Trajectory Data Preprocessing: Methods and Models

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper presents a review to explore the methods and techniques of trajectory data preprocessing, analyze previous research from different perspectives, and connect them with key theories. Researchers have proposed various data cleaning methods, such as filtering, interpolation, and outlier detection, to address this issue. In addition, this paper also reviews the techniques of trajectory data compression, segmentation, and map matching.

Strengths of this paper: The paper is well-organized and covers most of the key topics related to trajectory data preprocessing published over the last 30 years.

Weaknesses of the paper: The main weaknesses of the paper is the absence of illustrative figures and the unclear formatting of tables, which hinder the reader's ability to fully understand and engage with the content.

The following points should be considered:

1-Due to the limited number of figures, the paper is difficult to follow

2-The authors claim that the paper covers the last 50 years, but the references only span the past 30 years.

3-The authors are encouraged to discuss the limitations or drawbacks of each method presented in the paper, to provide a more balanced and critical evaluation.

4-The authors should clearly present the limitations of current methods and suggest future research directions that could address these limitations.

5-Due to the extensive use of abbreviations and nomenclature, a dedicated table is recommended for clarity and ease of reference.

6-The presentation of the tables should be reorganized to enhance clarity and readability.

Author Response

Comment 1: Limited number of figures

Response: Thank you for this valuable suggestion. We agree that additional figures would enhance the clarity and readability of the paper. To address this, we have added the Figure 1. PRISMA flow diagram as a starting point to illustrate the literature screening process. Should the reviewer feel that further figures (for instance, to elaborate on specific algorithms in Sections 3.2 or 3.4) are necessary for clarity, we need additional time to creating and incorporating such detailed illustrations in the next revision.  

Comment 2: Inconsistency in the claimed time span (50 years vs. 30 years in references)

Response: Thank you for pointing out this inconsistency. We have revised the text to accurately reflect the time span covered by the references. Although our earliest reference (Douglas & Peucker, 1973) dates back over 50 years, the majority of the cited works are from the past three decades. To avoid overstatement, we have: Revised the conclusion to state: "…published over the past several decades" Clarified in the introduction that the review focuses on foundational works from the 1970s onward, with an emphasis on developments in the last 30 years.  

Comment 3: Lack of discussion on limitations or drawbacks of each method

Response: We appreciate this insightful comment. To provide a more balanced and critical evaluation, we have added "Limitations and Challenges" at the end of each methodological subsection.  

Comment 4: Future research directions should be clearly linked to current limitations

Response: Thank you for this suggestion. We have restructured Section 5 (Discussion) to explicitly connect each future research direction with specific limitations identified in earlier sections.  

Comment 5: A nomenclature table is needed due to extensive use of abbreviations

Response: Thank you for pointing this out. To enhance readability, we have ensured that all abbreviations are clearly defined upon their first use in the text. We have also carefully reviewed the manuscript to guarantee consistency in the use of terminology and acronyms throughout.

Comment 6: Tables should be reorganized for better clarity and readability

Response: Thank you for this feedback. We have reformatted all tables to improve structure and readability.

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript makes a solid contribution by reviewing the state of the art in trajectory data preprocessing. The division of methods into data cleaning, compression, segmentation, and map matching is logical and helpful for readers. The inclusion of tables comparing prior surveys and compression techniques adds value.

I have the following suggestions for improvement:

Some of the subsections (e.g., on line simplification algorithms) are lengthy, and overwhelming the reader. Consider summarizing with schematic figures or to make comparative tables rather than extended text descriptions.

While methods are described individually, the paper could better emphasize comparisons between them—e.g., in terms of computational complexity, accuracy, and typical application scenarios (urban transport vs UAV vs robotics).

Please include several more references which are newest, for exaple from 2023–2025, particularly on unsupervised/semantic trajectory preprocessing and privacy-preserving approaches.

Figure 1, on trajectory compression is useful, but the captions could be expanded to make the diagrams self-explanatory. Also, adding an illustrative case study (even a simple synthetic trajectory) would help demonstrate how preprocessing stages work together.

Author Response

Comment 1: Some of the subsections (e.g., on line simplification algorithms) are lengthy.

Response: Thank you for your profound insights. We agree with your suggestion that simplifying the algorithm section is indeed too detailed. In order to improve clarity and conciseness, we have appropriately compressed the text description to enhance the readability of the manuscript.

 

Comment 2: While methods are described individually, the paper could better emphasize comparisons between them.

