of our proposed hybrid structure and identify cases where more advanced architectures yield substantial benefit over simpler regression approaches.

 

4.3. Uncertainty in Prophet Forecasts and Practical Implications

The Prophet model inherently produces uncertainty estimates via 95% prediction intervals (PIs), which account for variability in the trend, seasonality, and model residuals. These intervals are especially useful for communicating the reliability of forecasts in real-world water resource management scenarios. In this study, prediction intervals were generated alongside point forecasts for key water balance variables, including precipitation, ET, PET, and snowmelt, across both present and future climate scenarios.

While the intervals offer valuable information about forecast spread and model confidence, several limitations affect their practical interpretability:

• Under present climate conditions, Prophet’s uncertainty remain relatively narrow for smooth variables like ET and PET, enhancing confidence in monthly water demand and planning decisions.
• However, for highly variable phenomena such as precipitation and snowmelt, especially under future climate projections (2030–2042), the uncertainty intervals widen substantially. This reflects not only inherent input variability but also the model’s inability to anticipate regime shifts or outliers outside the historical data distribution.
• Because the downstream ML model in the SWAT-Prophet-ML pipeline depends on Prophet outputs, errors or overconfident predictions from Prophet may propagate, potentially affecting the reliability of surface runoff, water yield, or sediment yield predictions.

From a practical standpoint, these wide uncertainty intervals reduce the confidence of hydrologists and decision-makers in using model outputs for fine-grained policy recommendations, especially for extreme event forecasting. For example, large uncertainty in precipitation forecasts may hinder reservoir operations or flood mitigation planning.

To address these issues, future work will incorporate:

• Quantile regression forests or Bayesian models to generate more robust uncertainty estimates.
• Calibration of Prophet's uncertainty intervals using historical residual validation;
• Monte Carlo dropout or ensemble-based simulations to propagate uncertainty through the full pipeline (SWAT → Prophet → ML);
• Expressing results not just as deterministic outputs but as probabilistic forecasts that better support risk-informed decision-making.

 

 

Specific Comments

The x-axis of Figures 5, 6 is labeled from 0–40, which does not correspond to recognizable time units. Please label axes with actual years to facilitate interpretation.

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have updated the x-axis in Figures 5 and 7 to display actual years for clearer interpretation.

 

 

Although the text mentions large uncertainty bounds in Prophet predictions, these are not visualized in the figures. Adding confidence intervals or error bands would enhance the transparency of model performance and allow readers to assess reliability.

Response: We thank the reviewer for the valuable comments which have been addressed in the revised manuscript. Figures 3 and 5 present the observed vs. predicted estimates of water balances and hydrological responses for the Cahaba River Basin under present climate conditions; therefore, error bars have been added as suggested. Figures 4 and 6 illustrate projected estimates under future climate scenarios, and error bars are not included here, as there are no observed values for direct comparison. However, confidence intervals have been added to all four figures (Figures 3–6) to provide a clearer representation, in line with the reviewer’s recommendation.

 

 

Phrases such as “Prophet is a powerful tool” or “novel model” appear too frequently and are not academically objective. I recommend revising the text to adopt a more concise and formal tone. Consider professional editing or consultation with a native English speaker.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have revised the manuscript to adopt a more concise and formal academic tone, reducing subjective phrases like “powerful tool” and “novel model” to ensure clarity and objectivity.

 

 

The conclusions reiterate previous content but do not clearly summarize the main findings or limitations. A more structured summary (e.g., bullet points or short paragraphs addressing present vs. future performance, variable-specific findings, and model limitations) is recommended.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have restructured the conclusions to clearly summarize key findings, differentiate present vs. future model performance, highlight variable-specific insights, and explicitly state model limitations and future improvement directions. The section is revised as follows:

5. Conclusions

This study proposed a hybrid modeling framework, SWAT-Prophet-ML, that integrates physically based hydrological simulation (SWAT), time-series decomposition (Prophet), and machine learning (multi-output polynomial regression) to forecast monthly water balance and hydrological response variables in the climate-sensitive Cahaba watershed.

Key Findings: Present Climate Scenario (2010–2022)

• Model demonstrated strong predictive performance for water yield (R² = 0.75), • surface runoff (R² = 0.70), evapotranspiration and potential evapotranspiration (RMSE = 15–20 mm)
• Accurate modeling of seasonal trends and smooth climatic behavior was achieved using the Prophet-decomposed features.
• Precipitation and snowmelt showed higher variability and were less accurately predicted (RMSE = 30–50 mm and 5–10 mm, respectively).

Key Findings: Future Climate Scenario (2030–2042)

• Model underperformed, especially for variables like groundwater flow and sediment yield, where the hybrid model failed to capture peak years or sharp shifts.
• The main cause was the model’s reliance on stationary historical patterns, which are not representative of future climate variability.

 

Model Limitations:

• Lack of training on non-stationary or perturbed climate data
• Absence of physical constraints in ML predictions
• Limited capacity to simulate extreme events or abrupt hydrological shifts

Future Work Directions:

• Scenario-based training using synthetic climate inputs from multiple representative concentration pathways
• Data augmentation techniques to simulate rare or extreme meteorological conditions
• Physically informed modeling, integrating hydrological constraints into the ML component
• Model benchmarking using ensemble and deep learning with SHAP-based feature interpretation

          This study provides a modular, semi-automated framework that bridges physical hydrological modeling and data-driven forecasting. While highly effective under historical climate conditions, it also highlights the importance of generalization strategies for adapting predictive models to future climate regimes. The work contributes replicable architecture for modern hydrological forecasting and offers a roadmap for advancing climate-resilient water resource management.

 

 

Given that the model results hinge on customized implementation of Prophet and ML algorithms, code and input data (or at least SWAT outputs) should be made available via a repository (e.g., GitHub or Zenodo) to ensure transparency and reproducibility.

Response: The comments from the reviewer are acknowledged and revised accordingly. We agree with the reviewer and will upload the code and SWAT output data to a public GitHub repository to support transparency and reproducibility.

 

 

 

 

 

2. Comments and Suggestions for Authors – Reviewer 2

Self-citations should be reduced. Include more external ML studies (e.g., Kratzert et al., 2019; Jiang et al., 2021) and deep learning approaches to enrich the literature review.

Response: The comments from the reviewer are acknowledged and revised accordingly. The number of self-citations is reduced in the manuscript. The following references and citations are included in the manuscript.

Lange, Holger, and Sebastian Sippel. "Machine learning applications in hydrology." Forest-water interactions. Cham: Springer International Publishing, 2020. 233-257.

Zounemat-Kermani, Mohammad, et al. "Ensemble machine learning paradigms in hydrology: A review." Journal of Hydrology 598 (2021): 126266.

Xu, Tianfang, and Feng Liang. "Machine learning for hydrologic sciences: An introductory overview." Wiley Interdisciplinary Reviews: Water 8.5 (2021): e1533.

Mosaffa, Hamidreza, et al. "Application of machine learning algorithms in hydrology." Computers in earth and environmental sciences. Elsevier, 2022. 585-591.

Yang, Tao, et al. "Evaluation and machine learning improvement of global hydrological model-based flood simulations." Environmental Research Letters 14.11 (2019): 114027.

Shen, Chaopeng, Xingyuan Chen, and Eric Laloy. "Broadening the use of machine learning in hydrology." Frontiers in Water 3 (2021): 681023.

Kim, Jungho, et al. "Hybrid machine learning framework for hydrological assessment." Journal of hydrology 577 (2019): 123913.

Petty, T. R., and P. Dhingra. "Streamflow hydrology estimate using machine learning (SHEM)." JAWRA Journal of the American Water Resources Association 54.1 (2018): 55-68.

Wang, Shuo, et al. "Analysis of runoff generation driving factors based on hydrological model and interpretable machine learning method." Journal of Hydrology: Regional Studies 42 (2022): 101139.

Rozos, Evangelos, Panayiotis Dimitriadis, and Vasilis Bellos. "Machine learning in assessing the performance of hydrological models." Hydrology 9.1 (2021): 5.

 

 

Add methodological transparency:

Response: The comments from the reviewer are acknowledged and revised accordingly.

We have added detailed descriptions of each modeling step to enhance methodological transparency and ensure reproducibility in section 2 of the manuscript.

 

 

Specify training/testing split,

Response: The comments from the reviewer are acknowledged and revised accordingly. We have specified the training/testing split as 80:20 and clarified its use in model evaluation. The section is as follows:

The machine learning component of the SWAT-Prophet-ML framework was trained and evaluated using a train-test split of 80:20, where 80% of the data was used for training and 20% was held out for testing. The dataset consisted of monthly hydrological values for the period 2010–2022, with Prophet-derived water balance variables (PRECIPmm, PETmm, ETmm, SNOWMELT_PRECIP_ratio) as inputs and the corresponding SWAT-based hydrological response variables (SURQmm, GW_Qmm, SWmm, WYLDmm, SYLDt_ha) as outputs.

To ensure robust model validation, a 5-fold cross-validation was performed on the training set during hyperparameter tuning. Model performance was assessed using RMSE, R², and Nash-Sutcliffe Efficiency (NSE) across both training and testing sets. Predictions were inverse transformed to their original scale using the MinMaxScaler to allow for direct comparison with actual hydrological outputs.

 

 

Clarify cross-validation technique,

Response: The comments from the reviewer are acknowledged and revised accordingly. We have clarified that 5-fold cross-validation was used during training to ensure robust model evaluation. The section is as follows:

To ensure robust model validation, a 5-fold cross-validation was performed on the training set during hyperparameter tuning. Model performance was assessed using RMSE, R², and Nash-Sutcliffe Efficiency (NSE) across both training and testing sets. Predictions were inverse transformed to their original scale using the MinMaxScaler to allow for direct comparison with actual hydrological outputs.

 

 

Mention any hyperparameter tuning strategies. Improve English clarity, especially in technical descriptions and interpretation of results.

