The Operational Nitrogen Indicator (ONI): An Intelligent Index for the Wastewater Treatment Plant’s Optimization

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study presents an approach based on a virtual sensor architecture, which is designed to estimate total nitrogen levels in the effluent and assess plant performance using an operational indicator. This approach contributes to a shift toward smart wastewater treatment plants and further to improve process efficiency and reduce operational risk. Overall, this work can be recommended for publication after some points should be addressed in more details:

1. The authors should present the representative data for the samples for developing the model, such as the data collected and recorded per day.
2. More importantly, the authors should simply validate the performance of this approach in a real or simulated WWTP environment.
3. The dataset for validating agent approach should consist of four variables, but in Table 3, only three variables were listed. In addition, from the equation 1 for calculating Nreal%, it can be found that the lower the effluent nitrogen is, the higher the Nreal% value is. Is this equation correct?

Author Response

Comments 1: This study presents an approach based on a virtual sensor architecture, which is designed to estimate total nitrogen levels in the effluent and assess plant performance using an operational indicator. This approach contributes to a shift toward smart wastewater treatment plants and further to improve process efficiency and reduce operational risk. Overall, this work can be recommended for publication after some points should be addressed in more details:
1. The authors should present the representative data for the samples for developing the model, such as the data collected and recorded per day.
Response 1: Thank you for your comments. To provide more information about the dataset, the number of samples and the distribution of the variables present have been added. This new information is included on page 11, section 3.2, starting on line 395-398.

Comments 2: 2. More importantly, the authors should simply validate the performance of this approach in a real or simulated WWTP environment.
Response 2: We greatly appreciate this comment. Indeed, we are aware of the importance of validating the performance of the proposed approach in a wastewater treatment plant environment, whether real or simulated. However, this validation has not been possible at this stage of the work due to technical limitations and infrastructure access.
Nevertheless, this validation constitutes a priority line of research and is already included in the future work section of the manuscript. We plan to implement and evaluate the approach in a more realistic simulated environment, and subsequently in real-world conditions, in order to analyze its operational performance and practical applicability.

Comments 3: 3. The dataset for validating agent approach should consist of four variables, but in Table 3, only three variables were listed. In addition, from the equation 1 for calculating Nreal\%, it can be found that the lower the effluent nitrogen is, the higher the Nreal\% value is. Is this equation correct?
Response 3: Thank you for your comment. Indeed, the variable corresponding to total effluent nitrogen was not included in Table 3 due to an oversight. This has since been corrected, and the updated version of that table is shown below, which now correctly includes the four variables used for the validation of the agent approach.

Regarding equation 1 for the calculation of \%Nreal, we appreciate the comment. There was an error in the denominator of the original formula. Indeed, for the indicator to consistently reflect nitrogen yield in the effluent, the denominator must be the legally permitted nitrogen value, and not the value measured in the effluent, as was erroneously stated. The corrected formula is as follows:

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Review Comments

The paper presents a promising framework for smart WWTP optimization. Here are some comments which need to be addressed.

1. There is an inconsistency with the usage of acronym "ONI". It is defined as Operational Nitrogen Indicator (Abstract) but later called Oxidative Nitrogen Index.
2. Can ONI thresholds be auto-calibrated for individual WWTPs?
3. The virtual sensor requires real-time TKN measurements. TKN analysis is typically lab-based and slow. How is this addressed?
4. Outliers in high-nitrogen predictions (Fig. 10) may compromise ONI reliability. Was outlier handling explored?
5. Gradient Boosting (GB) outperformed all models (Table 12: R²=0.81), yet Decision Tree (DT) was chosen for the virtual sensor (page 26). No justification provided.