Response: We sincerely thank the reviewer for this excellent suggestion. To enhance the comparative analysis, we have now integrated a dedicated "Comparative Analysis" sub-section at the end of each major technical section (Sections 3.1-3.4). In these sub-sections, we explicitly discuss the trade-offs between methods regarding computational complexity, robustness to noise, and parameter sensitivity.

 

Comment 3: Please include several more references which are newest.

Response: Thank you for pointing this out. We have thoroughly updated the reference list to include several recent and high-impact publications from 2023-2025. Key additions include works on unsupervised semantic segmentation, privacy-preserving trajectory publishing, and advanced deep learning models for map matching. These new references ensure the review reflects the current state-of-the-art.

 

Comment 4: Figure on trajectory compression is useful, but the captions could be expanded to make the diagrams self-explanatory.

Response: We appreciate this constructive feedback. We have comprehensively expanded the captions for Figure 2 (Trajectory Compression Algorithms) to make them self-contained and self-explanatory. Each sub-figure now clearly describes the algorithmic step being illustrated and its outcome.

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript surveys the core pipeline of trajectory data preprocessing—including data cleansing, compression, segmentation, and map matching—by organizing both classical and emerging methods, consolidating commonly used public datasets, and outlining open challenges. It further highlights future directions such as real-time, large-scale computation, the integration of deep learning for semantic annotation, and privacy preservation. Overall, the study is practically oriented. To enhance the paper’s quality and advance the research, the following suggestions are offered:

1. Although the paper claims to be “the first comprehensive review covering the full technical stack of trajectory preprocessing,” this assertion should be substantiated through a systematic search protocol and quantitative comparisons that demonstrate how it surpasses prior surveys (e.g., distinctions from reviews focused on cleaning, compression, segmentation, and map matching).
2. The current sections present compression, segmentation, and map matching alongside time complexity and error metrics, yet category boundaries remain blurred (e.g., compression coupled with map matching; semantics-driven segmentation). It is recommended to provide an overview diagram that explicitly aligns “method–objective–metric,” and to define mutually exclusive vs. overlapping scopes with boundary cases.
3. While a broad set of representative algorithms is compiled, the paper lacks cross-method evaluation on shared datasets with unified metrics (e.g., geometric/topological/probabilistic/deep map matching under varying sampling rates, noise levels, and road-network densities). Please expand the discussion on comparative methodology and include reproducible quantitative benchmarks.
4. The discussion acknowledges the importance of Hadoop/Spark/Flink and GPU parallelization but remains descriptive. Consider adding a “data-rate–latency–resource” design map; for example, under V2X settings with N records/vehicle/second and M vehicles city-wide, provide achievable latency targets and resource configurations for each preprocessing module (cleaning/compression/matching).
5. The manuscript notes trends toward Transformer/GNN methods and the challenges of scarce labels and explainability. We recommend specifying at least one general input-representation paradigm (e.g., spatiotemporal tokenization, road-graph embeddings, semantic embeddings for speed/turning/stops) together with explainability strategies (attention attribution, prototype learning, counterfactual analysis).

Author Response

Comment 1: Although the paper claims to be “the first comprehensive review covering the full technical stack of trajectory preprocessing,” this assertion should be substantiated through a systematic search protocol and quantitative comparisons that demonstrate how it surpasses prior surveys.

Response: Thank you for this critical observation. To substantiate our claim of being a comprehensive review covering the full technical stack, we have strengthened our methodology in following way:

Enhanced Systematic Protocol: We have expanded Section 2 (Overview of Research Methods) to provide a more detailed account of our systematic literature search, including the exact search strings, inclusion/exclusion criteria application process, and a justification for the chosen databases and time span. This transparency reinforces the systematic nature of our survey.

Quantitative Comparison with Prior Surveys: We have significantly expanded Table 1 into a more detailed comparative matrix. This new table now explicitly lists the specific sub-topics covered by previous surveys (e.g., "Online Cleaning," "Semantic Compression," "Topological Map Matching") and uses checkmarks to visually demonstrate where our review provides coverage that prior works lack, thereby quantitatively showcasing its broader scope.

 

Comment 2: Clarifying Category Boundaries

Response: We agree that the interplay between categories is an important aspect of the field. To address this:

Explicit Discussion of Boundaries: In a new subsection "3.5 Interplay and Boundaries Between Preprocessing Tasks", we now explicitly discuss how methods can overlap and be combined. We use concrete examples such as "road-network constrained compression" (compression + map matching) and "semantic segmentation" (segmentation + semantic tagging) to illustrate these synergies, while also defining the core, distinct focus of each primary category.