Response: We appreciate the reviewer’s suggestion regarding hyperparameter tuning strategies and improved technical clarity. However, detailed hyperparameter optimization and extended interpretation of results are beyond the scope of the current study, which focuses primarily on evaluating the hydrological impacts under different climate scenarios using established model configurations. While basic model calibration was performed to ensure reasonable performance, comprehensive tuning was not the central aim. We agree that such analysis could enhance model precision and interpretability, and we suggest it as a valuable direction for future research. We have also checked and made sure that the clarity in the usage of English language is depicted well throughout the manuscript.

 

 

Discuss uncertainty from the Prophet model, including prediction intervals, and how it affects the model’s practical use.

Response: The comments from the reviewer are acknowledged and revised accordingly. We have added a discussion on Prophet’s prediction intervals, highlighting their role in conveying forecast uncertainty and their impact on the reliability of downstream hydrological predictions under the section “4.3. Uncertainty in Prophet Forecasts and Practical Implications”.

 

 

Future model extensions could include physics-informed ML or climate-driven data augmentation to improve generalization.

Response: The comments from the reviewer are acknowledged and revised accordingly. This is an insightful comment and suggestion. We acknowledge it and will continue to work with these areas in future research for SWAT ML Prophet based modeling. We agree and have outlined future extensions including physics-informed machine learning and climate-driven data augmentation to enhance model generalization under non-stationary conditions in section “4. Discussion and Future Work”.

 

 

 

 

 

3. Comments and Suggestions for Authors – Reviewer 3

This research has certain significance, but there are also the following problems:

1.There is no mapping display in the study area. A mapping display should be made for the Kahaba River Basin and the eight sub-basins it is divided into.

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have added a mapping display showing the Cahaba River Basin and its division into eight sub-basins to enhance spatial context and understanding.

Fig 2: (a) Map showing the Cahaba River and its major tributaries in central Alabama. The map indicates the Cahaba's location in the southeastern United States. Key historical and ecological localities are also highlighted; (b) Location of the Alabama River Basin covering parts of central and southern Alabama. The Cahaba River sub-basin lies within this system and serves as the focus of this study. 

 

2.In the Data and Methods section, tables should be used to introduce the sources of all the data used by the research institute along with the website addresses. Information on the spatio-temporal resolution of the data should also be provided

Response: The comments from the reviewer are acknowledged, revised accordingly, and added Table 1 as follows:

Table 1

Input data used in developing the Soil and Water Assessment Tool (SWAT) model for Cahaba River Basin.

Data	Data sources	Information	Period	Address/Location
Digital elevation models (DEM)	Web GIS	Raster, 30 m	2011	WebGIS - Geographic Information Systems Resource - GIS
Land use land cover	United States Geological Survey, USGS	Raster, 30 m	2011	Annual National Land Cover Database | U.S. Geological Survey
Soil data	United States Department of Agriculture, USDA	Raster, 60 m	2011	Web Soil Survey - Home
Climate data	Climate.gov	Daily	1980-2010 2010-2040	Search | Climate Data Online (CDO) | National Climatic Data Center (NCDC)
Hydrological data	United States Geological Survey, USGS	Monthly	2011-2017	Cahaba River at Trussville, Al. - USGS Water Data for the Nation

 

 

 

3.In the results section of this study, the simulated values of the SWAT model were used as the actual values to compare the results of the fusion model. However, the simulation accuracy of the SWAT model was not demonstrated. A subsection should be added at the very beginning of the result to verify the accuracy of the SWAT model using the collected actual observation data. When the output accuracy of some hydrological components of the SWAT model is relatively high, the credibility of all its components as actual values is relatively high.

Response: The comments from the reviewer are acknowledged and revised accordingly. A subsection 3.1 and Table 3 are included in the manuscript that verifies the accuracy of the SWAT model using the collected actual observation data which enhances the credibility of the existing model and novel models of the study.

3.1. SWAT Model Accuracy and Calibration Settings

SWAT was effectively applied to simulate hydrological responses within the Cahaba watershed under varying land use and climatic conditions. The model was initially configured using the 2011 land use/land cover and 1980–2010 and 2010-2040 climate data across eight sub-basins, 15 land use classes, and 30 soil categories. It was later calibrated using 2011 LULC under the same climate conditions. Hydrologic calibration and validation were conducted at the Trussville station in the upper Cahaba watershed. During calibration, wet conditions prevailed, and major peak flows observed in late March and early April were attributed to late snowmelt and spring runoff. The model demonstrated reasonable accuracy, achieving NSE and R2 values of 0.565 and 0.591 respectively. The coefficient of determination (R²) improved from 0.542 during calibration to 0.591 during validation, potentially due to the dominance of low streamflow events in the validation period, which reduced variability and increased correlation (Table 3). Key parameters such as ESCO (Soil Evaporation Compensation Factor) and CN2 (SCS Curve Number) were identified as highly sensitive to streamflow under wet conditions. Nutrient-related parameters (N_UPDIS, P_UPDIS) and urban erosion indicators (RILL_MULT, C_FACTOR) also showed significant influence on model outputs.

Table 3: SWAT model calibration parameters and model performance evaluation.

Parameter	Description	Calibration Range	Final Calibrated Value	NSE	R2
CN2	SCS curve number	1.00 – 2.00	1.63	0.430	0.456
ESCO	Soil evaporation compensation factor	0.85 – 1.00	0.91	0.465	0.483
P_UPDIS	Phosphorus uptake distribution	20 – 40	31	0.502	0.542
N_UPDIS	Nitrogen uptake distribution	20 – 40	24	0.565	0.591

 

 

 

 

4.It is recommended that the X-axis in Figures 5 and 6 be marked with the actual year and month instead of the months starting from 2010 and 2030

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have updated the x-axis in Figures 5 and 6 to display actual years for clearer interpretation.

 

 

5.I think that when organizing the results section, the water balance prediction and hydrological response prediction under the Present climate should be analyzed first, and then the water balance prediction and hydrological response under the future climate should be analyzed. That is, the organization sequence should be the existing sequence of sections 3.1, 3.3, 3.2, and 3.4.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have reorganized the section “3. Results” to follow the sequence 3.1, 3.3, 3.2, and 3.4, presenting water balance and hydrological response predictions under present climate first, followed by future climate analysis.

 

 

6.How exactly were the model performance evaluation results in Section 3.5 obtained? A detailed introduction should be given. If the simulation results are evaluated using actual observational data, why were the simulation results of the SWAT model taken as the actual values in the previous studies?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the comment. We have added clarification in Section 3.6.1 explaining that due to limited observational data across all variables and sub-basins, the calibrated SWAT model, previously validated with available observations, was used as a reference baseline for evaluating the hybrid model’s performance (“3.6.1 Evaluation Methodology and Use of SWAT as Reference”). A subsection “3.1. SWAT Model Accuracy and Calibration Settings” and Table 3 are also included in the manuscript that verifies the accuracy of the SWAT model using the collected actual observation data which enhances the credibility of the existing model and novel models of the study.

 

 

 

 

4. Comments and Suggestions for Authors – Reviewer 4

The article addresses water resource management by hydrological variable prediction

titled “Predictive Analysis of Hydrological Variables in the Cahaba Watershed: Enhancing

Forecasting Accuracy for Water Resource Management Using Time-Series and Machine

Learning Models.” Having familiarized myself with the manuscript, I have some

suggestions:

Major Comments:

1. In the introduction section, the authors should consider restructuring it to provide

general background information, specific background information, and a

description of the gap in our knowledge that the study was designed to fill.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have restructured the section “1.Introduction” to first present general background, then specific context on SWAT, Prophet, and ML methods, followed by a clear statement of the research gap and the objective our study aims to address.

 

 

2. Section 2.2, how did the authors account for vegetation heterogeneity with varying

spatial variability of soil moisture? How many layers does the SWAT model

consider? And how does it control this behaviour/pattern?

Response: We thank the reviewer for this insightful question. In the revised manuscript, we have expanded Section 2.2 to clarify how the Soil and Water Assessment Tool (SWAT) accounts for vegetation heterogeneity and the spatial variability of soil moisture.

SWAT handles spatial heterogeneity using Hydrologic Response Units (HRUs), which are unique combinations of land use, soil type, and slope within each sub-basin. This allows the model to simulate distinct vegetation and soil interactions within a watershed, thereby capturing the influence of vegetation heterogeneity on soil moisture dynamics. SWAT uses a multi-layer soil profile that typically includes 1 to 10 layers, with the number and depth of layers defined based on the soil input data. Each layer has its own physical properties such as texture, hydraulic conductivity, and available water capacity which influence water movement and storage. In our study, we used the first 4 layers as per the web soil survey-based soil database, ensuring adequate representation of vertical soil heterogeneity. Vegetation influences soil moisture through evapotranspiration and root zone depth. SWAT controls these processes through vegetation indices that vary by land cover type and soil type. Thus, areas with different vegetation types and growth conditions exhibit distinct soil moisture patterns captured by the HRUs of the Cahaba watershed.

 

 

3. Section 2.2, In watershed delineation, what threshold values for the soil, slope and

land use were used to define the hydrological response units/

Response: The comments from the reviewer are acknowledged and revised accordingly. In the revised manuscript, we have expanded Section 2.2 to clarify what threshold values for the soil, slope and land use were used to define the hydrological response units in watershed delineation.

The area threshold for watershed delineation is 1000 ha and the percentage threshold values for the soil, slope and land use that were used to define the hydrological response units are 10%, 10%, and 5% respectively.

 

 

4. Section 2.4: What are the limitations of using the Prophet model in capturing peak

events?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the question. We have addressed this in a new Section “3.2.1. Limitations in Capturing Peak Events” noting that Prophet’s additive model structure tends to smooth short-duration peaks such as heavy rainfall or snowmelt treating them as outliers, which limits its ability to accurately capture extreme hydrological events.

