A Vehicular IoT-Based Methane Sensing System for Large-Scale Urban Environmental Monitoring
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe article is devoted to the development, manufacture and field testing of a methane sensing system for monitoring methane concentrations in the urban environment. The undeniable advantage of the study is the long-term field trials conducted across a large area of ​​the São Paulo metropolitan region. It is worth noting the increased attention of authors to the development of a sampling system based on a dual-chamber architecture using ANSYS Fluent 2021 R2 for analyzing the internal flow distribution. The developed sensing system demonstrated the ability to detect methane emission hotspots no worse than the reference device (Protheo Huberg), while providing a denser spatial representation of the gas in real time. The data obtained can be used to justify the need for a more detailed inspection in local areas of urban infrastructure.
Despite the interest generated by the work done by the authors, there are some comments regarding the design of the article.
- It is not clear from Chapter 2 why only MOS sensor and NDIR sensors are compared. Only towards the end of the article does it become clear that the MOS sensor is being compared to the NDIR sensor because it is used in a Protheo Huberg installed on vehicles.
- Chapter 7 does not specify the material from which the chambers were made. How inert is this material to gases? Does this material cause gas molecules to settle and accumulate on the inner surface of the chambers over time?
- Chapter 9 lacks a more detailed description of the laboratory setup and the numerical values ​​of temperature and humidity during calibration. Please describe the calibration process in more detail.
- What mathematical model was used for calibration? Were temperature and humidity readings taken into account during calibration? Determining the exact methane concentration in urban field tests is impossible without calibrating sensors to account for temperature and humidity fluctuations. If the task was initially reduced to qualitative rather than quantitative monitoring of the urban environment, the calibration described in Chapter 9 did not make sense. In this case, there is no significant need to compare the quantitative parameters of the developed methane sensing system and reference device (given in Tables 1 and 2).
For example, a multivariable linear regression calibration method could be used for calibration, taking into account variations in environmental factors (Ren, X. et al. Effects of Environmental Factors on the Performance of Ground-Based Low-Cost CO2 Sensors. Sensors 2025, 25, 6114).
- To what extent do external factors, such as weather conditions, windy weather, vehicle speed, or exhaust fumes in traffic jams, influence the measurement results of the developed methane sensing system?
- The article provides a sensor warm up phase, which lasts approximately one minute before starting measurements (line 327). Is this period of time sufficient to warm up the sensor? How was this established?
- Installing measuring equipment on a vehicle results in budget savings, since, unlike stationary placement of measuring equipment throughout the city, it does not require the use of a large number of measuring instruments. At the same time, vehicles themselves are active polluters of the environment. What amount of exhaust gases did 16 vehicles emit into the atmosphere during 20 months of operation on 192,274 km of road? Should cars be used as carriers of sensing systems in the future?
- The excessive number of chapters into which the article is divided and the lack of a logical sequence in the presentation of the work done are noteworthy.
In the Introduction (Chapter 1) I recommend combining Chapters 1 and 2.
Chapter 2 can be divided into four subchapters:
Subchapter 2.1 – Chapter 8, describing the elements included in the developed system.
Subchapter 2.2 – Chapters 4 and 6 describe the design of a gas sampling system based on a chamber, pump and valves.
Subchapter 2.3 – Chapters 5 and 7, describing the calculation of dimensions and optimization of the camera geometry.
Subchapter 2.4 – Chapter 3, describing the processing of the signal received during the operation of the monitoring system.
Chapter 3 – Chapter 9, sensor calibration.
Chapter 4 “Results and Discussion” combines Chapters 11, 12, 13.
Chapter 5 – Chapter 14 “Conclusions”.
- The reference numbers do not correspond to the order in which they appear in the text. Please correct this.
Author Response
Comments 1: It is not clear from Chapter 2 why only MOS sensor and NDIR sensors are compared. Only towards the end of the article does it become clear that the MOS sensor is being compared to the NDIR sensor because it is used in a Protheo Huberg installed on vehicles.
Response 1: We sincerely thank the reviewer for this valuable observation. We agree that the rationale for comparing only MOS and NDIR sensing technologies was not sufficiently explained in the original version of the manuscript.
The primary objective of this project was to develop a low-cost, scalable methane monitoring system that could be easily replicated for large-scale urban deployment. During the early stages of the project, several sensing technologies were experimentally evaluated, including catalytic and pyroelectric sensors. However, these alternatives did not provide satisfactory performance for continuous mobile monitoring due to limitations related to long-term stability, durability under harsh environmental conditions, sensitivity, and overall implementation cost.
Among the evaluated technologies, the MOS sensor (Figaro TGS2611-C00) demonstrated the best compromise between cost, robustness, sensitivity, power consumption, and ease of integration into the proposed mobile platform. The NDIR technology was selected as the reference because it is employed in the Protheo Huberg methane detection system installed on the reference monitoring vehicles used during the field validation. Furthermore, the same MOS sensor was previously employed and characterized in our laboratory measurement platform (Enesens GT-7), providing additional confidence in its performance and calibration.
For these reasons, the comparison presented in this work focuses exclusively on MOS and NDIR technologies, as they represent the sensing principles adopted in the proposed system and in the reference commercial system used for validation.
Changes made in the manuscript: A new paragraph has been added to Chapter 2 explaining the rationale for selecting MOS and NDIR sensing technologies, clarifying that multiple sensing principles were initially evaluated and that these two technologies were selected because they demonstrated the best combination of performance, robustness, and practical applicability for mobile methane monitoring. The revised text also explicitly states that the NDIR sensor corresponds to the sensing technology employed by the commercial Protheo Huberg system used as the reference during the field validation.
Comments 2: Chapter 7 does not specify the material from which the chambers were made. How inert is this material to gases? Does this material cause gas molecules to settle and accumulate on the inner surface of the chambers over time?
Response 2: We thank the reviewer for this important comment. The gas chambers were manufactured using ABS (Acrylonitrile Butadiene Styrene) by fused deposition modeling (FDM) 3D printing. This information has now been included in Chapter 7.
ABS was selected because it combines good mechanical strength, dimensional stability, ease of fabrication, and low manufacturing cost, making it suitable for rapid prototyping and field deployment. From a chemical standpoint, ABS is considered sufficiently inert for atmospheric methane monitoring. Methane is a chemically stable, non-polar molecule that exhibits very low affinity for ABS surfaces under ambient operating conditions. Consequently, no significant chemical interaction or irreversible adsorption between methane and the chamber material is expected during normal operation. Although polymeric materials may exhibit very small gas permeability and weak surface adsorption, these effects are negligible for methane at atmospheric concentrations and sampling times of a few seconds. Therefore, no measurable accumulation of methane on the chamber walls was observed during laboratory calibration or extended field measurements, and no progressive baseline drift attributable to gas retention inside the chamber was detected. The periodic air exchange promoted by the active pump further minimizes any possibility of residual gas accumulation between successive measurements.
Changes made in the manuscript: The manuscript has been revised to specify that the sensing chambers were fabricated from ABS using 3D printing. An additional paragraph has been included discussing the chemical compatibility of ABS with methane, explaining that the material is sufficiently inert for the intended application and that the active airflow continuously renews the gas inside the chamber, preventing significant adsorption or accumulation effects during operation.