 

Comment 3: Incorporating Cross-Method Evaluation and Benchmarks

Response:This is a valuable suggestion for enhancing the practical utility of our review.

Expanded Comparative Discussion: We have added a new subsection "4.1 A Framework for Comparative Evaluation" under the "Public Dataset" section. Here, we propose a standardized benchmarking framework, suggesting specific datasets (e.g., a mix of GeoLife for urban drives and highD for highway data), unified metrics (SED for accuracy, compression ratio, runtime), and critical testing dimensions (varying sampling rates, adding synthetic noise, using road networks of different densities).

Synthesis of Existing Comparative Results: While performing new experiments is beyond the scope of this review, we now synthesize and discuss key findings from existing comparative studies in the field (e.g., Lin et al.'s evaluation of compression algorithms or Huang et al.'s survey on map-matching) within the relevant technical sections. This provides the reader with insights into the relative performance of different method categories under various conditions.

 

Comment 4: Consider adding a “data-rate–latency–resource” design map

Response: Thank you for this constructive suggestion. In response, we have enhanced the discussion on efficient computation in Section 5.1 by moving beyond a purely descriptive account. We now include a more concrete analysis of the performance considerations—specifically the interplay between data rate, latency requirements, and computational resources—for key preprocessing modules in large-scale scenarios. We acknowledge that establishing precise, universal quantitative benchmarks for all configurations is a complex task that depends heavily on specific infrastructure and data characteristics. Therefore, while our current revision provides a crucial foundational discussion and design framework, we recognize the value of developing more detailed, empirical performance models. We aim to pursue this line of inquiry in our future research, focusing on generating specific, reproducible resource configurations and latency targets for defined use cases.

 

Comment 5: Specifying Input Representations and Explainability Strategies

Response: We have deepened the discussion on deep learning as requested.

Specified Input Representation Paradigms: In the "Deep Learning-based Preprocessing" part of Section 5, we now specify concrete input-representation paradigms:

Spatiotemporal Tokenization: Representing a trajectory as a sequence of tokens incorporating (lat, lon, time).

Road-Graph Embeddings: Using pre-trained Graph Neural Networks to encode the underlying road network for map-matching tasks.

Semantic Embedding Tiers: Creating feature vectors that combine raw coordinates with derived features (speed, acceleration, heading change, stop likelihood).

Reviewer 4 Report

Comments and Suggestions for Authors

Here are the following comments.

1. The style of Table 1 seems broken and hard to follows. The first column linebreaks cause more confusion. In addition, the last row includes this paper survey which I have never seen before. Please fix them all

2. The last paragraph of Section 1 is too long. Normally, I quoted this paragraph due to the lack of details. However, in this paper, the details are too much. The authors need to fix that paragraph.

3. It seems that the processing steps were written in the second paragraph of Section 2. It would be the best if authors derive the procedure in the block diagram instead of words.

4. In subsection 2.1, the indentation of enumerate is not correct. Please fix them

5. In subsection 2.1.1, each paragraph represents the process of data cleaning, but cannot follow why the Kalman and particle filters need to be used in this cleaning. Also, no proper reference to back up your statement. Please fix them

6. In terms of subsection 2.1, are the subsections 2.1.1~2.1.4 lined up in order or not? Please clarify it.

7. In subsection 2.2.1., the explanation of line simplification trajectory compression is not well-described in this paper. Especially, the result of figure 1 is hard to follow. Please fix them.

8. In subsection 2.2.2, the description style is APA style. Do authors have to follow that style or are they just mistakes? Please fix them all.

9. Table 2 Style is also broken and no idea what I am reading. Please reorganize it.

10. Similar problem in Subsection 2.2~2.4, the tables and paragraph explanations are pretty lacking and difficult to follow. The authors need to work on fixing them substantially.

11. Suddenly, there is the list of dataset appeared in Section 3 and Section 4 showed some discussion about these paper topics. However, I am not exactly sure what analytical aspects are shown in this paper. Authors need to work a lot for the modification.