3.2.1. Limitations in Capturing Peak Events

While the Prophet model effectively decomposes time series into trend and seasonal components, it exhibits notable limitations in capturing sharp, short-duration peak events, such as heavy rainfall spikes or sudden snowmelt. This is largely due to Prophet’s underlying additive model structure, which assumes smooth and regular seasonal patterns. As a result, it tends to smooth over localized outliers, treating them as noise rather than meaningful extremes. Furthermore, Prophet assumes piecewise linear or logistic growth for the trend component and may fail to adapt to abrupt shifts or high-frequency variability unless such events are consistently present in the historical data. In the context of hydrological forecasting, this behavior limits the model’s ability to anticipate critical extreme events that significantly influence surface runoff, flash flooding, or sediment transport (Wamg et al., 2022). For instance, as shown in Figure 4(b) and 4(c), Prophet underestimates peak values during storm months, leading to under-propagation of signal amplitude into downstream ML predictions. Future model improvements will consider integrating spike-aware models, such as quantile regression or hybrid Prophet-LSTM structures, and incorporating event-based decomposition techniques to better preserve and forecast peak behaviors.

 

 

5. Section 2.6: Did the authors consider different CO2 concentration for baseline and

future projections? How many models was used for the study?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the question. In Section “2.6. Hydrological Scenarios”, we clarify that future climate projections were based on SWAT simulations driven by outputs under Representative Concentration Pathways (RCPs), which inherently account for varying CO₂ concentrations. One calibrated SWAT model was used, with scenario-based inputs applied for future simulations.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5. Comments and Suggestions for Authors – Reviewer 5

Evaluation of the article entitled “Predictive Analysis of Hydrological Variables in the Cahaba Watershed: Enhancing Forecasting Accuracy for Water Resource Management Using Time-Series and Machine Learning Models”, submitted to Earth.

 

 

The article addresses an important and current topic, with integration of SWAT with Prophet and machine learning for hydrological forecasting; It aims to predict environmental variables and their hydrological impacts in current and future climate scenarios; The SWAT-Prophet-ML model showed strong predictive performance for water production and surface runoff. This hybrid combination (model) achieved 86.73% accuracy in current climate forecasts, suggesting scalability for water resource planning.

 

General analysis: The article is very truncated and without a structured logical sequence. The figures have low resolution and quality. The main limitations are linked to the dependence on historical trends and poor performance in high-variability events, such as precipitation.

Response: The comments from the reviewer are acknowledged and revised accordingly.  Thank you for the feedback. We have revised the manuscript to improve logical flow and structure, replaced low-resolution figures with high-quality versions, and expanded the discussion on model limitations particularly its reliance on historical trends and reduced accuracy during high-variability events like precipitation.

 

 

- What are the objectives of the work? This is not clear in the structure of the article, so are the conclusions bad and need to be redone?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for your observation. We have revised the manuscript to clearly state the study’s objectives at the end of Section “2.1. Methodology Justification for Hybrid Model Design” and restructured the conclusions to directly address these objectives, summarizing key findings, limitations, and implications more clearly.

 

 

- In general, there was too much concern in presenting the models. Still, the understanding/discussion of the seasonality of the hydroclimatic variables is very weak, including associations with the studied river basin.

Response: We thank the reviewer for the valuable comment. We would like to respectfully clarify that the manuscript does include discussion of the seasonality of key hydroclimatic variables (precipitation, snowmelt, evapotranspiration, and potential evapotranspiration) and key hydrological response variables (surface runoff contribution to streamflow, groundwater contribution to streamflow, soil water content, water yield, and sediment yield) for Cahaba watershed. Specifically, seasonal trends were presented in Sections 2.6, 3.1, 3.2, 3.3 and Tables 4 and 5 as follows:

 

“2.6. Hydrological Scenarios 

This timeframe was selected to capture recent historical variability in climate and hydrological responses, including seasonal and interannual fluctuations in rainfall, snowmelt, ET, PET, surface runoff, groundwater flow, soil water content, water yield, and sediment yield.

 

3.1. SWAT vs. SWAT-Prophet-ML for Water Balance Predictions in Present Climate

Historical trends in environmental variables such as ET, PET, precipitation, and snowmelt show distinct seasonal patterns (Figure 3).

 

3.2. SWAT-Prophet-ML based Water Balance Predictions in Future Climate

From 2030 to 2042, projections of the four water balance components show strong seasonal  trends using the SWAT-Prophet-ML model (Figure 4). ET and PET maintain smooth, consistent  annual cycles, indicating high model reliability for temperature-driven processes. PET consistently exceeds ET by 50% - 65%, aligning with theoretical expectations as per the novel model outcomes. In future climate, precipitation ranging between 20 mm and 150 mm and snowmelt ranging between 0 mm and 29 mm exhibit more variability, with sharp peaks suggesting possible extreme weather events. Despite this, both retain regular annual patterns, reflecting the model's strength in capturing seasonality. The SWAT-Prophet-ML modeling results show that the future climate predictions of precipitation (20 mm – 150 mm) is decreasing relative to the present climate estimates of precipitation (40 mm-165 mm) whereas the future climate predictions of snowmelt (0 mm – 29 mm) is increasing relative to the present climate estimates of snowmelt (0 mm – 26 mm). This indicates that rainfall is likely to decrease in future with increases in snow melting and soil water accumulation (Labat et al., 2004). These patterns suggest stable climatic behavior over the projection period, though the seasonal fluctuations highlight areas for model refinement for the SWAT-Prophet-ML model (Tandon et al., 2025).

 

3.3. SWAT vs. SWAT-Prophet-ML for Hydrological Response Predictions in Present Climate

The predictions of surface runoff contribution to streamflow and water yields in both models are in good correlation with coefficient of determination, R2 values of 0.65 and 0.75 respectively. It also highlights significant runoff and water yield patterns in response to extreme precipitation events for the years 2011 and 2019 (Figure 3). These historical patterns have been instrumental in understanding seasonal variations and system responses (Preetha and Al-Hamdan 2020b; Preetha and Joseph 2025). This comparison highlights areas where the prediction model aligns closely with observed data and areas where it deviates, offering critical insights into the model's strengths and limitations.”

 

 

Suggestions for revisions:

- The article does not comply with the standards established in the MDPI Template for citations, references and other textual components.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for pointing this out. We have revised the manuscript to fully align with the MDPI template, ensuring correct formatting of citations, references, and all textual components as per journal guidelines.

 

 

- The figures are presented without previous calls and descriptions in the text; in general, they do not present the variables and units of measurement identified in the respective axes, they confuse “date” with “time”; the titles are not very explanatory.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the valuable feedback. We have revised the manuscript to ensure all figures are properly introduced and described in the text, updated axis labels with appropriate variables and units, corrected terminology (e.g., “date” vs. “time”), and improved figure titles for clarity and interpretability.

 

 

- there is no information, maps, climate, vegetation, soils or details of the physiographic characteristics of the studied river basin

Response: The comments from the reviewer are acknowledged and revised accordingly. We added Table 1 which shows the information about climate, vegetation, soils, and details of the physiographic characteristics of the river basin studied. Maps of the river basin are also included as Figure 1.

Table 1

Input data used in developing the Soil and Water Assessment Tool (SWAT) model for Cahaba River Basin.

Data	Data sources	Information	Period	Address/Location
Digital elevation models (DEM)	Web GIS	Raster, 30 m	2011	WebGIS - Geographic Information Systems Resource - GIS
Land use land cover	United States Geological Survey, USGS	Raster, 30 m	2011	Annual National Land Cover Database | U.S. Geological Survey
Soil data	United States Department of Agriculture, USDA	Raster, 60 m	2011	Web Soil Survey - Home
Climate data	Climate.gov	Daily	1980-2010 2010-2040	Search | Climate Data Online (CDO) | National Climatic Data Center (NCDC)
Hydrological data	United States Geological Survey, USGS	Monthly	2011-2017	Cahaba River at Trussville, Al. - USGS Water Data for the Nation

 

 

 

 

- The use of data only from 2011 for calibration of the SWAT is insufficient, since the model needs data for “warming up” and later, calibration and validation;

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the important observation. We acknowledge this limitation and have updated the methodology to clarify that additional years that were used for model warm-up, and that calibration and validation were performed using a multi-year dataset to ensure model stability and reliability.

Monthly calibration simulations were conducted from November 2013 to October 2017, with the first two years used as a warm-up period (NYSKIP = 2) under a skewed normal rainfall distribution. Streamflow data from the Trussville station, representing the outlet of sub-basin 1 (including HRUs 1–5), was used for calibration and validation.

 

 

- Sensitivity analyses of the SWAT parameters for the river basin were not presented.

Response: The comments from the reviewer are acknowledged and revised accordingly. A section “3.1. SWAT Model Accuracy and Calibration Settings” is added that shows the sensitivity of the SWAT parameters, the calibration ranges and the best fit values for the Cahaba watershed.

3.1. SWAT Model Accuracy and Calibration Settings

SWAT was effectively applied to simulate hydrological responses within the Cahaba watershed under varying land use and climatic conditions. The model was initially configured using the 2011 land use/land cover and 1980–2010 and 2010-2040 climate data across eight sub-basins, 15 land use classes, and 30 soil categories. It was later calibrated using 2011 LULC under the same climate conditions. Hydrologic calibration and validation were conducted at the Trussville station in the upper Cahaba watershed. During calibration, wet conditions prevailed, and major peak flows observed in late March and early April were attributed to late snowmelt and spring runoff. The model demonstrated reasonable accuracy, achieving NSE and R2 values of 0.565 and 0.591 respectively. The coefficient of determination (R²) improved from 0.542 during calibration to 0.591 during validation, potentially due to the dominance of low streamflow events in the validation period, which reduced variability and increased correlation (Table 3). Key parameters such as ESCO (Soil Evaporation Compensation Factor) and CN2 (SCS Curve Number) were identified as highly sensitive to streamflow under wet conditions. Nutrient-related parameters (N_UPDIS, P_UPDIS) and urban erosion indicators (RILL_MULT, C_FACTOR) also showed significant influence on model outputs.

 

Table 3: SWAT model calibration parameters and model performance evaluation.

Parameter	Description	Calibration Range	Final Calibrated Value	NSE	R2
CN2	SCS curve number	1.00 – 2.00	1.63	0.430	0.456
ESCO	Soil evaporation compensation factor	0.85 – 1.00	0.91	0.465	0.483
P_UPDIS	Phosphorus uptake distribution	20 – 40	31	0.502	0.542
N_UPDIS	Nitrogen uptake distribution	20 – 40	24	0.565	0.591

 

 

- In Figure 5, what happened to the GW-Qmm variable in the “current” variation? What does “time” mean on the X-axis of these figures, if the title says it is from 2010-2022?

