Bibliometric analysis of traffic-related air pollution: using CiteSpace to explore the knowledge structure and trends

Although traffic-related air pollution (TRAP) has been a long-standing problem, few bibliometric- and visual analysis-based literature reviews have been performed. In light of this issue, future research plans and directions in the field of TRAP must be determined. Therefore, this study performed a bibliometric analysis of the TRAP publishing trends, including the countries, institutional collaborations, author collaborations, keywords, and hotspots. The information visualization software CiteSpace was used to analyze the relevant literature collected from the Web of Science (WoS) from 2003 to 2022. The main findings of this study included the following: (1) the main keywords in TRAP research are particulate matter, exposure, health, nitrogen dioxide, and mortality; (2) current research is focused on the impacts of TRAP on humans; and (3) potential hotspots for future TRAP research are source apportionment, asthma, heart rate variability, and mobile monitoring. This article aims to develop a better understanding of current research trends in TRAP and provide directions for future research.


Introduction
Traffic-related air pollution (TRAP) has received increasing attention in various countries worldwide (Bai et al 2022).For example, the impact of TRAP on city and public health has been investigated in Austria, France, and Switzerland (Kunzli et al 2000), and the development of urban air standards has been promoted by the United States Environmental Protection Agency (USEPA) and World Health Organization (WHO) (Han and Naeher 2006).Developing methods of reducing the impact of traffic emissions on urban air pollution and controlling future traffic emissions is a common concern for all members of society, including car manufacturers and city officials (Winkler et al 2018).
Moreover, rapid urbanization has increased the demand within the automotive industry (Yin et al 2018).TRAP is mainly generated by the aviation and automotive industries, with aviation activities emitting large amounts of carbon dioxide and carbon monoxide (Zhang et al 2022) and rapid increases in the number of automobiles in operation representing a serious cause of air pollution (Zhang et al 2014b).Bai et al (2022) noted that air issues related to urban traffic have received attention from many groups and suggested that sustainable fuels, post-processing technologies, and new energy vehicles may represent future development directions (Bai et al 2022).Researchers have continued to optimize and reduce the emissions of fueled vehicles (Zhang et al 2014a;Zhang et al 2021).In recent years, China has been vigorously developing new energy vehicles, and the electrification of cars in China has led to climate and health benefits (Liang et al 2019).
Similarly, many academic scholars have begun to study other factors that can improve urban air quality related to traffic and conducted cross-disciplinary studies on various aspects of TRAP.For example, Shorshani et al (2015) reviewed the latest technologies for coupling traffic with atmospheric diffusion; Pan et al (2016) proposed a framework for estimating air pollution related to urban traffic; and Liang and Gong (2020) explored the relationship between traffic congestion and pollution.Medical scholars have also summarized the negative effects of TRAP on the nervous system (Costa et al 2017) and made outstanding contributions to TRAP research, thereby promoting the development of the traffic industry.