12. Same in Section 5, Conclusion. A lot of works need to be done.

Author Response

We sincerely thank the editor and reviewers for their diligent efforts and valuable feedback. We have thoroughly addressed all the points raised in the review. Specifically, we have restructured and reformatted Table 1 and Table 2 to improve clarity and readability, revised the overly long paragraph in Section 1 by breaking it into more digestible parts, and replaced the textual description of processing steps in Section 2 with a clear block diagram. Additionally, we have enhanced the description of line simplification in Subsection 2.2.1 and improved the interpretation of Figure 1, standardized the citation style in Subsection 2.2.2 to ensure consistency, and substantially revised the analytical aspects and presentation in Sections 3 and 4 regarding datasets and discussion. Finally, we have strengthened the conclusion in Section 5 to better summarize the paper’s contributions and future directions. We believe these revisions have significantly improved the manuscript and appreciate the opportunity to enhance our work.  

Round 2

Reviewer 4 Report

Comments and Suggestions for Authors

Still need some fixed but substantially improved.

 

1. Abstract need to be written in a paragraph instead of break-down style.

2. Still issues with section 2 and 3. Lack of organization and description causes too many subsections and missing the details. Please fix them. (less subsection.)

3. Table 2 still looks like inconsistent. Too much empty space and linebreak on the reference. Please fix them. (Table 3 has the same issue but much nicer.)

4. I am pretty sure that list of abbreviation can be done at the end of paper instead of the middle of paper as section 7. Please remove it.

Author Response

Dear Editor and Reviewers,

We are writing to express our sincere gratitude for your thoughtful feedback and valuable suggestions, which have greatly contributed to improving our manuscript. We have carefully considered all the comments provided by the reviewers and have undertaken substantial revisions to address their concerns. In response to the specific points raised, we have provided detailed, point-by-point responses to each comment and have revised the manuscript accordingly. Several new paragraphs and explanations have been added to ensure clarity and depth. All changes made to the manuscript have been highlighted in red for ease of reference.

We hope that our revisions satisfactorily address the issues raised and meet with your approval. Once again, we deeply appreciate the time and effort you have dedicated to reviewing our work. Your insightful comments have been invaluable in strengthening the paper.

Thank you for your guidance and support.

Sincerely,

Peiyu Li and Co-authors

Author Response File: Author Response.pdf

Round 3

Reviewer 4 Report

Comments and Suggestions for Authors

I looked at the manuscript. No major fixation whatsoever. Table 2 and 3 citation line is still broken and both tables have certain space which looks unprofessional and the abstract is not what it is supposed to be.

I am sorry, but I still cannot accept this paper unless the authors are willing to fix.

Author Response

We sincerely thank you for the valuable feedback provided during the peer-review process. We have carefully considered all the comments and have made significant revisions to the manuscript to address them. In particular, we wish to note that we have now fully incorporated the second round of comments, which we regrettably overlooked in our previous revision due to an oversight on our part. We apologize for this error. The current version includes comprehensive fixes to the specific points raised, such as the formatting of Tables 2 and 3 and the restructuring of the abstract. We believe the manuscript has been substantially improved as a result of this rigorous review process. Should you or the reviewers require any further modifications, please do not hesitate to inform us. Thank you for your time and consideration.

Question 1:

Abstract need to be written in a paragraph instead of break-down style.

Answer :

 Thank you for this suggestion regarding the structure of the abstract. We have rewritten the abstract into a single, continuous paragraph as recommended. The modifications are highlighted in red font as follow:

Abstract

Trajectory data from GPS and sensors are increasingly available, necessitating effective preprocessing techniques for data mining. To systematically review methods and models for trajectory data preprocessing. We conducted a systematic literature search in IEEE Xplore, Association for Computing Machinery Digital Library‌(ACM DL), Scopus, Web of Science, Transport Research International Documentation published over the past several decades, using keywords related to trajectory data preprocessing. Studies were screened and selected based on predefined inclusion and exclusion criteria. We included 138 studies, summarizing techniques in data cleaning, compression, segmentation, and map matching. Key algorithms and their performance are compared. This review synthesizes current preprocessing methods and identifies future research directions, including real-time processing, semantic labeling, and privacy protection.

 

Question 2:

Still issues with section 2 and 3. Lack of organization and description causes too many subsections and missing the details. Please fix them. (less subsection.).

Answer :

We sincerely thank you for this critical feedback regarding the organization of Sections 2 and 3. We agree that the previous structure was overly fragmented and lacked a cohesive narrative. In direct response to this comment, we have undertaken a restructuring of the entire methodological core of the paper to address the issues of excessive subsections and missing details. The modifications are highlighted in red font in manuscript.

Question 3:

Table 2 still looks like inconsistent. Too much empty space and linebreak on the reference. Please fix them. (Table 3 has the same issue but much nicer.)