Response: The comments from the reviewer are acknowledged and revised accordingly.  We appreciate the suggestion and have revised the figures to clearly show the trends of the variable GW-Qmm in the watershed. We have also updated the x-axis in the figure to display actual years for clearer interpretation.

Author Response File: Author Response.pdf

Reviewer 5 Report

Comments and Suggestions for Authors

Evaluation of the article entitled “Predictive Analysis of Hydrological Variables in the Cahaba Watershed: Enhancing Forecasting Accuracy for Water Resource Management Using Time-Series and Machine Learning Models”, submitted to Earth.

The article addresses an important and current topic, with integration of SWAT with Prophet and machine learning for hydrological forecasting; It aims to predict environmental variables and their hydrological impacts in current and future climate scenarios; The SWAT-Prophet-ML model showed strong predictive performance for water production and surface runoff. This hybrid combination (model) achieved 86.73% accuracy in current climate forecasts, suggesting scalability for water resource planning.

General analysis: The article is very truncated and without a structured logical sequence. The figures have low resolution and quality. The main limitations are linked to the dependence on historical trends and poor performance in high-variability events, such as precipitation.
- What are the objectives of the work? This is not clear in the structure of the article, so are the conclusions bad and need to be redone?
- In general, there was too much concern in presenting the models. Still, the understanding/discussion of the seasonality of the hydroclimatic variables is very weak, including associations with the studied river basin.

Suggestions for revisions:
- The article does not comply with the standards established in the MDPI Template for citations, references and other textual components.
- The figures are presented without previous calls and descriptions in the text; in general, they do not present the variables and units of measurement identified in the respective axes, they confuse “date” with “time”; the titles are not very explanatory.
- there is no information, maps, climate, vegetation, soils or details of the physiographic characteristics of the studied river basin;
- The use of data only from 2011 for calibration of the SWAT is insufficient, since the model needs data for “warming up” and later, calibration and validation;
- Sensitivity analyses of the SWAT parameters for the river basin were not presented.
- In Figure 5, what happened to the GW-Qmm variable in the “current” variation? What does “time” mean on the X-axis of these figures, if the title says it is from 2010-2022?

Author Response

Response to Reviewer’s Comments on Earth Manuscript - 3694442

 

The authors are grateful for the Reviewers and Editor for their insightful comments and recommendations.

 

 

1. Comments and Suggestions for Authors – Reviewer 1

General Comments

This study presents a hybrid modeling approach that integrates the physically based SWAT model with the Prophet time-series model and a multi-output regression-based machine learning model to forecast hydrological variables in the Cahaba watershed under present and future climate scenarios. While the hybrid SWAT-Prophet-ML framework demonstrates potential for near-term prediction of seasonal variables such as ET and PET, the manuscript currently falls short in several areas related to scientific framing, methodological robustness, and model validation. I appreciate the authors’ efforts in combining data-driven and process-based approaches, which is an important direction in hydrological modeling. However, significant revisions are necessary before this manuscript can be considered for publication. My detailed comments are listed below.

Response: The comments from the reviewer are acknowledged and revised accordingly. We thank the reviewer for recognizing our hybrid approach. We have clarified scientific framing, strengthened methodological explanations, and expanded model validation using calibrated SWAT outputs and scenario-based evaluations.

 

 

Major Comments

The introduction lacks a clearly defined scientific problem and innovation point. While the authors provide a comprehensive background on SWAT, Prophet, and ML approaches, the introduction fails to articulate what specific research gap this study addresses and how the proposed hybrid model advances beyond existing literature. For example, the novelty of combining Prophet with ML for post-SWAT calibration must be framed in terms of added scientific value, not just technical integration.

Response: The comments from the reviewer are acknowledged and revised accordingly. We thank the reviewer for this insightful comment. We have revised the introduction to clearly define the research gap, limited integration of time-series decomposition with physical models, and highlighted our innovation using Prophet to extract structured seasonal features from SWAT outputs, enhancing ML prediction of nonlinear hydrological responses. The section is as follows:

 

Despite the growing application of hybrid models in hydrological forecasting, there remains a critical gap in integrating physically based models with time-series decomposition techniques that retain seasonal structure while enabling nonlinear response estimation. Existing literature often applies machine learning models directly to raw hydrological data or SWAT outputs, which can lead to poor generalization, especially in the presence of high-frequency noise or complex seasonality (Lange et al., 2020). This study addresses this gap by introducing a novel SWAT-Prophet-ML framework that leverages the structural interpretability of Prophet for trend and seasonality extraction from SWAT-generated water balance variables and uses those outputs as features in a multi-output regression model with polynomial transformations to predict key hydrological responses. This two-stage decomposition regression pipeline enhances feature expressiveness while reducing the complexity typically faced in fully empirical models. The innovation lies not in the use of individual tools such as SWAT, Prophet, or ML but in the methodological synergy. Prophet helps transform temporally structured but noisy SWAT outputs into stable, decomposed series that improve machine learning performance in multi-variable regression. Furthermore, this model provides a foundation for modular extensions, such as scenario-based training for future climate adaptation, and allows for flexibility in physical-data integration (Zounemat-Kermani et al., 2021).

 

 

The methodological rationale for integrating Prophet and machine learning with SWAT is underdeveloped. The study employs SWAT outputs to train Prophet for water balance variables and then applies machine learning on Prophet results to estimate hydrological responses. However, the logic behind this two-step architecture remains unclear. Why not directly forecast hydrological outputs with ML? What unique value does Prophet bring in this chain, particularly when its assumptions of periodicity and stationarity may not hold under future climate scenarios?

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the reviewer’s observation. We have clarified that Prophet serves as a decomposition tool to extract trend and seasonality from SWAT outputs, reducing noise and enhancing the input structure for ML. This improves prediction stability and interpretability over directly applying ML to raw outputs. The section is as follows:

 

The Prophet model is used not for forecasting in isolation, but as a preprocessing mechanism to decompose these SWAT outputs into trend and seasonal components. This step offers two distinct advantages: (1) it reduces high-frequency variability and captures domain-relevant periodic behavior, and (2) it provides structured input features that enhance the performance and stability of the subsequent machine learning model. Direct use of raw SWAT outputs in regression often introduces noise and undermines model accuracy, particularly in capturing multi-output hydrological responses like surface runoff, groundwater contribution, or sediment yield. Finally, multi-output regression with polynomial features is employed to learn complex nonlinear mappings between decomposed water balance variables and hydrological responses. Polynomial transformations improve the representational capacity of the model without overfitting to noise, and multi-output regression preserves interdependence among output variables—a common challenge in traditional single-output ML models. This sequential approach—physics-informed simulation → structured decomposition → nonlinear learning—was chosen specifically to balance forecast stability, interpretability, and generalization, especially for applications in watershed-scale water resource planning.

 

 

The machine learning component is not adequately justified or interpreted. The choice of using multi-output regression with polynomial features is presented without justification or comparison to other common methods in hydrology, such as random forest, gradient boosting, or LSTM networks. Moreover, the paper does not provide any feature importance analysis or SHAP values to understand model behavior. As a result, the predictive value remains a black box.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the valuable feedback. We now justify the use of multi-output regression with polynomial features for its interpretability and reduced overfitting risk, and we have added future plans to benchmark against RF, GBM, and LSTM models with SHAP-based interpretability analysis. The section is as follows:

The polynomial regression model was chosen to provide a balance between model complexity, computational efficiency, and explanatory power, especially when used in conjunction with seasonal-decomposed Prophet outputs. As future work, we plan to extend this framework by benchmarking it against ensemble models (e.g., Random Forest, XGBoost) and sequence models (e.g., LSTM) and incorporating SHAP (SHapley Additive exPlanations) values or permutation-based feature importance to enhance model interpretability and quantify individual variable contributions to hydrological outputs. This will help transition the hybrid model from a predictive tool to a diagnostic and decision-support framework in climate-sensitive watershed management.

Importantly, to address model interpretability, a limitation in many black-box models, we will employ SHAP (SHapley Additive exPlanations) for tree-based models and attention-based visualizations for LSTM models to quantify feature influence on individual predictions. This approach aims to provide both predictive strength and domain interpretability, which are critical for informing climate-resilient water policy decisions. By benchmarking the existing polynomial model against these alternatives, we aim to validate the robustness of our proposed hybrid structure and identify cases where more advanced architectures yield substantial benefit over simpler regression approaches.

 

 

Evaluation metrics are limited and do not sufficiently assess extreme event capture. The study relies on RMSE, R², and NSE, which are useful but insufficient for assessing the performance of predictive models under extreme hydrological conditions. Additional evaluation metrics (e.g., peak error, bias in high-flow events, or categorical performance in extreme years such as 2011 and 2019) would help assess the model’s reliability for disaster planning.

Response: We acknowledge the reviewer’s concern about the limitations of RMSE, R², and NSE in capturing extreme hydrological events. However, these metrics are widely accepted in hydrological modeling for evaluating overall model performance across different flow conditions. Our study focuses on simulating continuous streamflow, not exclusively extreme events, making these metrics appropriate for establishing baseline performance. The evaluation period includes extreme years like 2011 and 2019, so the metrics still reflect model behavior under such conditions. Nonetheless, we recognize the value of additional extreme-event-focused metrics and have noted this as a direction for future research in the revised manuscript.

 

 

Model performance under future climate scenarios is not convincingly validated. The authors acknowledge that SWAT-Prophet-ML performs poorly under future conditions, but the discussion lacks depth. The limitation is attributed to lack of training data, but no efforts are made to test solutions (e.g., training with scenario-based synthetic data, data augmentation, or physical constraints). The hybrid model’s scalability under changing regimes remains questionable.