Comments 3: Chapter 9 lacks a more detailed description of the laboratory setup and the numerical values ​​of temperature and humidity during calibration. Please describe the calibration process in more detail.
Response 3: Thank you for this valuable comment. Chapter 9 has been substantially revised to provide a more comprehensive description of the laboratory calibration procedure.
Changes made in the manuscript: The revised manuscript now includes:
- A more detailed description of the laboratory calibration chamber, including its configuration, gas circulation system, purge procedure, reference analyzer connection, sensor warm-up sequence, and data acquisition process.
- A complete description of the calibration methodology, including methane injection, gas homogenization, baseline establishment, and the acquisition of sensor resistance and reference methane concentration.
- The numerical values of temperature and relative humidity recorded during calibration (41–44 °C and 30–33% RH), together with an explanation of how these environmental parameters were monitored throughout the experiments.
- A detailed description of the environmental compensation procedure implemented to account for the temperature and humidity dependence of the TGS2611-C00 sensor. A new compensation model based on the manufacturer's temperature–humidity characteristics has been introduced (Equation 7), where the measured sensor resistance is normalized using a bilinearly interpolated compensation factor prior to methane concentration estimation.
- An updated description of the nonlinear calibration procedure, which is now explicitly based on the temperature- and humidity-compensated sensor resistance.
- A new flowchart (Figure 10) illustrating the complete methane concentration estimation algorithm implemented in the embedded firmware, including environmental compensation, calibration equation selection, and real-time methane concentration calculation.
We believe these additions significantly improve the description of the calibration procedure and clarify how environmental conditions were incorporated into the sensor calibration and embedded implementation.
Comments 4: What mathematical model was used for calibration? Were temperature and humidity readings taken into account during calibration? Determining the exact methane concentration in urban field tests is impossible without calibrating sensors to account for temperature and humidity fluctuations. If the task was initially reduced to qualitative rather than quantitative monitoring of the urban environment, the calibration described in Chapter 9 did not make sense. In this case, there is no significant need to compare the quantitative parameters of the developed methane sensing system and reference device (given in Tables 1 and 2).
For example, a multivariable linear regression calibration method could be used for calibration, taking into account variations in environmental factors (Ren, X. et al. Effects of Environmental Factors on the Performance of Ground-Based Low-Cost CO2 Sensors. Sensors 2025, 25, 6114).
Response 4: Thank you for this valuable observation and for recommending the work of Ren et al. We agree that the influence of temperature and relative humidity must be considered for quantitative methane measurements using MOS sensors.
To address this concern, Chapter 9 has been substantially revised. The revised manuscript now explicitly describes the mathematical calibration model used in this work. The calibration procedure consists of two sequential stages. First, the measured sensor resistance is compensated for temperature and relative humidity using an environmental compensation model derived from the manufacturer's temperature–humidity characteristic curves. The compensation factor is obtained by bilinear interpolation, and the compensated sensor resistance is calculated according to the newly introduced Equation (7). Second, the compensated sensor resistance is used as the input variable for the nonlinear piecewise rational regression model (Equation (8)), whose coefficients were obtained by nonlinear least-squares fitting using the laboratory calibration dataset.
Therefore, unlike the previous manuscript version, temperature and relative humidity are explicitly incorporated into the calibration algorithm and are also implemented in the embedded firmware used during field measurements. Consequently, every methane concentration reported in the field experiments corresponds to a temperature- and humidity-compensated measurement.
Regarding the suggested multivariable regression approach, we acknowledge that it represents an effective calibration strategy. In the present work, however, we adopted a sequential compensation methodology, in which the environmental effects are first removed from the sensor resistance before applying the nonlinear methane calibration model. This approach provides comparable environmental compensation while significantly reducing the computational complexity required for real-time execution on the embedded platform. We have clarified this methodology in the revised manuscript and believe that it is well suited to the proposed embedded sensing system.
Changes made in the manuscript: Chapter 9 revised.
Comments 5: To what extent do external factors, such as weather conditions, windy weather, vehicle speed, or exhaust fumes in traffic jams, influence the measurement results of the developed methane sensing system?
Response 5: Thank you for this important observation. External environmental and operational factors can influence mobile methane measurements, and these aspects have been considered in both the system design and the revised manuscript.
Regarding temperature and humidity, the proposed sensing platform continuously measures both environmental parameters and applies the compensation model described in Section 9 before estimating methane concentration. Additionally, because excessive humidity may lead to condensation inside the sampling chamber and affect the MOS sensor response, the embedded firmware incorporates a protection mechanism. When high humidity conditions are detected, the sampling solenoid valves are automatically closed and the sensing system enters a protection mode. During this period, the system performs a short air sampling cycle every five minutes to verify whether environmental conditions have returned to acceptable operating levels before automatically resuming normal measurements.
Regarding vehicle speed, its primary influence is not on methane quantification but on the spatial localization of the measurements. Since the sensing chamber, sampling system, and sensor exhibit a finite response time, increasing vehicle speed increases the spatial displacement between the actual methane source and the recorded measurement location. This effect becomes more significant for localized methane plumes with concentrations above approximately 200 ppm. However, the field experiments were conducted under normal urban driving conditions, where the maximum legal speed limit in the city of São Paulo is 50 km/h. Under these operating conditions, the adopted sampling strategy and the 30 s moving-average processing window provide adequate spatial resolution for identifying methane concentration patterns and emission hotspots.
Concerning vehicle exhaust emissions, the revised Discussion section now explicitly addresses the cross-sensitivity characteristics of the TGS2611-C00 sensor. The sensor exhibits negligible sensitivity to carbon dioxide (COâ‚‚), while limited cross-sensitivity may occur in the presence of carbon monoxide (CO) and other reducing gases, which is an inherent characteristic of metal-oxide semiconductor sensing technology. Consequently, poorly maintained vehicles emitting elevated concentrations of reducing gases may occasionally produce localized responses. However, this effect is mitigated by repeated measurements, laboratory calibration, environmental compensation, and spatial analysis of persistent concentration patterns rather than isolated measurements.
Finally, although the State of São Paulo has a vehicle fleet of approximately 35.3 million vehicles, only about 0.4% are powered by natural gas. Therefore, methane emissions originating directly from vehicle exhaust represent only a small fraction of the overall fleet. The proposed monitoring methodology focuses on identifying persistent methane hotspots through repeated spatial observations, which reduces the influence of occasional transient emissions from individual vehicles and enhances the reliability of the generated methane concentration maps.
Changes made in the manuscript: Discussion section revised.
Comments 6: The article provides a sensor warm up phase, which lasts approximately one minute before starting measurements (line 327). Is this period of time sufficient to warm up the sensor? How was this established?
Response: Thank you for this important observation. According to the TGS2611-C00 datasheet, the recommended warm-up period after power-up may range from approximately 2 to 5 minutes, depending on the operating conditions and the required stabilization level. While this recommendation is appropriate for laboratory measurements, such a long warm-up period would significantly reduce the amount of useful data collected at the beginning of each vehicle route in the proposed mobile monitoring application.