Data collection
The data were obtained from the WoS Core Collection and included Science Citation Index Expanded (SCI-EXPANDED) data but excluded other databases, such as the Conference Proceedings Citation Index-Science (CPCI-S) database.This practice is currently recognized by bibliometrists, and Ho (2020) revealed that data obtained in this manner were highly reliable.The accurate use of the WoS Core Collection is crucial (Ho 2018).The WoS version used in this study was WoS 2022; thus, some filter names may have been updated.
The data used were collected on September 4, 2023.The following search criteria were implemented.(1) TS = "traffic air pollution" OR TS = "traffic air quality" OR TS = "traffic-related air pollution" OR TS = "traffic-related air quality" OR TS = "transport * air quality" OR TS = "transport * air pollution".TS is the abbreviation for topic in the WoS 2022, and it includes the title, abstract, author keywords, and KeyWords Plus.The wildcard asterisk ( * ) can indicate any group of characters (Busygina and Rykova 2020); for example, transport * would return results on transport, transporting, transported, and transportation.The use of quotation marks and brackets in WoS can result in completely different situations.Quotation marks are used to find phrases (Busygina and Rykova 2020), and the word AND represents the Boolean relationship between words in the same quotation mark (i.e., 'traffic air pollution').However, the word OR represents the Boolean relationship between words in the same bracket.In this experiment, the relationship between traffic and air pollution must be emphasized; thus, quotation marks must be used to determine the AND Boolean relationship.The reason for this setting is that although some authors have not used the standard expression of TRAP, their articles are essentially consistent with the TRAP topic.The search criteria also include (2) 2003 to 2022 as the time period, with a total interval of 20 years; and (3) articles as the only file type.Other types of articles, such as proceedings papers, meeting abstracts, reviews, and letters, were not considered.Fu and Ho (2016) demonstrated that considering only literature with an article file type is a reasonable approach; moreover, articles constitute the majority of publications and represent independent research themes and achievements (Ho et al 2010).
In the first step, the 'front page' filter was applied as the first filter, as proposed by Fu et al (2012).As previously mentioned, the topic option in WoS includes KeyWords Plus content.However, Zheng et al (2017) pointed out that the mechanism of KeyWords Plus involves adding keywords based on additional searches, although these additional keywords are less related to the article itself, such as those from reference books.This may result in selected articles that are not related to the topic of 'traffic air pollution' (Fu and Ho 2015).Therefore, with the help of the 'front page' filter, articles that can only be searched by KeyWords Plus were excluded.The final Boolean operation equation is TS = "traffic air pollution" OR TS = "traffic air quality" OR TS = "traffic-related air pollution" OR TS = "traffic-related air quality" OR TS = "transport * air quality" OR TS = "transport * air pollution" NOT KP = (traffic OR traffic-related OR transport * ) NOT KP = (air pollution OR air quality) and PY = (2003-2022) and DT = (Article).In WoS, KP represents KeyWords Plus, PY refers to Year Published, and DT is the abbreviation for document type.The detailed filtering criteria are provided at https://webofscience.clarivate.cn/wos/woscc/summary/2cb542cd-c4f7-4ee8-90cf-ebcc61661f1b-c7cf9010/times-cited-descending/1.With further multiple screening, 304 records need to exclude by title, or were letters to the editor, systematic reviews, only abstracts.After performing selections with the help of the 'front page' filter, a total of n = 1,179 articles were selected.The selected literature was exported as Full Records and Cited References, which not only include basic information, such as author, article title, and abstract, but also contain the number of citations per article per year.

Processing highly cited article data
Microsoft Excel was used to process the above data for the next step (Li and Ho 2008).The WoS Core Collection is only a database that helps scholars search for literature; thus, it cannot be directly used for bibliometric research (Ho 2018).Therefore, the obtained data must be processed manually using Microsoft Excel software, which is a critical practice that has not yet been adopted by all scholars.
The second filter was the Times Cited (TC) filter, and it was set to TC 2022 > 100, where TC year represents the total number of citations of an article from publication to the end of a certain year (Ho 2012).The TC 2022 selections were analyzed using Microsoft Excel to determine the number of citations for each selected article from publication until the end of 2022.
Owing to TC year being a deterministic value derived from 'All Databases,' it has greater repeatability compared to 'WoS Core Times Cited' (Fu et al 2012;Mo et al 2018).The TC year and 'front page' filters aim to screen for the most popular literature in the field (Mo et al 2018).Finally, the obtained literature was integrated and exported in 'Plain Text File format' (.txt).The document should include all records, such as titles, authors, institutions, and countries.According to the CiteSpace software working environment requirements, the documents were named 'download_ * .txt' in order by the number of citations.After performing screening with the above two filters, the literature was imported to CiteSpace software to remove duplicate articles.This process resulted in n = 158 highly cited articles that were used in the data analysis stage.Based on the PRISMA 2020 review framework (Page et al 2021), a PRISMA framework was established for this study, as shown in figure 1.