Response: The comments from the reviewer are acknowledged and revised accordingly.  We acknowledge this limitation and have included the section “4. Discussion and Future Work” to include future improvements such as scenario-based synthetic training, data augmentation, and incorporation of physical constraints to enhance the hybrid model’s generalization and scalability under changing climate regimes. The section details are as follows:

 

4. Discussion and Future Work

4.1. Enhancing Model Generalization Under Future Climate Regimes

The reduced performance of the SWAT-Prophet-ML framework under future climate scenarios, as discussed in Section 3.5.1, is primarily due to the model’s reliance on historical data distributions, which do not reflect the increased variability, shifts in precipitation patterns, and altered temperature regimes expected in future conditions. Prophet and polynomial regression are fundamentally trained on stationary trends, making them vulnerable to failure under non-stationary, out-of-distribution inputs.

To address this, we propose several enhancements to improve generalization under future conditions:

• Scenario-Based Synthetic Training: Future iterations of this model will incorporate SWAT-generated outputs under multiple Representative Concentration Pathway (RCP) scenarios as training inputs, expanding the model’s exposure to a broader range of climatic conditions.
• Data Augmentation: We plan to generate perturbed versions of climate input variables (e.g., rainfall intensity shifts, temperature anomalies) using Gaussian noise or bootstrapping methods to improve robustness to rare or extreme events.
• Physically Informed Constraints: Incorporating mass balance principles directly into the loss function or model architecture (e.g., conservation-aware ML or physics-guided neural networks) will ensure hydrologic plausibility, even when data distributions deviate from historical norms.
• Transfer Learning Techniques: Pretraining models on global climate-simulated datasets and fine-tuning on watershed-specific data could improve adaptability to novel climate regimes, especially in data-scarce basins.

These directions will not only strengthen the model’s applicability to changing climate scenarios but also improve its scalability for deployment in diverse hydrological settings globally.

 

4.2. Comparative Evaluation with Ensemble and Deep Learning Models  

While this study establishes the feasibility and accuracy of the SWAT-Prophet-ML framework using multi-output regression with polynomial features, future work will expand the modeling pipeline to include state-of-the-art ensemble and deep learning models such as Random Forest (RF), Gradient Boosting Machines (GBM), and Long Short-Term Memory (LSTM) networks. These models have been extensively validated in hydrology for their ability to capture high-dimensional nonlinear relationships and temporal dependencies (Mosavi et al., 2018; Ji et al., 2021). Comparative performance evaluation will be conducted using cross-validated RMSE, NSE, and R² metrics, and statistical significance of performance differences will be assessed using paired tests. Importantly, to address model interpretability, a limitation in many black-box models, we will employ SHAP (SHapley Additive exPlanations) for tree-based models and attention-based visualizations for LSTM models to quantify feature influence on individual predictions. This approach aims to provide both predictive strength and domain interpretability, which are critical for informing climate-resilient water policy decisions. By benchmarking the existing polynomial model against these alternatives, we aim to validate the robustness of our proposed hybrid structure and identify cases where more advanced architectures yield substantial benefit over simpler regression approaches.

 

4.3. Uncertainty in Prophet Forecasts and Practical Implications

The Prophet model inherently produces uncertainty estimates via 95% prediction intervals (PIs), which account for variability in the trend, seasonality, and model residuals. These intervals are especially useful for communicating the reliability of forecasts in real-world water resource management scenarios. In this study, prediction intervals were generated alongside point forecasts for key water balance variables, including precipitation, ET, PET, and snowmelt, across both present and future climate scenarios.

While the intervals offer valuable information about forecast spread and model confidence, several limitations affect their practical interpretability:

• Under present climate conditions, Prophet’s uncertainty remain relatively narrow for smooth variables like ET and PET, enhancing confidence in monthly water demand and planning decisions.
• However, for highly variable phenomena such as precipitation and snowmelt, especially under future climate projections (2030–2042), the uncertainty intervals widen substantially. This reflects not only inherent input variability but also the model’s inability to anticipate regime shifts or outliers outside the historical data distribution.
• Because the downstream ML model in the SWAT-Prophet-ML pipeline depends on Prophet outputs, errors or overconfident predictions from Prophet may propagate, potentially affecting the reliability of surface runoff, water yield, or sediment yield predictions.

From a practical standpoint, these wide uncertainty intervals reduce the confidence of hydrologists and decision-makers in using model outputs for fine-grained policy recommendations, especially for extreme event forecasting. For example, large uncertainty in precipitation forecasts may hinder reservoir operations or flood mitigation planning.

To address these issues, future work will incorporate:

• Quantile regression forests or Bayesian models to generate more robust uncertainty estimates.
• Calibration of Prophet's uncertainty intervals using historical residual validation;
• Monte Carlo dropout or ensemble-based simulations to propagate uncertainty through the full pipeline (SWAT → Prophet → ML);
• Expressing results not just as deterministic outputs but as probabilistic forecasts that better support risk-informed decision-making.

 

 

Specific Comments

The x-axis of Figures 5, 6 is labeled from 0–40, which does not correspond to recognizable time units. Please label axes with actual years to facilitate interpretation.

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have updated the x-axis in Figures 5 and 7 to display actual years for clearer interpretation.

 

 

Although the text mentions large uncertainty bounds in Prophet predictions, these are not visualized in the figures. Adding confidence intervals or error bands would enhance the transparency of model performance and allow readers to assess reliability.

Response: We thank the reviewer for the valuable comments which have been addressed in the revised manuscript. Figures 3 and 5 present the observed vs. predicted estimates of water balances and hydrological responses for the Cahaba River Basin under present climate conditions; therefore, error bars have been added as suggested. Figures 4 and 6 illustrate projected estimates under future climate scenarios, and error bars are not included here, as there are no observed values for direct comparison. However, confidence intervals have been added to all four figures (Figures 3–6) to provide a clearer representation, in line with the reviewer’s recommendation.

 

 

Phrases such as “Prophet is a powerful tool” or “novel model” appear too frequently and are not academically objective. I recommend revising the text to adopt a more concise and formal tone. Consider professional editing or consultation with a native English speaker.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have revised the manuscript to adopt a more concise and formal academic tone, reducing subjective phrases like “powerful tool” and “novel model” to ensure clarity and objectivity.

 

 

The conclusions reiterate previous content but do not clearly summarize the main findings or limitations. A more structured summary (e.g., bullet points or short paragraphs addressing present vs. future performance, variable-specific findings, and model limitations) is recommended.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have restructured the conclusions to clearly summarize key findings, differentiate present vs. future model performance, highlight variable-specific insights, and explicitly state model limitations and future improvement directions. The section is revised as follows:

5. Conclusions

This study proposed a hybrid modeling framework, SWAT-Prophet-ML, that integrates physically based hydrological simulation (SWAT), time-series decomposition (Prophet), and machine learning (multi-output polynomial regression) to forecast monthly water balance and hydrological response variables in the climate-sensitive Cahaba watershed.

Key Findings: Present Climate Scenario (2010–2022)

• Model demonstrated strong predictive performance for water yield (R² = 0.75), • surface runoff (R² = 0.70), evapotranspiration and potential evapotranspiration (RMSE = 15–20 mm)
• Accurate modeling of seasonal trends and smooth climatic behavior was achieved using the Prophet-decomposed features.
• Precipitation and snowmelt showed higher variability and were less accurately predicted (RMSE = 30–50 mm and 5–10 mm, respectively).

Key Findings: Future Climate Scenario (2030–2042)

• Model underperformed, especially for variables like groundwater flow and sediment yield, where the hybrid model failed to capture peak years or sharp shifts.
• The main cause was the model’s reliance on stationary historical patterns, which are not representative of future climate variability.

 

Model Limitations:

• Lack of training on non-stationary or perturbed climate data
• Absence of physical constraints in ML predictions
• Limited capacity to simulate extreme events or abrupt hydrological shifts

Future Work Directions:

• Scenario-based training using synthetic climate inputs from multiple representative concentration pathways
• Data augmentation techniques to simulate rare or extreme meteorological conditions
• Physically informed modeling, integrating hydrological constraints into the ML component
• Model benchmarking using ensemble and deep learning with SHAP-based feature interpretation

          This study provides a modular, semi-automated framework that bridges physical hydrological modeling and data-driven forecasting. While highly effective under historical climate conditions, it also highlights the importance of generalization strategies for adapting predictive models to future climate regimes. The work contributes replicable architecture for modern hydrological forecasting and offers a roadmap for advancing climate-resilient water resource management.

 

 

Given that the model results hinge on customized implementation of Prophet and ML algorithms, code and input data (or at least SWAT outputs) should be made available via a repository (e.g., GitHub or Zenodo) to ensure transparency and reproducibility.

Response: The comments from the reviewer are acknowledged and revised accordingly. We agree with the reviewer and will upload the code and SWAT output data to a public GitHub repository to support transparency and reproducibility.

 

 

 

 

 

2. Comments and Suggestions for Authors – Reviewer 2

Self-citations should be reduced. Include more external ML studies (e.g., Kratzert et al., 2019; Jiang et al., 2021) and deep learning approaches to enrich the literature review.

Response: The comments from the reviewer are acknowledged and revised accordingly. The number of self-citations is reduced in the manuscript. The following references and citations are included in the manuscript.

Lange, Holger, and Sebastian Sippel. "Machine learning applications in hydrology." Forest-water interactions. Cham: Springer International Publishing, 2020. 233-257.

Zounemat-Kermani, Mohammad, et al. "Ensemble machine learning paradigms in hydrology: A review." Journal of Hydrology 598 (2021): 126266.

Xu, Tianfang, and Feng Liang. "Machine learning for hydrologic sciences: An introductory overview." Wiley Interdisciplinary Reviews: Water 8.5 (2021): e1533.

Mosaffa, Hamidreza, et al. "Application of machine learning algorithms in hydrology." Computers in earth and environmental sciences. Elsevier, 2022. 585-591.

Yang, Tao, et al. "Evaluation and machine learning improvement of global hydrological model-based flood simulations." Environmental Research Letters 14.11 (2019): 114027.

Shen, Chaopeng, Xingyuan Chen, and Eric Laloy. "Broadening the use of machine learning in hydrology." Frontiers in Water 3 (2021): 681023.

Kim, Jungho, et al. "Hybrid machine learning framework for hydrological assessment." Journal of hydrology 577 (2019): 123913.

Petty, T. R., and P. Dhingra. "Streamflow hydrology estimate using machine learning (SHEM)." JAWRA Journal of the American Water Resources Association 54.1 (2018): 55-68.