To determine a suitable warm-up period for the intended field operation, additional laboratory experiments were performed prior to the calibration procedure. The sensor response was continuously monitored immediately after power-up under controlled methane concentrations, and the stabilization behavior was evaluated. These experiments showed that the sensor output reached a sufficiently stable condition after approximately 60 seconds, providing repeatable measurements compatible with the objectives of the proposed mobile sensing platform.
Based on these experimental observations, a 60 s warm-up period was adopted in the embedded firmware as a compromise between sensor stabilization and operational efficiency. This configuration minimizes the loss of spatial data at the beginning of each monitoring route while maintaining satisfactory measurement repeatability for mobile urban methane monitoring.
Changes made in the manuscript: A clarification of this experimental criterion has been added to the revised manuscript.
Comments 7: Installing measuring equipment on a vehicle results in budget savings, since, unlike stationary placement of measuring equipment throughout the city, it does not require the use of a large number of measuring instruments. At the same time, vehicles themselves are active polluters of the environment. What amount of exhaust gases did 16 vehicles emit into the atmosphere during 20 months of operation on 192,274 km of road? Should cars be used as carriers of sensing systems in the future?
Response 7: Thank you for this insightful observation. We agree that the environmental impact of the carrier platform is an important consideration when designing large-scale mobile sensing systems.
The objective of the present study was to evaluate the feasibility of a mobile methane sensing platform rather than to assess the environmental footprint of the vehicles used during the measurement campaigns. Consequently, the exhaust emissions produced by the participating vehicles were not quantified. The monitoring campaign was conducted using vehicles from different manufacturers and models operating under real urban driving conditions, making a reliable estimation of their combined exhaust emissions beyond the scope of this work.
The decision to use conventional passenger vehicles was primarily motivated by practical considerations. During the development of the project, electric vehicles suitable for long-term instrumentation were not readily available. In addition, vehicle manufacturers commonly impose warranty restrictions on the installation of experimental electronic equipment, making the permanent instrumentation of privately owned or fleet vehicles difficult. Therefore, the sensing modules were deployed using ride-hailing vehicles operating under the Uber platform, vehicles owned by the research team, and test vehicles provided by the local natural gas distribution company (Comgás), which provided an efficient and economically viable solution for collecting extensive urban methane datasets over a wide geographic area.
We agree that the long-term evolution of mobile environmental monitoring should aim to minimize the environmental impact of the sensing platform itself. As low-power sensing technologies continue to advance, future implementations may rely on electric vehicles or even personal mobile devices, such as smartphones and wearable devices, integrating miniaturized gas sensors with the positioning, communication, and cloud connectivity capabilities already available in these platforms. Such developments could enable large-scale distributed environmental monitoring with a substantially reduced environmental footprint.
Changes made in the manuscript: A brief discussion highlighting these practical considerations and future perspectives has been added to the revised manuscript.
Comments 8: The excessive number of chapters into which the article is divided and the lack of a logical sequence in the presentation of the work done are noteworthy.
In the Introduction (Chapter 1) I recommend combining Chapters 1 and 2.
Chapter 2 can be divided into four subchapters:
Subchapter 2.1 – Chapter 8, describing the elements included in the developed system.
Subchapter 2.2 – Chapters 4 and 6 describe the design of a gas sampling system based on a chamber, pump and valves.
Subchapter 2.3 – Chapters 5 and 7, describing the calculation of dimensions and optimization of the camera geometry.
Subchapter 2.4 – Chapter 3, describing the processing of the signal received during the operation of the monitoring system.
Chapter 3 – Chapter 9, sensor calibration.
Chapter 4 “Results and Discussion” combines Chapters 11, 12, 13.
Chapter 5 – Chapter 14 “Conclusions”.
Response 8: Thank you very much for this valuable suggestion and for the detailed proposal regarding the manuscript organization. We sincerely appreciate the time and effort devoted to carefully analyzing the structure of the manuscript and providing such a comprehensive recommendation.
The original organization of the manuscript was primarily developed by a multidisciplinary engineering team during the research and development phase of the project. Since the work was initially documented by engineers directly involved in the laboratory development, prototype implementation, and experimental validation, the manuscript naturally followed the chronological sequence adopted during the engineering design process rather than the conventional structure typically used in scientific journal articles.
We fully agree that the organization proposed by the reviewer would provide a more logical, concise, and reader-friendly presentation of the work. Unfortunately, due to the extensive technical revisions required throughout the manuscript, together with the limited time available before the revision deadline, it was not possible to completely restructure the paper according to the suggested chapter organization without introducing a significant risk of inconsistencies in figure, table, and equation numbering.
Nevertheless, we carefully considered the reviewer's recommendations and implemented several improvements throughout the manuscript, including the addition of new explanatory sections, clarification of the calibration methodology, expansion of the discussion, and improvements to the logical flow of several sections. We recognize that the proposed chapter organization represents a valuable recommendation and intend to adopt a structure closer to the reviewer's suggestion in future revisions and subsequent publications on this research topic.
We sincerely thank the reviewer for this constructive recommendation, which we believe will also be valuable for improving the presentation of our future work.
Changes made in the manuscript: Manuscript revision.
Comments 9: The reference numbers do not correspond to the order in which they appear in the text. Please correct this.
Answer 9: Thank you for pointing out this editorial issue. The reference list has been carefully reviewed, and all citations have been reordered to ensure that the reference numbering now follows the order of their first appearance in the manuscript. In addition, all in-text citations were verified to ensure consistency with the revised reference list
Reviewer 2 Report
Comments and Suggestions for AuthorsThe paper represents a decent engineering effort, but its scientific soundness, quality and novelty are questionable as a number of statements require clarification.
1) Most important and general question is what are the possible sources of methane in the urban area. Usually, the methane emissions are associated with the safety of infrastructure of petroleum gas transportation and conversion to other valuable products, some methane emissions are related to the natural gas plumes, some methane is produced by cattle farms. So, the methane emissions monitoring due to the needs of global climate change mitigation is not related to the urban area. The measurements with reference IR instrument in the given manuscript clearly show it. Provide clear motivation for methane monitoring in urban area.
2) Protheo Huberg device is mentioned as a reference instrument for "gold standard" methane measurement. The more detailed description of the instrument is required - measurement principle and technology (IR-based is not enough as there are a number of IR-based gas analysis techniques and they differ a lot in the accuracy and sensitivity).
3) The Figure 1 is better to transform into a table and give a quantitative comparison of MOS and NDIR technologies. For example, it is not clear how sensitivity is compared.
4) Signal processing topic contains common knowledge on the moving average technique, it should be better illustrated with the use of actual MOS sensor primary measurement data. The absence of primary raw data in the paper is a weak point in general, it should be provided in some extent at least. Otherwise, it is not clear how the methane concentration is estimated.
5) Line 179 - is 1 s enough to make a measurement with MOS sensor? The equation or formula on how the concentration is calculated should be given.
6) Give more clear comparison between chamber designs 1 and 2. Indeed it is not clear should the type 1 chamber be mentioned at all.
7) Fig. 7 is not informative. Provide indications on the parts of the device.