Answer :

Thank you for their meticulous attention to the formatting details of our tables. We have carefully revised Table 2, Table 3 and Table 4 to resolve the inconsistencies in layout. The specific fixes include:

We have adjusted the column widths to significantly reduce the excessive empty white space.

The line breaks in the reference column have been fixed to ensure a cleaner and more compact presentation. The modifications are highlighted in red font as follow:

 

Table 2. Overview of trajectory compression methods

 

Classification		Algorithm	Time Complexity	Error Metric	Article
Compression based on line segment simplification	Offline Compression	Douglas-Peucker		Perpendicular Euclidean Distance	(Douglas and Peucker, 1973)
		Path Hull		Perpendicular Euclidean Distance	(Hershberger and Snoeyink, 1992)
		TD-TR		Time-ratio Distance	(Meratnia and de By, 2004)
		DPTS		Euclidean Distance, Direction, Speed	(Long et al., 2013)
		CTEV		Euclidean distance	(Bashir et al., 2022)
	Online Compression	Sliding Window		Perpendicular Euclidean Distance	(Keogh et al., 2001)
		Opening Window		Time-synchronous Distance	(Meratnia and de By, 2004)
		STTrace		Time-synchronous Distance, Direction, Speed	(Potamias et al., 2006)
		SQUISH		Time-synchronous Distance, Direction, Speed	(Muckell et al., 2011)
		SQUISH-E		Time-synchronous Distance, Direction, Speed	(Muckell et al., 2014)
		BQS		Euclidean Distance	(Liu et al., 2015)
		ROPW		Perpendicular Euclidean Distance	(Li et al., 2021b)
Road network constrained compression		Nonmaterialized		Perpendicular Euclidean Distance	(Cao and Wolfson, 2005)
		Shortest Path		Euclidean Distance	(Lerin et al., 2012)
		MMTC		A weighted average of network distance and time distance	(Kellaris et al., 2009, 2013)
		PRESS		Time Synchronized Network Distance, Network Synchronized Time Difference	(Song et al., 2014)
		COMPRESS		Time Synchronized Network Distance, Network Synchronized Time Difference	(Han et al., 2017)
		CiNCT		Bit-wise rank value	(Koide et al., 2018)
		TrajCompressor		Perpendicular Euclidean Distance	(Chen et al., 2019b)
		VTracer		Perpendicular Euclidean Distance	(Chen et al., 2019a)
		CLEAN		Time Synchronized Network Distance, Network Synchronized Time Difference	(Zhao et al., 2020)
Semantic Compression		STC		Average Spatiotemporal Distance	(Richter et al., 2012; Schmid et al., 2009)
		EHSTC		Perpendicular Euclidean Distance	(Feng et al., 2013)
		STMaker		N/A	(Su et al., 2014)
		SATC		synchronous Euclidean distance	(Ta et al., 2016)
		GR-B		synchronous Euclidean distance	(Reyes Zambrano, 2019)
		STSS		Homomorphic Distance	(Liu et al., 2021)
		ROCE		Point-to-Segment Euclidean Distance	(Yin et al., 2022)

(N represents the number of trajectory points, represents the buffer size, λrepresents target compression ratio, M represents the number of line segments or edges in a trajectory, B represents a bit vector that controls the size of the internal blocks, m represents the number of replacement polylines, L represents the number of layers in CascadeSync model, T represents the time steps in each round)

...

Table 3. An overview of map segmentation methods.

Classification	Method	Time Complexity	Segment Metric	Article
Supervised trajectory segmentation	SPD		Time and distant threshold	(Zheng et al., 2011)
	WS-II		Error threshold	(Etemad et al., 2020)
Unsupervised trajectory segmentation	TRACLUS		Distance	(Lee et al., 2007)
	SMoT		Time	(Alvares et al., 2007)
	CB-SMoT		Speed	(Palma et al., 2008)
	DB-SMoT		Direction	(Rocha et al., 2010)
	Greedy segmentation		Location, heading, speed, velocity, curvature, sinuosity, and curviness.	(Buchin et al., 2011; Buchin et al., 2010)
	SeTraStream		Correlation of features	(Yan et al., 2011)
	Warped K-Means		Criterion function	(Leiva and Vidal, 2013)
	GRASP-UTS		Homogeneity of features	(Soares Júnior et al., 2015)
	OWS		Error signal	(Etemad et al., 2019)
	SWS		Error signal	(Etemad et al., 2021)
	BTCN		Transportation mode	(Markos et al., 2021)
	TS-MF		Similarity of multiple motion features	(Xu and Dong, 2022)
Semi-supervised trajectory segmentation	RGRASP-SemTS		Homogeneity of features	(Junior et al., 2018)
	SECA		Homogeneity of features	(Dabiri et al., 2019)

(N represents the number of trajectory points, C represents candidate stop,  represents the sample vector dimension, m represents the total number of iterations.)