Wang, Shuo, et al. "Analysis of runoff generation driving factors based on hydrological model and interpretable machine learning method." Journal of Hydrology: Regional Studies 42 (2022): 101139.

Rozos, Evangelos, Panayiotis Dimitriadis, and Vasilis Bellos. "Machine learning in assessing the performance of hydrological models." Hydrology 9.1 (2021): 5.

 

 

Add methodological transparency:

Response: The comments from the reviewer are acknowledged and revised accordingly.

We have added detailed descriptions of each modeling step to enhance methodological transparency and ensure reproducibility in section 2 of the manuscript.

 

 

Specify training/testing split,

Response: The comments from the reviewer are acknowledged and revised accordingly. We have specified the training/testing split as 80:20 and clarified its use in model evaluation. The section is as follows:

The machine learning component of the SWAT-Prophet-ML framework was trained and evaluated using a train-test split of 80:20, where 80% of the data was used for training and 20% was held out for testing. The dataset consisted of monthly hydrological values for the period 2010–2022, with Prophet-derived water balance variables (PRECIPmm, PETmm, ETmm, SNOWMELT_PRECIP_ratio) as inputs and the corresponding SWAT-based hydrological response variables (SURQmm, GW_Qmm, SWmm, WYLDmm, SYLDt_ha) as outputs.

To ensure robust model validation, a 5-fold cross-validation was performed on the training set during hyperparameter tuning. Model performance was assessed using RMSE, R², and Nash-Sutcliffe Efficiency (NSE) across both training and testing sets. Predictions were inverse transformed to their original scale using the MinMaxScaler to allow for direct comparison with actual hydrological outputs.

 

 

Clarify cross-validation technique,

Response: The comments from the reviewer are acknowledged and revised accordingly. We have clarified that 5-fold cross-validation was used during training to ensure robust model evaluation. The section is as follows:

To ensure robust model validation, a 5-fold cross-validation was performed on the training set during hyperparameter tuning. Model performance was assessed using RMSE, R², and Nash-Sutcliffe Efficiency (NSE) across both training and testing sets. Predictions were inverse transformed to their original scale using the MinMaxScaler to allow for direct comparison with actual hydrological outputs.

 

 

Mention any hyperparameter tuning strategies. Improve English clarity, especially in technical descriptions and interpretation of results.

Response: We appreciate the reviewer’s suggestion regarding hyperparameter tuning strategies and improved technical clarity. However, detailed hyperparameter optimization and extended interpretation of results are beyond the scope of the current study, which focuses primarily on evaluating the hydrological impacts under different climate scenarios using established model configurations. While basic model calibration was performed to ensure reasonable performance, comprehensive tuning was not the central aim. We agree that such analysis could enhance model precision and interpretability, and we suggest it as a valuable direction for future research. We have also checked and made sure that the clarity in the usage of English language is depicted well throughout the manuscript.

 

 

Discuss uncertainty from the Prophet model, including prediction intervals, and how it affects the model’s practical use.

Response: The comments from the reviewer are acknowledged and revised accordingly. We have added a discussion on Prophet’s prediction intervals, highlighting their role in conveying forecast uncertainty and their impact on the reliability of downstream hydrological predictions under the section “4.3. Uncertainty in Prophet Forecasts and Practical Implications”.

 

 

Future model extensions could include physics-informed ML or climate-driven data augmentation to improve generalization.

Response: The comments from the reviewer are acknowledged and revised accordingly. This is an insightful comment and suggestion. We acknowledge it and will continue to work with these areas in future research for SWAT ML Prophet based modeling. We agree and have outlined future extensions including physics-informed machine learning and climate-driven data augmentation to enhance model generalization under non-stationary conditions in section “4. Discussion and Future Work”.

 

 

 

 

 

3. Comments and Suggestions for Authors – Reviewer 3

This research has certain significance, but there are also the following problems:

1.There is no mapping display in the study area. A mapping display should be made for the Kahaba River Basin and the eight sub-basins it is divided into.

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have added a mapping display showing the Cahaba River Basin and its division into eight sub-basins to enhance spatial context and understanding.

Fig 2: (a) Map showing the Cahaba River and its major tributaries in central Alabama. The map indicates the Cahaba's location in the southeastern United States. Key historical and ecological localities are also highlighted; (b) Location of the Alabama River Basin covering parts of central and southern Alabama. The Cahaba River sub-basin lies within this system and serves as the focus of this study. 

 

2.In the Data and Methods section, tables should be used to introduce the sources of all the data used by the research institute along with the website addresses. Information on the spatio-temporal resolution of the data should also be provided

Response: The comments from the reviewer are acknowledged, revised accordingly, and added Table 1 as follows:

Table 1

Input data used in developing the Soil and Water Assessment Tool (SWAT) model for Cahaba River Basin.

Data	Data sources	Information	Period	Address/Location
Digital elevation models (DEM)	Web GIS	Raster, 30 m	2011	WebGIS - Geographic Information Systems Resource - GIS
Land use land cover	United States Geological Survey, USGS	Raster, 30 m	2011	Annual National Land Cover Database | U.S. Geological Survey
Soil data	United States Department of Agriculture, USDA	Raster, 60 m	2011	Web Soil Survey - Home
Climate data	Climate.gov	Daily	1980-2010 2010-2040	Search | Climate Data Online (CDO) | National Climatic Data Center (NCDC)
Hydrological data	United States Geological Survey, USGS	Monthly	2011-2017	Cahaba River at Trussville, Al. - USGS Water Data for the Nation

 

 

 

3.In the results section of this study, the simulated values of the SWAT model were used as the actual values to compare the results of the fusion model. However, the simulation accuracy of the SWAT model was not demonstrated. A subsection should be added at the very beginning of the result to verify the accuracy of the SWAT model using the collected actual observation data. When the output accuracy of some hydrological components of the SWAT model is relatively high, the credibility of all its components as actual values is relatively high.

Response: The comments from the reviewer are acknowledged and revised accordingly. A subsection 3.1 and Table 3 are included in the manuscript that verifies the accuracy of the SWAT model using the collected actual observation data which enhances the credibility of the existing model and novel models of the study.

3.1. SWAT Model Accuracy and Calibration Settings

SWAT was effectively applied to simulate hydrological responses within the Cahaba watershed under varying land use and climatic conditions. The model was initially configured using the 2011 land use/land cover and 1980–2010 and 2010-2040 climate data across eight sub-basins, 15 land use classes, and 30 soil categories. It was later calibrated using 2011 LULC under the same climate conditions. Hydrologic calibration and validation were conducted at the Trussville station in the upper Cahaba watershed. During calibration, wet conditions prevailed, and major peak flows observed in late March and early April were attributed to late snowmelt and spring runoff. The model demonstrated reasonable accuracy, achieving NSE and R2 values of 0.565 and 0.591 respectively. The coefficient of determination (R²) improved from 0.542 during calibration to 0.591 during validation, potentially due to the dominance of low streamflow events in the validation period, which reduced variability and increased correlation (Table 3). Key parameters such as ESCO (Soil Evaporation Compensation Factor) and CN2 (SCS Curve Number) were identified as highly sensitive to streamflow under wet conditions. Nutrient-related parameters (N_UPDIS, P_UPDIS) and urban erosion indicators (RILL_MULT, C_FACTOR) also showed significant influence on model outputs.

Table 3: SWAT model calibration parameters and model performance evaluation.

Parameter	Description	Calibration Range	Final Calibrated Value	NSE	R2
CN2	SCS curve number	1.00 – 2.00	1.63	0.430	0.456
ESCO	Soil evaporation compensation factor	0.85 – 1.00	0.91	0.465	0.483
P_UPDIS	Phosphorus uptake distribution	20 – 40	31	0.502	0.542
N_UPDIS	Nitrogen uptake distribution	20 – 40	24	0.565	0.591

 

 

 

 

4.It is recommended that the X-axis in Figures 5 and 6 be marked with the actual year and month instead of the months starting from 2010 and 2030

Response: The comments from the reviewer are acknowledged and revised accordingly. We appreciate the suggestion and have updated the x-axis in Figures 5 and 6 to display actual years for clearer interpretation.

 

 

5.I think that when organizing the results section, the water balance prediction and hydrological response prediction under the Present climate should be analyzed first, and then the water balance prediction and hydrological response under the future climate should be analyzed. That is, the organization sequence should be the existing sequence of sections 3.1, 3.3, 3.2, and 3.4.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have reorganized the section “3. Results” to follow the sequence 3.1, 3.3, 3.2, and 3.4, presenting water balance and hydrological response predictions under present climate first, followed by future climate analysis.

 

 

6.How exactly were the model performance evaluation results in Section 3.5 obtained? A detailed introduction should be given. If the simulation results are evaluated using actual observational data, why were the simulation results of the SWAT model taken as the actual values in the previous studies?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the comment. We have added clarification in Section 3.6.1 explaining that due to limited observational data across all variables and sub-basins, the calibrated SWAT model, previously validated with available observations, was used as a reference baseline for evaluating the hybrid model’s performance (“3.6.1 Evaluation Methodology and Use of SWAT as Reference”). A subsection “3.1. SWAT Model Accuracy and Calibration Settings” and Table 3 are also included in the manuscript that verifies the accuracy of the SWAT model using the collected actual observation data which enhances the credibility of the existing model and novel models of the study.

 

 

 

 

4. Comments and Suggestions for Authors – Reviewer 4

The article addresses water resource management by hydrological variable prediction

titled “Predictive Analysis of Hydrological Variables in the Cahaba Watershed: Enhancing

Forecasting Accuracy for Water Resource Management Using Time-Series and Machine

Learning Models.” Having familiarized myself with the manuscript, I have some

suggestions:

Major Comments:

1. In the introduction section, the authors should consider restructuring it to provide

general background information, specific background information, and a

description of the gap in our knowledge that the study was designed to fill.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the suggestion. We have restructured the section “1.Introduction” to first present general background, then specific context on SWAT, Prophet, and ML methods, followed by a clear statement of the research gap and the objective our study aims to address.