8) The weakest part of the paper is the huge difference of detected methane concentrations by MOS device and reference instrument (table 2). Some considerations on it are given in lines 424-428, however the question of MOS sensor cross-sensitivity has not been raised. The map, which is given on figure 15, clearly indicates the correlation between the estimated methane concentration and proximity to the roads and center of urban air. It is definitely related to the response of the sensors towards other common urban pollutants, related to automotive exhaust, industrial fume etc.
9) So, the considerations in lines 464-469 are not convincing. The proposed prototype is giving more than 30% of falsely elevated methane concentrations.
10) Lines 484-496 are quite misleading as the designed and tested prototype was initially designed in order to provide more dense data on the methane concentration in air compared to reference IR instrument. This is not an achievement to me.
Author Response
Comments 1: Most important and general question is what are the possible sources of methane in the urban area. Usually, the methane emissions are associated with the safety of infrastructure of petroleum gas transportation and conversion to other valuable products, some methane emissions are related to the natural gas plumes, some methane is produced by cattle farms. So, the methane emissions monitoring due to the needs of global climate change mitigation is not related to the urban area. The measurements with reference IR instrument in the given manuscript clearly show it. Provide clear motivation for methane monitoring in urban area.
Response 1: Response: Thank you for this important comment. We agree that methane emissions are traditionally associated with natural gas infrastructure, petroleum production and transportation, landfills, wastewater treatment facilities, agricultural activities, and livestock. However, recent studies have shown that urban environments also contain multiple distributed methane sources that contribute to local atmospheric methane concentrations and are therefore relevant for greenhouse gas monitoring and climate change mitigation.
The motivation for the present work originated from a project aimed at detecting natural gas leaks in the distribution network of the São Paulo metropolitan area, where methane concentrations above approximately 200 ppm are typically associated with leakage events. During the field campaigns, however, repeated measurements revealed several urban regions exhibiting persistent methane concentrations between approximately 30 and 40 ppm, even in locations where no natural gas pipelines were present. Furthermore, approximately 37% of all measurements exceeded the background methane concentration adopted in this study (1.9 ppm, based on NOAA atmospheric reference values), suggesting the presence of additional methane sources within the urban environment.
Although identifying the specific origin of each methane plume is beyond the scope of the present work, the spatial distribution maps indicate that elevated methane concentrations are frequently associated with locations near rivers, open areas with accumulated organic waste, regions potentially affected by sewer infrastructure, and, in some cases, vehicles fueled by natural gas operating under poor maintenance conditions. These observations motivated the development of a mobile sensing platform capable of identifying persistent methane hotspots through repeated spatial measurements.
The primary objective of the proposed system is therefore not to attribute methane emissions to a specific source, but to generate high-resolution spatial methane distribution maps that support subsequent environmental investigations by researchers, utility companies, and public agencies. By identifying regions with consistently elevated methane concentrations, the proposed platform provides a practical screening tool that can assist future studies aimed at source attribution, emission quantification, and mitigation strategies. To clarify this motivation, the Introduction and Discussion sections have been revised accordingly.
Changes made in the manuscript: Paragraph was added to the Introduction to address the reviewer's comment
Comments 2: Protheo Huberg device is mentioned as a reference instrument for "gold standard" methane measurement. The more detailed description of the instrument is required - measurement principle and technology (IR-based is not enough as there are a number of IR-based gas analysis techniques and they differ a lot in the accuracy and sensitivity).
Response 2: Thank you for this valuable comment. We agree that the roles of the reference instruments were not sufficiently explained in the original manuscript.
Two different reference instruments were employed in this study for different purposes. The laboratory calibration procedure described in Section 9 was performed using a Gastec GT40 methane analyzer, which is based on metal-oxide semiconductor (MOS) sensing technology. The Gastec GT40 was used exclusively under controlled laboratory conditions to establish the calibration relationship between sensor resistance and methane concentration.
The field validation experiments, however, were performed using a Protheo Compact mobile methane analyzer (Huberg GmbH, Germany). Unlike the Gastec GT40, the Protheo Compact is a dedicated infrared methane analyzer providing selective methane measurements with a measurement resolution of 1 ppm and a sampling period of 1 s. Because of its high selectivity, continuous acquisition capability, and widespread use for natural gas leak detection, it was adopted as the reference instrument for the field comparison experiments.
Changes made in the manuscript: The revised manuscript has been updated to clearly distinguish the roles of both reference instruments and to provide a more detailed description of the Protheo Compact analyzer and its measurement characteristics.
Comments 3: The Figure 1 is better to transform into a table and give a quantitative comparison of MOS and NDIR technologies. For example, it is not clear how sensitivity is compared.
Response 3: Thank you for this valuable suggestion. We agree that a tabular presentation provides a clearer and more objective comparison between the MOS and NDIR methane sensing technologies than the original radar chart.
Following the reviewer's recommendation, Figure 1 has been replaced by a comparative table summarizing the main characteristics of both sensing technologies. The new table presents a side-by-side comparison of the measurement principle, response time, measurement resolution, selectivity, cross-sensitivity, long-term stability, power consumption, sampling interval, warm-up time, integration complexity, relative cost, and typical application of each technology. In addition, the table clarifies the roles of the two sensing approaches in this work, highlighting that the MOS sensor was selected for the proposed low-cost mobile sensing platform, whereas the infrared analyzer was employed as the reference instrument during the field validation experiments.
We believe that the new table significantly improves the clarity of the comparison and addresses the reviewer's concern regarding the qualitative interpretation of the original radar chart.
Changes made in the manuscript: Figure 1 has been replaced by a comparative table.
Comments 4: Signal processing topic contains common knowledge on the moving average technique, it should be better illustrated with the use of actual MOS sensor primary measurement data. The absence of primary raw data in the paper is a weak point in general, it should be provided in some extent at least. Otherwise, it is not clear how the methane concentration is estimated.
Response 4: Thank you for this valuable comment. We agree that the original manuscript did not sufficiently illustrate the complete signal processing chain using primary MOS sensor data.
To address this issue, the revised manuscript now includes the primary calibration dataset in Table 2, which presents the raw resistance values measured from the MOS sensors, the corresponding temperature and relative humidity readings, the injected methane volumes, the calculated methane concentrations, and the reference concentrations measured by the Gastec GT40 analyzer.
In addition, Figure 9 now illustrates the experimental response of the TGS2611 methane sensors as a function of methane concentration, showing how the raw sensor resistance decreases nonlinearly with increasing methane concentration. The manuscript also explains how the measured resistance is compensated for temperature and humidity effects using Equation (7), and then converted into methane concentration using the nonlinear piecewise calibration model described by Equation (8).
Finally, Figure 10 was added to illustrate the complete methane concentration estimation algorithm implemented in the embedded firmware, including raw resistance acquisition, temperature and humidity compensation, calibration equation selection, and real-time methane concentration estimation.
We believe that these additions clarify how the methane concentration is estimated from the primary MOS sensor measurements and address the reviewer’s concern regarding the absence of raw sensor data and the limited illustration of the signal processing procedure.
Changes made in the manuscript: Chapter 9 - Revised and Expanded.
Comments 5: Line 179 - is 1 s enough to make a measurement with MOS sensor? The equation or formula on how the concentration is calculated should be given.
Response 5: Thank you for this important comment. We agree that the original manuscript did not sufficiently clarify the relationship between the 1 s sampling interval and the dynamic response of the MOS sensor.