...

 

Table 4. An overview of map matching techniques.

Classification	Method	Time Complexity	Article
Geometric based model	PTP, PTC, CTC		(Bernstein and Kornhauser, 1996)
	PTP, PTC, CTC		(White et al., 2000)
	RRF		(Taylor et al., 2001)
	Global Map-Matching	for Fr´echet dist, for Weak Fr´echet dist	(Brakatsoulas et al., 2005)
Topology based model	MPM		(Alt et al., 2003)
	MM		(Quddus et al., 2003)
	GeoTrackMapper		(Chawathe, 2007)
	ATMM		(Zhao et al., 2018a)
	HFTMM		(Yu et al., 2022)
Probability statistics-based model	MHT-MM	N/A	(Pyo et al., 2001)
	Adaptation MHT-MM	N/A	(Marchal et al., 2005; Schuessler and Axhausen, 2009)
	MDP-MM		(Chen et al., 2014)
	ST-CRF		(Liu et al., 2016)
	MCM		(Li et al., 2023)
	PMHT-MM	N/A	(Wang et al., 2023)
Advanced model	HMM		(Newson and Krumm, 2009)
	OHMM		(Goh et al., 2012)
	FMM		(Yang and Gidofalvi, 2018)
	OM2		(Xie et al., 2020)
	INC-RB		(Luo et al., 2020)
	ST-Matching		(Lou et al., 2009)
	IVMM		(Yuan et al., 2010)
	STP-IWC		(Teng and Wang, 2019)
	AMM		(Hu et al., 2023)
	HRIS		(Zheng et al., 2012)
	HMM+RCM		(Jagadeesh and Srikanthan, 2017)
	DeepMM	N/A	(Zhao et al., 2019)
	TMM	N/A	(Jin et al., 2022)
	L2MM	N/A	(Jiang et al., 2023a)

(N represents the number of GPS points in the trajectory, k represents the average number of candidate points per GPS point, M represents the total number of edges or intersections in the road network, and S represents the number of segments.)

Question 4:

I am pretty sure that list of abbreviation can be done at the end of paper instead of the middle of paper as section 7. Please remove it.

Answer :

Thank you for this suggestion regarding the placement of the list of abbreviations. We have removed Section 7 (the list of abbreviations) from the main body of the paper and have placed it as an appendix at the end of the manuscript, as is standard practice.

Thank you for this helpful suggestion to improve the paper's structure.

Question 5:

I looked at the manuscript. No major fixation whatsoever. Table 2 and 3 citation line is still broken and both tables have certain space which looks unprofessional and the abstract is not what it is supposed to be.

Answer :

We sincerely apologize for this oversight. It appears there was a misunderstanding on our part during the previous revision, and we regrettably missed the specific instructions regarding Table 2, Table 3, and the abstract in your last round of comments. We thank you for your patience and for bringing this to our attention again.

We have now comprehensively addressed all these points in the current version of the manuscript:

We have completely reformatted both tables. The citation lines have been fixed to ensure there are no awkward line breaks. We have also optimized the column widths and overall layout to eliminate the excessive white space, creating a more compact and professional appearance.

The abstract has been rewritten into a single, continuous paragraph as required, removing the previous bullet-point or broken-down style.

We assure you that the manuscript has now been updated to fully incorporate all your valuable feedback. Thank you again for your meticulous review and for helping us improve the quality of our work.

Author Response File: Author Response.pdf

Round 4

Reviewer 4 Report

Comments and Suggestions for Authors

The very last one which is the list of abbreviation is not fixed yet. Please fix it, then it is allowed to be published.

Author Response

Dear Reviewer, Thank you for your comment regarding the placement of the list of abbreviations. I apologize for the previous misunderstanding. Following your feedback, I have now moved the List of Abbreviations to the very end of the manuscript, placing it after the References section. Please let me know if this placement is appropriate or if any further adjustments are needed. I am happy to revise it accordingly. Thank you again for your guidance. Sincerely, Peiyu Li

Author Response File: Author Response.pdf