 

 

2. Section 2.2, how did the authors account for vegetation heterogeneity with varying

spatial variability of soil moisture? How many layers does the SWAT model

consider? And how does it control this behaviour/pattern?

Response: We thank the reviewer for this insightful question. In the revised manuscript, we have expanded Section 2.2 to clarify how the Soil and Water Assessment Tool (SWAT) accounts for vegetation heterogeneity and the spatial variability of soil moisture.

SWAT handles spatial heterogeneity using Hydrologic Response Units (HRUs), which are unique combinations of land use, soil type, and slope within each sub-basin. This allows the model to simulate distinct vegetation and soil interactions within a watershed, thereby capturing the influence of vegetation heterogeneity on soil moisture dynamics. SWAT uses a multi-layer soil profile that typically includes 1 to 10 layers, with the number and depth of layers defined based on the soil input data. Each layer has its own physical properties such as texture, hydraulic conductivity, and available water capacity which influence water movement and storage. In our study, we used the first 4 layers as per the web soil survey-based soil database, ensuring adequate representation of vertical soil heterogeneity. Vegetation influences soil moisture through evapotranspiration and root zone depth. SWAT controls these processes through vegetation indices that vary by land cover type and soil type. Thus, areas with different vegetation types and growth conditions exhibit distinct soil moisture patterns captured by the HRUs of the Cahaba watershed.

 

 

3. Section 2.2, In watershed delineation, what threshold values for the soil, slope and

land use were used to define the hydrological response units/

Response: The comments from the reviewer are acknowledged and revised accordingly. In the revised manuscript, we have expanded Section 2.2 to clarify what threshold values for the soil, slope and land use were used to define the hydrological response units in watershed delineation.

The area threshold for watershed delineation is 1000 ha and the percentage threshold values for the soil, slope and land use that were used to define the hydrological response units are 10%, 10%, and 5% respectively.

 

 

4. Section 2.4: What are the limitations of using the Prophet model in capturing peak

events?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the question. We have addressed this in a new Section “3.2.1. Limitations in Capturing Peak Events” noting that Prophet’s additive model structure tends to smooth short-duration peaks such as heavy rainfall or snowmelt treating them as outliers, which limits its ability to accurately capture extreme hydrological events.

3.2.1. Limitations in Capturing Peak Events

While the Prophet model effectively decomposes time series into trend and seasonal components, it exhibits notable limitations in capturing sharp, short-duration peak events, such as heavy rainfall spikes or sudden snowmelt. This is largely due to Prophet’s underlying additive model structure, which assumes smooth and regular seasonal patterns. As a result, it tends to smooth over localized outliers, treating them as noise rather than meaningful extremes. Furthermore, Prophet assumes piecewise linear or logistic growth for the trend component and may fail to adapt to abrupt shifts or high-frequency variability unless such events are consistently present in the historical data. In the context of hydrological forecasting, this behavior limits the model’s ability to anticipate critical extreme events that significantly influence surface runoff, flash flooding, or sediment transport (Wamg et al., 2022). For instance, as shown in Figure 4(b) and 4(c), Prophet underestimates peak values during storm months, leading to under-propagation of signal amplitude into downstream ML predictions. Future model improvements will consider integrating spike-aware models, such as quantile regression or hybrid Prophet-LSTM structures, and incorporating event-based decomposition techniques to better preserve and forecast peak behaviors.

 

 

5. Section 2.6: Did the authors consider different CO2 concentration for baseline and

future projections? How many models was used for the study?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the question. In Section “2.6. Hydrological Scenarios”, we clarify that future climate projections were based on SWAT simulations driven by outputs under Representative Concentration Pathways (RCPs), which inherently account for varying CO₂ concentrations. One calibrated SWAT model was used, with scenario-based inputs applied for future simulations.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5. Comments and Suggestions for Authors – Reviewer 5

Evaluation of the article entitled “Predictive Analysis of Hydrological Variables in the Cahaba Watershed: Enhancing Forecasting Accuracy for Water Resource Management Using Time-Series and Machine Learning Models”, submitted to Earth.

 

 

The article addresses an important and current topic, with integration of SWAT with Prophet and machine learning for hydrological forecasting; It aims to predict environmental variables and their hydrological impacts in current and future climate scenarios; The SWAT-Prophet-ML model showed strong predictive performance for water production and surface runoff. This hybrid combination (model) achieved 86.73% accuracy in current climate forecasts, suggesting scalability for water resource planning.

 

General analysis: The article is very truncated and without a structured logical sequence. The figures have low resolution and quality. The main limitations are linked to the dependence on historical trends and poor performance in high-variability events, such as precipitation.

Response: The comments from the reviewer are acknowledged and revised accordingly.  Thank you for the feedback. We have revised the manuscript to improve logical flow and structure, replaced low-resolution figures with high-quality versions, and expanded the discussion on model limitations particularly its reliance on historical trends and reduced accuracy during high-variability events like precipitation.

 

 

- What are the objectives of the work? This is not clear in the structure of the article, so are the conclusions bad and need to be redone?

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for your observation. We have revised the manuscript to clearly state the study’s objectives at the end of Section “2.1. Methodology Justification for Hybrid Model Design” and restructured the conclusions to directly address these objectives, summarizing key findings, limitations, and implications more clearly.

 

 

- In general, there was too much concern in presenting the models. Still, the understanding/discussion of the seasonality of the hydroclimatic variables is very weak, including associations with the studied river basin.

Response: We thank the reviewer for the valuable comment. We would like to respectfully clarify that the manuscript does include discussion of the seasonality of key hydroclimatic variables (precipitation, snowmelt, evapotranspiration, and potential evapotranspiration) and key hydrological response variables (surface runoff contribution to streamflow, groundwater contribution to streamflow, soil water content, water yield, and sediment yield) for Cahaba watershed. Specifically, seasonal trends were presented in Sections 2.6, 3.1, 3.2, 3.3 and Tables 4 and 5 as follows:

 

“2.6. Hydrological Scenarios 

This timeframe was selected to capture recent historical variability in climate and hydrological responses, including seasonal and interannual fluctuations in rainfall, snowmelt, ET, PET, surface runoff, groundwater flow, soil water content, water yield, and sediment yield.

 

3.1. SWAT vs. SWAT-Prophet-ML for Water Balance Predictions in Present Climate

Historical trends in environmental variables such as ET, PET, precipitation, and snowmelt show distinct seasonal patterns (Figure 3).

 

3.2. SWAT-Prophet-ML based Water Balance Predictions in Future Climate

From 2030 to 2042, projections of the four water balance components show strong seasonal  trends using the SWAT-Prophet-ML model (Figure 4). ET and PET maintain smooth, consistent  annual cycles, indicating high model reliability for temperature-driven processes. PET consistently exceeds ET by 50% - 65%, aligning with theoretical expectations as per the novel model outcomes. In future climate, precipitation ranging between 20 mm and 150 mm and snowmelt ranging between 0 mm and 29 mm exhibit more variability, with sharp peaks suggesting possible extreme weather events. Despite this, both retain regular annual patterns, reflecting the model's strength in capturing seasonality. The SWAT-Prophet-ML modeling results show that the future climate predictions of precipitation (20 mm – 150 mm) is decreasing relative to the present climate estimates of precipitation (40 mm-165 mm) whereas the future climate predictions of snowmelt (0 mm – 29 mm) is increasing relative to the present climate estimates of snowmelt (0 mm – 26 mm). This indicates that rainfall is likely to decrease in future with increases in snow melting and soil water accumulation (Labat et al., 2004). These patterns suggest stable climatic behavior over the projection period, though the seasonal fluctuations highlight areas for model refinement for the SWAT-Prophet-ML model (Tandon et al., 2025).

 

3.3. SWAT vs. SWAT-Prophet-ML for Hydrological Response Predictions in Present Climate

The predictions of surface runoff contribution to streamflow and water yields in both models are in good correlation with coefficient of determination, R2 values of 0.65 and 0.75 respectively. It also highlights significant runoff and water yield patterns in response to extreme precipitation events for the years 2011 and 2019 (Figure 3). These historical patterns have been instrumental in understanding seasonal variations and system responses (Preetha and Al-Hamdan 2020b; Preetha and Joseph 2025). This comparison highlights areas where the prediction model aligns closely with observed data and areas where it deviates, offering critical insights into the model's strengths and limitations.”

 

 

Suggestions for revisions:

- The article does not comply with the standards established in the MDPI Template for citations, references and other textual components.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for pointing this out. We have revised the manuscript to fully align with the MDPI template, ensuring correct formatting of citations, references, and all textual components as per journal guidelines.

 

 

- The figures are presented without previous calls and descriptions in the text; in general, they do not present the variables and units of measurement identified in the respective axes, they confuse “date” with “time”; the titles are not very explanatory.

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the valuable feedback. We have revised the manuscript to ensure all figures are properly introduced and described in the text, updated axis labels with appropriate variables and units, corrected terminology (e.g., “date” vs. “time”), and improved figure titles for clarity and interpretability.

 

 

- there is no information, maps, climate, vegetation, soils or details of the physiographic characteristics of the studied river basin

Response: The comments from the reviewer are acknowledged and revised accordingly. We added Table 1 which shows the information about climate, vegetation, soils, and details of the physiographic characteristics of the river basin studied. Maps of the river basin are also included as Figure 1.

Table 1

Input data used in developing the Soil and Water Assessment Tool (SWAT) model for Cahaba River Basin.

Data	Data sources	Information	Period	Address/Location
Digital elevation models (DEM)	Web GIS	Raster, 30 m	2011	WebGIS - Geographic Information Systems Resource - GIS
Land use land cover	United States Geological Survey, USGS	Raster, 30 m	2011	Annual National Land Cover Database | U.S. Geological Survey
Soil data	United States Department of Agriculture, USDA	Raster, 60 m	2011	Web Soil Survey - Home
Climate data	Climate.gov	Daily	1980-2010 2010-2040	Search | Climate Data Online (CDO) | National Climatic Data Center (NCDC)
Hydrological data	United States Geological Survey, USGS	Monthly	2011-2017	Cahaba River at Trussville, Al. - USGS Water Data for the Nation

 

 

 

 

- The use of data only from 2011 for calibration of the SWAT is insufficient, since the model needs data for “warming up” and later, calibration and validation;

Response: The comments from the reviewer are acknowledged and revised accordingly. Thank you for the important observation. We acknowledge this limitation and have updated the methodology to clarify that additional years that were used for model warm-up, and that calibration and validation were performed using a multi-year dataset to ensure model stability and reliability.