Author Response

Comments 1: The paper presents a promising framework for smart WWTP optimization. Here are some comments which need to be addressed.
    1. There is an inconsistency with the usage of acronym "ONI". It is defined as Operational Nitrogen Indicator (Abstract) but later called Oxidative Nitrogen Index.
Response 1: Thank you for pointing out this inconsistency. Indeed, it was a terminological confusion, and it has been corrected throughout the manuscript. We have carefully reviewed the use of the acronym "ONI," and it is now used consistently with the correct definition throughout all sections of the document.

Comments 2: 2. Can ONI thresholds be auto-calibrated for individual WWTPs?
Response 2: Thank you for your question, which we consider very pertinent.

The values taken into account to calculate the four sub-indicators that make up the ONI, and subsequently the ONI indicator itself, are based on data measured directly at the plant, on values established by current regulations, and on information derived from the literature. Therefore, there is no parameter specifically adjusted for this WWTP in the case study. In this sense, the approach is fully generalizable to other treatment plants.Regarding the thresholds defined to interpret the final ONI value, these are proposed as indicative alert ranges. While they can serve as an initial reference in different contexts, they can be adjusted manually according to the preferences or needs of the operator responsible for monitoring or controlling the plant, rather than due to a lack of adaptability to other facilities.

Comments 3: 3. The virtual sensor requires real-time TKN measurements. TKN analysis is typically lab-based and slow. How is this addressed?
Response 3: Thank you for your question, which we consider very important. It is true that real-time measurement of TKN can be a limitation in many plants, as it is traditionally performed through laboratory analysis. However, in the case study considered, the plant has an automated and integrated laboratory that allows for frequent TKN measurements. Furthermore, advanced sensor developments (e.g., using UV spectroscopy or correlation technologies) are currently being developed that allow TKN to be estimated with sufficient accuracy for operational purposes, opening the door to real-time application in the near future.

Comments 4: 4. Outliers in high-nitrogen predictions (Fig. 10) may compromise ONI reliability. Was outlier handling explored?
Response 4: Once again, we sincerely appreciate your input. However, we are not currently considering handling outliers with any method. While validating the virtual sensor's performance in a real-world setting, we cannot determine the impact of these outliers. If the impact is significant, we will consider statistical methods such as the interquartile range (IQR) or standard deviations to detect and eliminate them.

Comments 5: 5. Gradient Boosting (GB) outperformed all models (Table 12: R²=0.81), yet Decision Tree (DT) was chosen for the virtual sensor (page 26). No justification provided.
Response 5: Again, thank you very much for your review. This was a typo where Decision Tree was written instead of Gradient Boosting. We apologize for the error, and it has been corrected, choosing Gradient Boosting as the model. This was corrected on page 28, section 4.4.3 on line 796.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors The authors set the stage to present a novel approach based on a virtual sensor architecture designed to estimate total nitrogen levels in the effluent and assess plant performance using an operational indicator. Heart of the system is an intelligent agent that integrates real-time sensor data with machine learning models to infer nitrogen dynamics and anticipate deviations from operating system conditions. Central component is the Operational Nitrogen Indicator (ONI), a weighted  aggregation of four sub-indicators (legal compliance, nitrogen dynamic trend, removal efficiency & microbial balance). each of which captures a critical dimension of the nitrogen removal process. It enables early detection of stress conditions and facilitates adaptive decision-making by quantifying operational status in terms of regulatory thresholds, biological requirements, and dynamic stability. Target of the authors is to implement the complete system in a real wastewater treatment plant with aim of validating its performance and evaluate its robustness, adaptability and practical utility in real-life scenarios. This study is interesting but there are some issues that have to be addressed.   1) In introduction, the authors should mention other decision-making models for WWT in literature and compare them with their own.   2) In Table, the author should add the limits (if there are) of the four indicators   3) Except oxygen control sensor, have the authors take into account the pH and the temperature effect?   4) The authoρs should add a figure of the WWT plant   5) Linear and polynomial regression equations should be added   6) A representative figure of a decision tree should be added