The 1 s interval does not represent the response time of the TGS2611 sensor. Instead, it corresponds to the air sampling cycle of the dual-chamber system, during which the sampling pump continuously renews the gas volume inside each measurement chamber. The TGS2611 sensing element remains continuously powered throughout the measurement process and exhibits a typical response time (T90) of approximately 30 s, as specified by the manufacturer.
To obtain reliable methane concentration estimates, the proposed system does not use a single instantaneous resistance measurement. Instead, the embedded firmware continuously acquires the sensor resistance at 1 s intervals and applies a moving-average filter over a 30 s window (15 consecutive measurements). The filtered resistance is subsequently compensated for temperature and relative humidity using Equation (7), and the compensated resistance is converted into methane concentration using the nonlinear piecewise calibration model described by Equation (8).
The manuscript has been revised to clarify the distinction between the air sampling interval, the intrinsic sensor response time, and the methane concentration estimation procedure.
Changes made in the manuscript: The manuscript has been revised to clarify the distinction between the air sampling interval, the intrinsic sensor response time, and the methane concentration estimation procedure.
Comments 6: Give more clear comparison between chamber designs 1 and 2. Indeed it is not clear should the type 1 chamber be mentioned at all.
Response 6: Thank you for this valuable observation. We agree that the distinction between Chamber Designs 1 and 2 was not sufficiently clear in the original manuscript.
The manuscript has been revised to better explain the role of each chamber design. Chamber Design 1 represents the initial analytical design, whose dimensions were determined from the pump flow rate, sampling time, and required chamber volume. This preliminary geometry served as the starting point for the computational fluid dynamics (CFD) analysis.
Based on the CFD results, Chamber Design 2 was developed by optimizing the internal geometry while preserving the same internal volume. The optimized design provides a more uniform airflow distribution, reduces stagnant flow regions, and improves gas exchange within the sensing chamber. All laboratory calibration and field experiments presented in this paper were performed exclusively using Chamber Design 2.
The manuscript has been revised to clarify this development process and to emphasize that Chamber Design 1 is included only to illustrate the engineering methodology that led to the final optimized sensing chamber.
Changes made in the manuscript: The manuscript has been revised.
Comments 7: Fig. 7 is not informative. Provide indications on the parts of the device.
Response 7: Response: Thank you for this valuable suggestion. We agree that the original Figure 7 did not provide sufficient information regarding the main components of the proposed sensing platform.
The figure has been revised by adding labels and callouts identifying the principal hardware components, including Flow pump, solenoid valves, ESP32 embedded controller, power supply input, and air inlet/outlet connections. These additions improve the readability of the figure and provide a clearer understanding of the sensing platform architecture.
Changes made in the manuscript: Figure 7 revised.
Comments 8: The weakest part of the paper is the huge difference of detected methane concentrations by MOS device and reference instrument (table 2). Some considerations on it are given in lines 424-428, however the question of MOS sensor cross-sensitivity has not been raised. The map, which is given on figure 15, clearly indicates the correlation between the estimated methane concentration and proximity to the roads and center of urban air. It is definitely related to the response of the sensors towards other common urban pollutants, related to automotive exhaust, industrial fume etc.
Response 8: Thank you for this important and insightful comment. We agree that cross-sensitivity is an inherent limitation of metal-oxide semiconductor (MOS) sensors and should be explicitly discussed.
The revised manuscript has been expanded to address this issue in the Discussion section. The TGS2611-C00 sensor exhibits limited cross-sensitivity to certain reducing gases, such as carbon monoxide (CO), hydrogen (Hâ‚‚), alcohol vapors, and hydrocarbons, which is an intrinsic characteristic of MOS sensing technology. Conversely, the sensor exhibits negligible sensitivity to carbon dioxide (COâ‚‚), which is one of the major components of vehicle exhaust gases.
The objective of the proposed sensing platform is not to perform chemical source attribution, but rather to identify persistent spatial methane hotspots through repeated mobile measurements. To reduce the influence of environmental and operational factors, the sensing platform incorporates laboratory calibration, temperature and humidity compensation, and repeated measurements collected over a 20-month monitoring campaign covering more than 48 million measurements.
Furthermore, the field results were compared with an independent infrared methane analyzer (Protheo Compact), which is selective to methane. Although differences between instantaneous measurements are expected because of the different sensing principles and sampling methodologies, both systems consistently identified similar spatial concentration patterns and methane hotspot regions.
We acknowledge that localized interference from other reducing gases may contribute to individual measurements in complex urban environments. This limitation has been explicitly discussed in the revised manuscript and represents an important direction for future work, including the integration of complementary gas sensors to improve source discrimination.
Changes made in the manuscript: The manuscript Discussion has been revised.
Comments 9: So, the considerations in lines 464-469 are not convincing. The proposed prototype is giving more than 30% of falsely elevated methane concentrations.
Response 9: Thank you for this important observation. We respectfully believe that the percentage of measurements above the adopted atmospheric background concentration should not be interpreted as false positive methane detections.
In this work, the threshold of 1.9 ppm was adopted as the atmospheric background methane concentration based on NOAA reference values. Consequently, measurements above this level simply indicate locations where methane concentrations exceeded the expected atmospheric background and do not imply erroneous sensor responses.
Furthermore, the observed elevated concentrations were not isolated events. Similar methane hotspot locations were repeatedly detected by multiple vehicles during a 20-month monitoring campaign comprising more than 48 million measurements. The persistence of these spatial patterns, together with the comparison against the Protheo Compact infrared methane analyzer, suggests that the detected concentration anomalies cannot be explained solely by random sensor cross-sensitivity or transient interference.
We fully acknowledge that MOS sensors exhibit limited cross-sensitivity to certain reducing gases, and this limitation has been explicitly discussed in the revised manuscript. For this reason, the proposed platform is not intended to perform chemical source attribution or replace high-precision infrared analyzers. Instead, its objective is to identify persistent methane hotspot regions through repeated large-scale mobile measurements. Subsequent investigation of these hotspots can then be performed using more selective analytical instruments. To avoid possible misinterpretation, the Discussion section has been revised to emphasize this distinction.
Changes made in the manuscript: Discussion section has been revised
Comments 10: Lines 484-496 are quite misleading as the designed and tested prototype was initially designed in order to provide more dense data on the methane concentration in air compared to reference IR instrument. This is not an achievement to me.
Response 10: Thank you for this valuable comment. We agree that the higher spatial sampling density is an inherent characteristic of the proposed sensing architecture and should not be presented as a scientific achievement by itself.
The intention of this discussion was to highlight an operational advantage of the proposed low-cost mobile sensing platform rather than to claim a superior sensing performance compared with the reference infrared analyzer. The main contribution of this work is the demonstration that a low-cost MOS-based sensing platform is capable of reproducing the principal spatial methane concentration patterns identified by a commercial infrared reference analyzer while enabling scalable and cost-effective deployment over extensive urban areas.
To avoid this possible misinterpretation, the Discussion section has been revised to emphasize that the higher measurement density is a consequence of the proposed system architecture, whereas the scientific contribution lies in the ability of the proposed platform to identify persistent methane hotspot regions and reproduce the principal spatial concentration patterns observed by the reference instrument.