Monthly calibration simulations were conducted from November 2013 to October 2017, with the first two years used as a warm-up period (NYSKIP = 2) under a skewed normal rainfall distribution. Streamflow data from the Trussville station, representing the outlet of sub-basin 1 (including HRUs 1–5), was used for calibration and validation.

 

 

- Sensitivity analyses of the SWAT parameters for the river basin were not presented.

Response: The comments from the reviewer are acknowledged and revised accordingly. A section “3.1. SWAT Model Accuracy and Calibration Settings” is added that shows the sensitivity of the SWAT parameters, the calibration ranges and the best fit values for the Cahaba watershed.

3.1. SWAT Model Accuracy and Calibration Settings

SWAT was effectively applied to simulate hydrological responses within the Cahaba watershed under varying land use and climatic conditions. The model was initially configured using the 2011 land use/land cover and 1980–2010 and 2010-2040 climate data across eight sub-basins, 15 land use classes, and 30 soil categories. It was later calibrated using 2011 LULC under the same climate conditions. Hydrologic calibration and validation were conducted at the Trussville station in the upper Cahaba watershed. During calibration, wet conditions prevailed, and major peak flows observed in late March and early April were attributed to late snowmelt and spring runoff. The model demonstrated reasonable accuracy, achieving NSE and R2 values of 0.565 and 0.591 respectively. The coefficient of determination (R²) improved from 0.542 during calibration to 0.591 during validation, potentially due to the dominance of low streamflow events in the validation period, which reduced variability and increased correlation (Table 3). Key parameters such as ESCO (Soil Evaporation Compensation Factor) and CN2 (SCS Curve Number) were identified as highly sensitive to streamflow under wet conditions. Nutrient-related parameters (N_UPDIS, P_UPDIS) and urban erosion indicators (RILL_MULT, C_FACTOR) also showed significant influence on model outputs.

 

Table 3: SWAT model calibration parameters and model performance evaluation.

Parameter	Description	Calibration Range	Final Calibrated Value	NSE	R2
CN2	SCS curve number	1.00 – 2.00	1.63	0.430	0.456
ESCO	Soil evaporation compensation factor	0.85 – 1.00	0.91	0.465	0.483
P_UPDIS	Phosphorus uptake distribution	20 – 40	31	0.502	0.542
N_UPDIS	Nitrogen uptake distribution	20 – 40	24	0.565	0.591

 

 

- In Figure 5, what happened to the GW-Qmm variable in the “current” variation? What does “time” mean on the X-axis of these figures, if the title says it is from 2010-2022?

Response: The comments from the reviewer are acknowledged and revised accordingly.  We appreciate the suggestion and have revised the figures to clearly show the trends of the variable GW-Qmm in the watershed. We have also updated the x-axis in the figure to display actual years for clearer interpretation.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I have no further comments.

Author Response

Response to Reviewer’s Comments on Earth Manuscript - 3694442

 

The authors are grateful for the Reviewers and Editor for their insightful comments and recommendations.

 

1. Comments and Suggestions for Authors – Reviewer 1

 

I have no further comments.

Response: The comment from the reviewer is acknowledged.

 

 

2. Comments and Suggestions for Authors – Reviewer 3

 

All the initial review comments have been revised and answered. I would like to offer some minor suggestions on the revised content.

Response: The comment from the reviewer is acknowledged.

 

 

Firstly, among the eight sub-basins added in part (b) of Figure 2, some have the same numbers (there are two 5s and two 7s), and the content of the legend is unclear

Response: The comments from the reviewer are acknowledged and revised accordingly. Figure 2 (b) is revised to show the delineated Cahaba River watershed using SWAT model. The content of the legend is revised to improve the clarity of the figure.

Fig 2: (a) Map showing the Cahaba River and its major tributaries in central Alabama. The map indicates the Cahaba's location in the southeastern United States. Key historical and ecological localities are also highlighted; (b) Location of the Cahaba River Basin covering parts of central and southern Alabama delineated using SWAT model. The Cahaba River sub-basin lies within this system and serves as the focus of this study. 

 

 

Secondly, please modify all the tables to the same format. It seems that only the added table 1 is in the format of a three-line table.

Response: The comments from the reviewer are acknowledged and revised accordingly. All the tables in the manuscript are revised to the same format as that of Table 1.

 

 

3. Comments and Suggestions for Authors – Reviewer 5

The authors justified and provided changes to the comments from the first review; however, they did not highlight them in the file (using different colors), making it difficult to quickly identify the corrections/changes.

Response: The comments from the reviewer are acknowledged and revised accordingly. All the changes provided to the comments from the first review are highlighted in the file (using yellow color), making it easy to quickly identify the corrections/changes.

 

 

The most sensitive points related to SWAT calibration were better clarified, and the relationships between objectives and conclusions were also explored (in a less than creative way in terms of scientific writing).

Response: The comments from the reviewer are acknowledged and revised accordingly.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

All the initial review comments have been revised and answered. I would like to offer some minor suggestions on the revised content.

Firstly, among the eight sub-basins added in part (b) of Figure 2, some have the same numbers (there are two 5s and two 7s), and the content of the legend is unclear

Secondly, please modify all the tables to the same format. It seems that only the added table 1 is in the format of a three-line table.

Author Response

Response to Reviewer’s Comments on Earth Manuscript - 3694442

 

The authors are grateful for the Reviewers and Editor for their insightful comments and recommendations.

 

1. Comments and Suggestions for Authors – Reviewer 1

 

I have no further comments.

Response: The comment from the reviewer is acknowledged.

 

 

2. Comments and Suggestions for Authors – Reviewer 3

 

All the initial review comments have been revised and answered. I would like to offer some minor suggestions on the revised content.

Response: The comment from the reviewer is acknowledged.

 

 

Firstly, among the eight sub-basins added in part (b) of Figure 2, some have the same numbers (there are two 5s and two 7s), and the content of the legend is unclear

Response: The comments from the reviewer are acknowledged and revised accordingly. Figure 2 (b) is revised to show the delineated Cahaba River watershed using SWAT model. The content of the legend is revised to improve the clarity of the figure.

Fig 2: (a) Map showing the Cahaba River and its major tributaries in central Alabama. The map indicates the Cahaba's location in the southeastern United States. Key historical and ecological localities are also highlighted; (b) Location of the Cahaba River Basin covering parts of central and southern Alabama delineated using SWAT model. The Cahaba River sub-basin lies within this system and serves as the focus of this study. 

 

 

Secondly, please modify all the tables to the same format. It seems that only the added table 1 is in the format of a three-line table.

Response: The comments from the reviewer are acknowledged and revised accordingly. All the tables in the manuscript are revised to the same format as that of Table 1.

 

 

3. Comments and Suggestions for Authors – Reviewer 5

The authors justified and provided changes to the comments from the first review; however, they did not highlight them in the file (using different colors), making it difficult to quickly identify the corrections/changes.

Response: The comments from the reviewer are acknowledged and revised accordingly. All the changes provided to the comments from the first review are highlighted in the file (using yellow color), making it easy to quickly identify the corrections/changes.

 

 

The most sensitive points related to SWAT calibration were better clarified, and the relationships between objectives and conclusions were also explored (in a less than creative way in terms of scientific writing).

Response: The comments from the reviewer are acknowledged and revised accordingly.

Author Response File: Author Response.docx

Reviewer 5 Report

Comments and Suggestions for Authors

The authors justified and provided changes to the comments from the first review; however, they did not highlight them in the file (using different colors), making it difficult to quickly identify the corrections/changes.

The most sensitive points related to SWAT calibration were better clarified, and the relationships between objectives and conclusions were also explored (in a less than creative way in terms of scientific writing).

Author Response

Response to Reviewer’s Comments on Earth Manuscript - 3694442

 

The authors are grateful for the Reviewers and Editor for their insightful comments and recommendations.

 

1. Comments and Suggestions for Authors – Reviewer 1

 

I have no further comments.

Response: The comment from the reviewer is acknowledged.

 

 

2. Comments and Suggestions for Authors – Reviewer 3

 

All the initial review comments have been revised and answered. I would like to offer some minor suggestions on the revised content.

Response: The comment from the reviewer is acknowledged.

 

 

Firstly, among the eight sub-basins added in part (b) of Figure 2, some have the same numbers (there are two 5s and two 7s), and the content of the legend is unclear

Response: The comments from the reviewer are acknowledged and revised accordingly. Figure 2 (b) is revised to show the delineated Cahaba River watershed using SWAT model. The content of the legend is revised to improve the clarity of the figure.

Fig 2: (a) Map showing the Cahaba River and its major tributaries in central Alabama. The map indicates the Cahaba's location in the southeastern United States. Key historical and ecological localities are also highlighted; (b) Location of the Cahaba River Basin covering parts of central and southern Alabama delineated using SWAT model. The Cahaba River sub-basin lies within this system and serves as the focus of this study. 

 

 

Secondly, please modify all the tables to the same format. It seems that only the added table 1 is in the format of a three-line table.

Response: The comments from the reviewer are acknowledged and revised accordingly. All the tables in the manuscript are revised to the same format as that of Table 1.

 

 

3. Comments and Suggestions for Authors – Reviewer 5

The authors justified and provided changes to the comments from the first review; however, they did not highlight them in the file (using different colors), making it difficult to quickly identify the corrections/changes.

Response: The comments from the reviewer are acknowledged and revised accordingly. All the changes provided to the comments from the first review are highlighted in the file (using yellow color), making it easy to quickly identify the corrections/changes.

 

 

The most sensitive points related to SWAT calibration were better clarified, and the relationships between objectives and conclusions were also explored (in a less than creative way in terms of scientific writing).

Response: The comments from the reviewer are acknowledged and revised accordingly.

Author Response File: Author Response.docx