Analytical methods
Bibliometrics can be used to classify the main trends in a field based on published studies (Shi et al 2022).Ellegaard (2018) noted that bibliometric analyses are widely used by the scientific community, and he also pointed out that this method is not only used in the field of library and information science but is also rapidly being used by scholars in other fields (Ellegaard 2018).
Visualization analysis technology based on bibliometrics has also been widely applied in scientific research (Geng et al 2022).Using visualization software, the development trends of a field can be intuitively displayed in a large number of studies (Kao et al 2022).This study utilized the CiteSpace software developed by Dr Chen Chaomei for the knowledge graph analysis.CiteSpace software can be used to detect and visualize publication trends and the latest developments (Chen 2006).It is a mature analysis software that was developed over many years and is widely used in various research fields for visual analysis (Geng et al 2022).In this study, CiteSpace software helped determine rapidly growing topics and hotspots to identify publication trends and turning points.More importantly, CiteSpace includes excellent visualization technology, thus making it easier for users to detect and visualize the structure, temporal patterns, and trends of specific research fields (Liu et al 2015).
Therefore, this study integrated traditional bibliometric methods and visual analysis software.CiteSpace (6.2.R4 and 6.2.R6) software was used to provide visualized comparisons and perform a network analysis.The main modeling method of this article is as follows.(1) For the analysis of publication trends, this article used 1,179 articles obtained during data collection for analysis.The main reason for including this number of studies was that the analysis of publication trends requires data over an extended time period.Thus, although some studies were not well cited, the work had to be included to determine trends over time.(2) For the collaboration analysis and research focus analysis, this study included 158 articles obtained after processing highly cited article data, which focused on analyzing countries, institutional collaborations, author collaborations, keywords, and hotspots.The main reason for considering only highly cited literature is that the trends in hotspots, effective cooperation networks, and authors' reputations in the field can be considered.The results showed an increasing publication trend over time.Before 2018, the annual publication volume did not exceed 100 articles, which represents a relatively small number.At this stage, relatively few researchers were involved in TRAP research, the number of articles was insufficient, and interdisciplinary research on traffic and urban air pollution was not popular worldwide.However, the articles published during this period provided a foundation for future research on TRAP.From 2018 to 2022, the number of articles showed a rapid growth trend, with a peak of 114 in 2021.This indicates that over time, scholars have begun to pay attention to the topic of TRAP and a large number of scholars have begun to conduct in-depth research on TRAP, thereby providing support to advance this field.The findings reveal that the topic of TRAP is worthy of long-term discussion, and a trend of interdisciplinary intersection has been observed in academic circles.Scholars in the field of traffic or the urban environment can participate in this intersection in the future, which will jointly promote the development of intersectional research between traffic and the urban environment.

Analysis of journal circulation
The 1,179 articles collected over the 20-year study period were classified by journal to determine the most influential journal in the field of TRAP.By utilizing the screening function of WoS, the top ten journals with the largest circulations were obtained, as shown in figure 4. The journal with the largest circulation was Environmental Research, which published 98 articles related to TRAP over the study period, thus accounting for nearly 10% of the total.Atmospheric Environment, and Environmental Health Perspectives are also major journals in this field.Thus, Environmental Research, Atmospheric Environment, and Environmental Health Perspectives have provided significant support for research on TRAP.
To evaluate the circulation trends of these 10 journals in the field of TRAP over time, this article constructed a three-dimensional histogram, as shown in figure 5.The Z axis in the histogram represents the circulation of the topic TRAP for each journal, the Y axis lists the top journals, and the X axis shows the research phases, which were divided into four parts (2003-2007, 2008-2012, 2013-2017, and 2018-2022).The construction of a histogram is common in bibliometric studies (Huang et al 2022).

Collaboration analysis
Identifying collaborative networks in research among different countries, institutions, and authors is important (Zhou et al 2018) because it can help identify leading and cutting-edge countries, institutions, and authors in this  field (Yu and Chen 2021).Moreover, assessing the collaboration mode and degree can help demonstrate collaborative relationships at different stages in the field of TRAP.