Changes made in the manuscript: Discussion section has been revised.
Reviewer 3 Report
Comments and Suggestions for AuthorsThe presented work is well written and presents applicative research with real-field results that are often time consuming and difficult to highlight in scientific publications.
The presented works are solid, the results interesting. The reviewer thinks that the following elements coud improve the quality of the paper.
- Figure 1 is presenting comparative results but cross sensitivity presented as a key parameter to check is missing. It should be added.
- l129 : the intrinsic response is of 30 seconds and then measurements are done to consolidate the date : 15 samples. We must wait for the line 178 to understand from where the 30s indicated l133 comes from. The reviewer suggests that here the calculation should be detailed : 30 seconds of intrinsic response + 15 samples at 2 second of delay (cf. section 5) but that means 1 minute of measurement or 15 samples taken during the 30 seconds of transient response but that lacks coherence. Why taking measurement during the transient response?
- The analog Front End (l246) should be detailed somewhere . What are the components, what are the performances?
- l281 : the reviewer disagrees with the assumption using a smartphone can be more convenient and faster in implementation for authors but is not a solution that reduces cost, power consumption and size. Dedicated electronic circuits can be used, more compact and less power consuming.
- l352 : the Protheo Hubert unit should be presented specifically in a dedicated subsection to understand what is compared to what. We must wait for the line 476 to only have a reference.
- Table 2 l237 should be more explained : Protheo Huberg seems to be unable to detect strong concentration. Why? Is it a limitation of Proteo or is the presented work missing some calibration?
Author Response
Comments 1: Figure 1 is presenting comparative results but cross sensitivity presented as a key parameter to check is missing. It should be added.
Response 1: Thank you for this valuable suggestion. We agree that cross-sensitivity is an important parameter when comparing methane sensing technologies. Following the reviewer's recommendation, the original Figure 1 was replaced with a comparative table summarizing the main characteristics of the MOS and NDIR sensing technologies. The revised table now includes cross-sensitivity as an independent comparison parameter, together with selectivity, measurement principle, response time, measurement resolution, long-term stability, power consumption, and other relevant characteristics. This addition provides a more comprehensive comparison between the two sensing technologies and better highlights one of the main limitations of MOS-based methane sensors.
Changes made in the manuscript: Change figure 1 for table 1.
Comments 2: l129 : the intrinsic response is of 30 seconds and then measurements are done to consolidate the date : 15 samples. We must wait for the line 178 to understand from where the 30s indicated l133 comes from. The reviewer suggests that here the calculation should be detailed : 30 seconds of intrinsic response + 15 samples at 2 second of delay (cf. section 5) but that means 1 minute of measurement or 15 samples taken during the 30 seconds of transient response but that lacks coherence. Why taking measurement during the transient response?
Response 2: Thank you for this important observation. We agree that the original text could lead to confusion regarding the relationship between the intrinsic response time of the MOS sensor and the 15-sample moving-average window.
The 30 s window does not correspond to an additional stabilization period after the intrinsic sensor response. Instead, it corresponds to the total duration of the moving-average window used during continuous operation. Since the sampling cycle produces one measurement every 2 s, a window of 15 consecutive samples corresponds to:
This value was selected to match the typical response time of the TGS2611 sensor, which is approximately 30 s. Therefore, the system does not wait 30 s and then acquire an additional 15 samples. Rather, it continuously acquires measurements while the sensing element remains powered and exposed to the sampled gas, and the moving-average filter provides a smoothed concentration estimate over a time window consistent with the sensor dynamics.
Changes made in the manuscript: We have revised the manuscript to clarify this point earlier in the text and to avoid the interpretation that the total measurement time is 60 s
Comments 3: The analog Front End (l246) should be detailed somewhere . What are the components, what are the performances?
Response 3: Thank you for this valuable comment. We agree that the description of the analog front-end was insufficient in the original manuscript.
To address this point, Section 8 (Electronic System) has been substantially expanded to provide a detailed description of the analog front-end (AFE) developed for the TGS2611-C00 sensor. The revised manuscript now explains the operating principle of the measurement electronics, including the heater driver, programmable constant-current excitation circuit, voltage-conditioning stage, current-conditioning stage, and analog filtering before analog-to-digital conversion.
In addition, the mathematical formulation used to determine the sensor resistance (Rs) has been included. The manuscript now shows that the sensing resistance is calculated from the simultaneously measured sensor voltage and excitation current according to Ohm's law, while accounting for the gains of the analog conditioning stages. This measured resistance is subsequently compensated for temperature and relative humidity using the calibration procedure described in Section 9, and finally converted into methane concentration through the experimentally obtained calibration model.
These additions clarify the complete measurement chain, from the electrical response of the MOS sensing element to the final methane concentration estimate, making the operation of the proposed sensing system considerably more transparent.
We believe that these revisions fully address the reviewer's concern regarding the description and performance of the analog front-end.
Changes made in the manuscript: Section 8 has been substantially expanded to provide a detailed description of the analog front-end.
Comments 4: l281 : the reviewer disagrees with the assumption using a smartphone can be more convenient and faster in implementation for authors but is not a solution that reduces cost, power consumption and size. Dedicated electronic circuits can be used, more compact and less power consuming.
Response 4: Thank you for this valuable comment. We agree that a dedicated embedded communication platform could provide a more compact and energy-efficient solution than a commercial smartphone.
The objective of using a smartphone in the present work was not to minimize the size, power consumption, or hardware cost of the sensing module itself. Instead, the smartphone was adopted as a readily available communication gateway integrating Bluetooth, GNSS positioning, 4G/5G connectivity, local data storage, user interface, and cloud communication into a single commercial device.
This architecture significantly simplified the prototype development by eliminating the need for dedicated cellular communication hardware, GNSS receivers, display interfaces, operating system development, and cloud communication software, thereby reducing development complexity and accelerating field deployment.
We agree that, for a commercial implementation, these functions could be integrated into a dedicated embedded platform, resulting in lower power consumption, smaller physical dimensions, and a more optimized hardware architecture. This aspect has been clarified in the revised manuscript.
Changes made in the manuscript: Smartphone aspect has been clarified in the revised manuscript.
Comments 5: l352 : the Protheo Hubert unit should be presented specifically in a dedicated subsection to understand what is compared to what. We must wait for the line 476 to only have a reference.
Response 5: Thank you for this valuable suggestion. We agree that the reference instrument should be introduced before being used in the comparison methodology.
To improve the organization of the manuscript, a new subsection entitled "Reference Methane Analyzer" has been added to the system description. This subsection presents the Protheo Compact analyzer, including its operating principle, methane selectivity, measurement resolution (1 ppm), sampling interval (1 s), and its role as the reference instrument employed during the field validation experiments.
The laboratory reference instrument (Gastec GT40) and the field reference instrument (Protheo Compact) are now clearly distinguished. The Gastec GT40 was used exclusively for laboratory calibration of the proposed sensing platform, whereas the Protheo Compact was employed only during the urban field campaigns for comparison of the spatial methane concentration patterns. The subsequent sections now refer the reader to this new subsection, improving the logical flow of the manuscript.
Changes made in the manuscript: New subsection entitled "Reference Methane Analyzer" has been added.