Author Response

Comments 1: The authors set the stage to present a novel approach based on a virtual sensor architecture designed to estimate total nitrogen levels in the effluent and assess plant performance using an operational indicator. Heart of the system is an intelligent agent that integrates real-time sensor data with machine learning models to infer nitrogen dynamics and anticipate deviations from operating system conditions. Central component is the Operational Nitrogen Indicator (ONI), a weighted  aggregation of four sub-indicators (legal compliance, nitrogen dynamic trend, removal efficiency & microbial balance). each of which captures a critical dimension of the nitrogen removal process. It enables early detection of stress conditions and facilitates adaptive decision-making by quantifying operational status in terms of regulatory thresholds, biological requirements, and dynamic stability. Target of the authors is to implement the complete system in a real wastewater treatment plant with aim of validating its performance and evaluate its robustness, adaptability and practical utility in real-life scenarios. This study is interesting but there are some issues that have to be addressed.   1) In introduction, the authors should mention other decision-making models for WWT in literature and compare them with their own.
Response 1: We appreciate the comment. A new paragraph has been added to the introduction briefly comparing existing approaches in the literature related to decision-making in WWTPs, particularly those based on indicators or decision support systems. The added paragraph is shown below:

\corrected{Although this proposal introduces a novel approach, it is important to compare it with other decision-making models applied in WWTPs. Among these, multi-criteria decision-making frameworks (MCDM), such as AHP (Analytic Hierarchy Process) or TOPSIS, are widely used to evaluate and prioritize treatment alternatives or operational strategies [24,25]. Likewise, artificial intelligence approaches or expert systems have been developed to provide operational recommendations based on historical data analysis [26]. Compared to these models, the approach presented in this work is distinguished by integrating a virtual sensor and a dynamic operational indicator into an autonomous agent capable of real-time adaptation and proactive decision-making. This integration enhances response capability and reduces dependence on predefined static rules.

In addition, the proposed model stands out for its simplicity and interpretability, thanks to the use of normalized indicators (0–100\%) and measurable variables, making it more accessible to plant operators. While other methods may require large volumes of historical data or complex multi-objective optimization procedures, this framework is oriented toward real-time monitoring, adaptability, and efficient implementation over existing infrastructures.}

Comments 2: 2) In Table, the author should add the limits (if there are) of the four indicators
Response 2: Thank you for your feedback. The four sub-indicators are expressed as percentages, which is explicitly stated in the text when naming and describing them. As is typical for this type of representation, their values range from 0 to 100, with 0 representing the worst possible performance and 100 representing the best. This presentation facilitates interpretation of the results, allowing one to quickly identify how far a value is from or close to the optimal scenario.

Comments 3: 3) Except oxygen control sensor, have the authors take into account the pH and the temperature effect?
Response 3: The data available for the study were limited to the parameters indicated in the manuscript, and the proposed model was constructed within this context, including the design of the virtual sensor to estimate the total nitrogen value. In this case, specific pH and temperature measurements were not available, so their potential effects were not incorporated. However, we are aware of the relevance of these variables to biological processes in WWTPs, and therefore, we consider future work to evaluate the influence of these parameters, as well as their possible integration to improve or complement both the calculation of the ONI and the construction of more robust and accurate virtual sensors.

Comments 4: 4) The authors should add a figure of the WWT plant
Response 4: Thank you for your comment. A representative diagram of the lines and treatments of the WWTP case study has been added. The added figure is in the Case Study section and is shown below:

Comments 5: 5) Linear and polynomial regression equations should be added
Response 5: We appreciate all your comments and the time you spent reviewing the article. The linear and polynomial regression equations were included. The linear regression equation is shown on page 12, section 4.2.1. Linear Regression on line 436. The polynomial regression equation was added on page 13, section 4.2.2. Linear Regression on line 447.

Comments 6: 6) A representative figure of a decision tree should be added
Response 6: Again, thank you very much for your comments. A representative diagram of decision trees has been included in the page 13, section 4.2.4 on line 471.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

-