Country and region analysis
CiteSpace software can be used to plot a cooperation network between countries or regions, as shown in figure 6.This study found that the TRAP-related cooperation network included 80 national nodes and 431 cooperative relationships.The top 15 countries with the largest circulation and centrality are listed in table 1.
Three conclusions can be drawn from the data shown in figure 6 and table 1.
(1) The publication volume from developed countries is significantly higher than that of developing countries, indicating that developed countries are more concerned about the intersection of traffic and the urban environment and their research on TRAP is in a leading position.(2) A total of 589 counts were from the USA, making the USA the largest node.Moreover, the USA had the highest centrality value (0.49) among all countries, which indicates that the USA has a leading position in research on TRAP and thus can attract other countries to cooperate.(3) The most collaborative countries are also considered major countries in the automotive industry.The USA is famous for its General Motors (GM), England is noted for the world-renowned luxury brands Bentley and McLaren, and Germany is renowned for the best-selling global car brands Mercedes Benz, Volkswagen, and BMW.Therefore, these findings suggest that automotive manufacturers are paying attention to the urban air pollution caused by   (4) Although Canada accounts for a large number of publications, its centrality is not high (lower than that of the USA, England and Germany), as shown in figure 6 and table 1.This indicates that over the past 20 years, although Canada has made great efforts in the field of urban traffic-air pollution control, its research institutions are more willing to cooperate with their own institutions and are not prominent in the international cooperation network.This is a noteworthy issue, and Geng et al (2022) suggested that strengthening academic exchange and cooperation with other countries represents a win-win situation.Therefore, the intricate connections reveal that TRAP has become a globally involved research topic.Collaborating and exchanging ideas with researchers from other countries will also help further advance TRAP research.
Based on this, this study examines citation bursts in all countries as an indicator to identify the strength of trends (Kim and Chen 2015).This indicator reflects the concentration and short-term explosive strength of certain countries and is not necessarily related to the volume of publications.Figure 7 shows the 15 counties with the strongest citation bursts.
Among the top 15 countries with the strongest citation bursts, China ranked highest with a strength of 6.87.Since 2017, China has produced many high-level articles.It is worth noting that some powerful developing countries also presented strong citation bursts, with China (6.87), Mexico (3.18), and Saudi Arabia (3.02) ranking among the top three countries.This indicates that among these important developing countries, the strength of scientific research in the later stages is strong and significant attention has been focused on the field of TRAP in the short term.Research in these three countries played a propulsive role within a certain time frame.
In summary, the USA started early in the field of TRAP and currently maintains a leading and core position in academic cooperation.The relevant research achievements in Canada, England and Germany have had an important impact and cannot be ignored, whereas the research in China, Mexico, and Saudi Arabia is still in its developmental stage but has had a strong impact.

Core institution analysis
The top 15 collaborative institutions in TRAP research and their centralities are listed in table 2. Owing to the affiliation of the Harvard T.H. Chan School of Public Health to Harvard University, post-processing is needed to merge the findings for the two.The situation is similar for the University of California Berkeley and University of California System.This processing requires the use of the CiteSpace function 'add to the alias list'.As shown in table 2, the University of California System in the USA has the highest number of publications in this field (121 counts) and a relatively high centrality at 0.11.This finding indicates that although the University of California System has conducted extensive research in the fields of traffic and urban air pollution, there are many specific subdivisions in this fields, possibly due to the differences in research between different campuses.Similarly, Harvard University in the USA has a large number of publications at 89 and high centrality at 0.16.These findings indicate that it is at the core of this research topic (Cao et al 2023).The cooperation network between core institutions can be drawn using CiteSpace software, as shown in figure 8.This network includes 274 institutional nodes and 1,657 cooperative relationships.The following conclusions can be drawn.(1) universities are the main institutions for studying TRAP issues, and a close cooperation network occurs between different research institutions.(2) extensive academic exchanges were conducted with the University of California System and Harvard University as centers.(3) a close cooperative relationship occurs between research institutions in the USA and Europe, with little participation from Chinese institutions.
This indicates that Chinese universities and institutions should continue to promote exchanges and cooperation with other scientific research institutions to strengthen China's research in the field of TRAP.Specifically, cooperation with institutions in the USA should be strengthened because they are located at the center of this field.From a global perspective, a large number of articles has been published on TRAP, which is an urgent and potential threat to human health, and strong centrality is observed.Institutions with strong centrality should actively strengthen their contact with other institutions worldwide to accelerate research in this field.At the end of this section, information will be provided to help relevant scholars determine which institutions have the potential for joint research.

Core author analysis
Figure 9 shows the cooperation network between core authors for TRAP research drawn using CiteSpace software.The top 15 authors in terms of highest publication circulation were also exported and are summarized in table 3.
Cooperation among authors is more dispersed than that among countries or institutions.Brunekreef and Bert, Brauer and Michael, and Sunyer and Jordi are the three central author teams and contributed to the most articles.These authors prefer to communicate with familiar researchers.The team of Brunekreef and Bert, which is the most central team, is the most prominent and has a large number of researchers and a large number of published articles.Meanwhile, some small peripheral teams shown in figure 9 have not yet established a cooperative relationship with the central team.Taking Hatzopoulou and Marianne's team as an example, they published 13 papers and ranked 8th among the top 15 authors in terms of TRAP research; however, their  centrality was 0, which indicates a certain distance from the core research team.Geng et al (2022) pointed out that this is not an ideal state and that researchers need to work together to achieve breakthroughs in the field of traffic and urban air pollution rather than relying on scattered individualism.