Comments 6: Table 2 l237 should be more explained : Protheo Huberg seems to be unable to detect strong concentration. Why? Is it a limitation of Proteo or is the presented work missing some calibration?
Response 6: Thank you for this important observation. We would like to clarify that the Protheo Compact analyzer is not unable to detect high methane concentrations. The differences observed in Table 5 should not be interpreted as a limitation of the reference instrument nor as evidence of insufficient calibration of the proposed sensing platform.
The two systems employ fundamentally different sensing principles and sampling methodologies. The Protheo Compact is a commercial infrared methane analyzer specifically designed for natural gas leak detection, whereas the proposed system is based on a metal-oxide semiconductor sensor combined with a dual-chamber sampling architecture and environmental compensation.
Although both systems were installed on the same vehicle, they operate with different sampling geometries, different gas transport dynamics, different intrinsic sensor response times, and independent acquisition chains. Consequently, narrow methane plumes may be sampled at slightly different positions and times, leading to differences in the instantaneous peak concentrations reported by each instrument.
For this reason, the objective of the comparison presented in this work was not to demonstrate point-by-point agreement between individual measurements but rather to evaluate whether both systems identify similar spatial methane concentration patterns and hotspot regions. As discussed in the revised manuscript, both instruments consistently identified the same areas of elevated methane concentration despite differences in the reported peak values.
Changes made in the manuscript: o avoid possible misinterpretation, the Discussion section has been expanded to clarify these aspects.
Reviewer 4 Report
Comments and Suggestions for AuthorsDear Authors,
Congratulations for this useful paper. I recommend that it be published after the below issues are addressed:
1) By far the most important issue is the very limited agreement between the TGS2611 measurements and the Protheo Huberg measurements shown in Table 2. These significant differences need to be understood much better than they are at present: why do the TGS2611 methane concentration values appear to be systematically significantly higher than the Protheo Huberg methane concentration values?? And which of the two is correct?? Are the TGS2611C00 values tending to be misled / biased high by concentrations of other gases, such as ethanol vapors, and/or gasoline vapors?? If so, would a fairly simple replacement of the TGS2611C00 sensors with TGS2611E00 sensors help ameliorate such effects? And if that is not the cause, what is it? Some testing of the two detection systems in a garage together with a reference-grade sensor as a "tiebreaker," with generated and reasonably known concentrations of methane in air, and other hydrocarbons (and/or hydrogen) in air, in the testing garage would be useful. One could then see if the quite simple replacement of TGS2611C00 sensors with TGS2611E00 sensors in the detection system would be beneficial or not. And especially if not, if simultaneous measurement of temperature and humidity, and correction for such factors, in the TGS2611-based detection system could help obtain more reliable absolute measurements (without adding a significant amount to the ultimate cost of a TGS2611-based system)?
More minor comments follow:
a) In Figure 4: Gas Chamber Simulation, a sentence should be added to the caption explaining what the difference is between the upper and lower simulated chambers within the figure. It took me a few seconds to realize that the upper simulated chamber represents when the output port valve is closed, and the lower simulated chamber represents when the output port valve is opened; those few seconds to realize this could have been better spent on other things.
b) In addition to the overall map in Figure 15, separate methane concentration maps from TGS2611-based measurements and from Protheo Huberg-based measurements should be provided (as an additional aide for readers toward seeing the agreement (or lack thereof) between the two systems).
c) Are there any ideas about what might be causing the collection of a few red dots near Taboao da Serra in Figure 15? Just a sentence or two (especially for readers like me who are not natives of Sao Paulo and have no clue what possibly might be causing such red dots there) within the text on this would be very useful.
Thanks very much again to the authors for their excellent work so far. I hope the above rather minor issues and improvements can be made, and the article published.
Author Response
Comments 1: By far the most important issue is the very limited agreement between the TGS2611 measurements and the Protheo Huberg measurements shown in Table 2. These significant differences need to be understood much better than they are at present: why do the TGS2611 methane concentration values appear to be systematically significantly higher than the Protheo Huberg methane concentration values?? And which of the two is correct?? Are the TGS2611C00 values tending to be misled / biased high by concentrations of other gases, such as ethanol vapors, and/or gasoline vapors?? If so, would a fairly simple replacement of the TGS2611C00 sensors with TGS2611E00 sensors help ameliorate such effects? And if that is not the cause, what is it? Some testing of the two detection systems in a garage together with a reference-grade sensor as a "tiebreaker," with generated and reasonably known concentrations of methane in air, and other hydrocarbons (and/or hydrogen) in air, in the testing garage would be useful. One could then see if the quite simple replacement of TGS2611C00 sensors with TGS2611E00 sensors in the detection system would be beneficial or not. And especially if not, if simultaneous measurement of temperature and humidity, and correction for such factors, in the TGS2611-based detection system could help obtain more reliable absolute measurements (without adding a significant amount to the ultimate cost of a TGS2611-based system)?
Response 1 Thank you for this important and detailed comment. We agree that the limited point-by-point agreement between the TGS2611-based platform and the Protheo Huberg infrared analyzer requires clearer discussion.
The differences observed between the two systems should not be interpreted as indicating that one instrument is simply “correct” and the other is “incorrect”. The Protheo Huberg analyzer is a methane-selective infrared instrument and therefore provides higher gas selectivity for instantaneous methane measurements. In contrast, the proposed platform is based on the TGS2611-C00 MOS sensor, whose response may be affected by environmental conditions, sensor dynamics, drift, and cross-sensitivity to other reducing gases, including hydrocarbons, hydrogen, alcohol vapors, and exhaust-related compounds. Therefore, the TGS2611-C00 measurements may be biased high under specific urban conditions where interfering reducing gases are present.
To address this issue, the revised manuscript now explicitly discusses MOS cross-sensitivity as a limitation of the proposed system. The Discussion section has been expanded to clarify that the TGS2611-C00 sensor is not intended to provide the same metrological selectivity as an infrared methane analyzer. Instead, the objective of the proposed platform is large-scale methane hotspot screening through repeated spatial measurements. The manuscript now emphasizes that the field comparison with the Protheo Huberg analyzer was intended to evaluate spatial consistency of methane concentration patterns rather than to establish a direct metrological calibration between individual measurements.
In addition, the calibration section has been substantially revised. The revised manuscript now includes temperature and relative humidity compensation based on the manufacturer’s environmental response curves, using a bilinearly interpolated compensation factor. The compensated sensor resistance is then converted into methane concentration using a nonlinear piecewise calibration model. This compensation procedure is implemented in the embedded firmware and is applied before every methane concentration estimate.
We agree with the reviewer that additional controlled experiments involving the proposed system, the Protheo Huberg analyzer, and an independent reference-grade methane analyzer would be valuable to further quantify the contribution of methane, interfering gases, temperature, and humidity to the observed differences. Such experiments, including comparative tests with TGS2611-C00 and TGS2611-E00 sensors under controlled methane and interfering gas concentrations, are highly relevant and have been identified as future work. The possible use of the TGS2611-E00 sensor may reduce sensitivity to certain interfering gases and will be evaluated in future versions of the sensing platform.
The revised manuscript has therefore been updated to avoid overstating the quantitative agreement between the MOS-based platform and the infrared reference analyzer. The conclusions now emphasize that the proposed system is suitable for scalable hotspot screening and spatial pattern identification, while high-selectivity infrared instruments remain necessary for confirmatory measurements and precise source attribution.
Changes made in the manuscript: Manuscript has therefore been updated
Comments : a) In Figure 4: Gas Chamber Simulation, a sentence should be added to the caption explaining what the difference is between the upper and lower simulated chambers within the figure. It took me a few seconds to realize that the upper simulated chamber represents when the output port valve is closed, and the lower simulated chamber represents when the output port valve is opened; those few seconds to realize this could have been better spent on other things.
Response (a): Thank you for this helpful suggestion. We agree that the original caption did not clearly distinguish the two operating conditions illustrated in Figure 4.
The figure has been revised by explicitly identifying the two CFD simulations as (a) chamber filling stage with the outlet valve closed and (b) chamber purge stage with the outlet valve open. In addition, the figure caption has been expanded to explain the operating condition represented in each simulation and to clarify that the color scale corresponds to the simulated airflow velocity inside the sensing chamber. These modifications improve the readability of the figure and make the chamber operation immediately understandable.
Changes made in the manuscript: The figure has been revised.
Comments : b) In addition to the overall map in Figure 15, separate methane concentration maps from TGS2611-based measurements and from Protheo Huberg-based measurements should be provided (as an additional aide for readers toward seeing the agreement (or lack thereof) between the two systems).
Response: (b) Thank you for this valuable suggestion. We agree that separate methane concentration maps could facilitate the visual comparison between the proposed sensing platform and the Protheo Huberg reference system.
However, the Protheo Huberg measurements were not acquired over the entire monitoring area. Unlike the proposed sensing platform, which was deployed on multiple vehicles throughout the monitoring campaign, the Protheo Huberg analyzer was available only during specific field campaigns and therefore covered only selected road segments.
Consequently, it is not possible to generate a complete methane concentration map for the Protheo Huberg system that is directly comparable with the full-area map obtained from the proposed platform. Presenting such a map could inadvertently suggest that both systems surveyed the same geographical area, which is not the case.
To address this issue, the revised manuscript now includes Figures 15 and 16, which present detailed comparisons restricted to the regions where measurements from both systems are available. These figures provide a direct visual comparison of the methane concentration distributions obtained by the two sensing systems over the common surveyed road segments and therefore represent the most appropriate basis for comparison.
Changes made in the manuscript: Clarified this limitation in the manuscript to emphasize that the comparison is restricted to the overlapping measurement regions.
Comments : c) Are there any ideas about what might be causing the collection of a few red dots near Taboao da Serra in Figure 15? Just a sentence or two (especially for readers like me who are not natives of Sao Paulo and have no clue what possibly might be causing such red dots there) within the text on this would be very useful.
Response: (c) Thank you for this helpful suggestion. We agree that readers who are unfamiliar with the São Paulo metropolitan area may benefit from additional context regarding the methane hotspot observed near Taboão da Serra.
To address this point, the Discussion section has been expanded to include a brief description of this region. The revised manuscript now explains that Taboão da Serra is a densely urbanized area characterized by intense vehicular traffic, aging sewer infrastructure, canalized streams, and other potential urban methane sources. We also clarify that the objective of the present work was not to identify the specific origin of each methane plume. Instead, the repeated detection of elevated methane concentrations by different vehicles over multiple surveys suggests the presence of persistent emission sources, making the region an appropriate target for future investigations using high-selectivity analytical instruments.
Changes made in the manuscript: Discussion section has been expanded.
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsI thank the authors of the article for the great work they have done. I have no further comments.
Author Response
We sincerely thank the reviewer for the careful evaluation of our manuscript and for the constructive comments provided throughout the review process. We greatly appreciate the reviewer's positive assessment and are pleased that the revisions have satisfactorily addressed all concerns. Your valuable feedback has significantly contributed to improving the quality and clarity of the manuscript.
Reviewer 2 Report
Comments and Suggestions for AuthorsThe authors have addressed most of the questions, which were raised during the review of the first version of the manuscript. Now the objective of the research and the discussion of the results are more consistent. However, I would like to requires one more correction to the paper, related to the Comments 4, which was made during the previous round of review: "Signal processing topic contains common knowledge on the moving average technique, it should be better illustrated with the use of actual MOS sensor primary measurement data. The absence of primary raw data in the paper is a weak point in general, it should be provided in some extent at least. Otherwise, it is not clear how the methane concentration is estimated." Now the signal calculation is clear but the manuscript still is lacking a piece of raw data. Including a fragment of transient response of the gas sensor with the duration of 5-10 minutes will make the matter much better understandable by the reader. If the raw data with 1 second sampling time is unavailable then provide at least a calculated signal transient which, according to the manuscript, should also have 1 second sampling rate.
Author Response
Comments: The authors have addressed most of the questions, which were raised during the review of the first version of the manuscript. Now the objective of the research and the discussion of the results are more consistent. However, I would like to requires one more correction to the paper, related to the Comments 4, which was made during the previous round of review: "Signal processing topic contains common knowledge on the moving average technique, it should be better illustrated with the use of actual MOS sensor primary measurement data. The absence of primary raw data in the paper is a weak point in general, it should be provided in some extent at least. Otherwise, it is not clear how the methane concentration is estimated." Now the signal calculation is clear but the manuscript still is lacking a piece of raw data. Including a fragment of transient response of the gas sensor with the duration of 5-10 minutes will make the matter much better understandable by the reader. If the raw data with 1 second sampling time is unavailable then provide at least a calculated signal transient which, according to the manuscript, should also have 1 second sampling rate.
Response: Thank you for this valuable suggestion. We agree that the inclusion of representative primary sensor data significantly improves the understanding of the proposed signal processing methodology.
Following the reviewer's recommendation, a new figure has been added to the revised manuscript showing the experimental transient response of the TGS2611-C00 sensor obtained during the laboratory calibration procedure after the injection of a known methane concentration (9 ppm). The figure presents the raw sensor resistance together with the methane concentration simultaneously measured by the Gastec GT40 reference analyzer as a function of time. The experimental data are presented before the application of temperature–humidity compensation, moving-average filtering, or nonlinear calibration. The markers correspond to the original measured data, while the smooth curves were included only to improve the visualization of the transient response.
Together with the calibration dataset presented in Table 2, the temperature–humidity compensation model (Equation (5)), the nonlinear calibration model (Equation (6)), and the methane concentration estimation flowchart shown in Figure 11, this additional experimental result provides a complete description of how the primary MOS sensor measurements are converted into methane concentration estimates.
We believe that the inclusion of these representative primary experimental data substantially improves the clarity and reproducibility of the proposed signal processing methodology and fully addresses the reviewer's concern regarding the absence of representative raw sensor measurements.
Changes made in the manuscript: A new figure has been added presenting the representative transient response of the TGS2611-C00 sensor during laboratory calibration, together with the corresponding explanatory text describing the primary sensor measurements and their role in the methane concentration estimation process. As a result of the insertion of this new figure, all subsequent figures and their corresponding references in the manuscript have been renumbered accordingly.