Research focus analysis
3.3.1.Keyword analysis Grant (2010) noted that keywords play an important role in the academic field.A bibliometric analysis of keyword co-occurrence can help explore the development trajectory and hotspots of a given research field (Cao et al 2023).Additional processing was required for subsequent analyses.
(1) The terms 'traffic,' 'air,' and 'pollution' and any combination words had to be deleted because these words are artificially set in WoS searches and do not objectively reflect keywords in the field.This step requires the use of the CiteSpace function 'add to the exclusion list', as shown in figure 10. (2) Keywords with same meaning must be merged.For example, 'NO 2 ' was merged into 'nitrogen dioxide'; 'ultrafine particles,' 'ultrafine,' 'PM 2.5 ,' 'PM 10 ,' 'ambient particulate matter,' 'airborne particulate matter,' and 'fine particulate matter' were merged into 'particulate matter.'This step requires the use of the CiteSpace function 'add to the alias list', which is also shown in figure 10.Moreover, it represents a common subsequent processing method in the field of bibliometrics for analyzing air pollution (Cao et al 2023).After the above operations, the top 15 keywords with the highest frequency of co-occurrence were obtained and are listed in table 4.
Among the selected keywords, (1) 'particulate matter,' 'nitrogen dioxide,' 'particles,' and 'pollutants' indicated the main chemical substances involved in pollution as well as the main contaminants that are currently the focus of researchers.(2) 'Exposure' and 'long term exposure' revealed the contact pathways between humans with pollutants, which is important because long-term exposure to pollutants leads to the development of a number of diseases (Liu et al 2022).(3) 'Association' and 'oxidative stress' reveal the main scientific methods and indicators of research in this field, indicating that scholars are attempting to explore the impact of TRAP using these methods.(4) 'Health,' 'mortality,' 'children,' and 'risk' reveal that some researchers are concerned about the impact of TRAP on humans and the next generation.These researchers paid more attention to the sustainable development of traffic.Subsequently, the keywords with the strongest citation bursts were combined (figure 11) and meaningless nouns were filtering out.The results showed that that research related to 'use regression models,' 'physical activity,' 'positive matrix factorization,' and 'particles' has presented a strong citation burst over a short period of time.Therefore, over the time period shown in figure 11, the associated researchers had a high level of passion and enthusiasm for these keywords.

Cluster analysis
Keywords of the same type can be aggregated into the same cluster using the 'cluster' function of CiteSpace, and the most frequently co-occurring keyword can be used as the name of a cluster.Keywords in the same cluster usually have a certain degree of similarity.In figure 12, each colored border represents a cluster, the cluster names are marked in red, the cluster ID is indicated before the cluster name, and the keywords within each cluster are represented in black.It should be noted that CiteSpace has certain threshold requirements for the average evaluation contours (S) and modular values (Q) generated after clustering.For example, S > 0.5 represents reasonable clustering, S > 0.7 represents convincing clustering, and modular Q > 0.3 indicates significant clustering (Cao et al 2023).In this analysis, S was 0.7056 and Q was 0.3393.This clustering was considered convincing and significant.The top six clusters were then selected for analysis.The clustering labels show that the classification of clusters can be divided into two categories: (1) methods, models, or indicators of TRAP and (2) the impact of TRAP.
(1) The methods of studying TRAP mainly involve clusters '# 0 source apportionment,' '# 4 air quality,' and '# 5 mobile monitoring.'Among them, '# 0 source apportionment' is a commonly used model in atmospheric pollution that reconstructs the impact of different sources of atmospheric pollutant emissions, such as the impact of particulate matter, based on detected environmental data (Viana et al 2008).Therefore, keywords related to particulate matter, such as 'PM 10 ' and 'PM 2.5 ,' are all in this cluster.Cluster '# 4 air quality' is an indicator for studying urban air pollution.The keywords 'nitrogen dioxide (NO 2 ),' 'sulfur dioxide (SO 2 ),' 'carbon monoxide (CO),' and 'ozone (O 3 )' all belong to air quality (Pelaez et al 2020), and they were correctly placed in cluster '# 4 air quality' during the cluster analysis.This further confirms the reliability of the analysis.Cluster '# 5 mobile monitoring' includes popular detection methods for studying urban air pollution.Wu et al (2020) noted that immovable monitoring stations cannot capture the heterogeneity of air pollution in complex areas.Therefore, researchers are now utilizing fleet vehicles equipped with mobile atmospheric monitoring systems to measure and evaluate pollutants (Zhao et al 2021).The Portable Emission Measurement System (PEMS) is one of the key mobile monitoring components for collecting real-time emission data from vehicles (Zhu et al 2022).In addition, unmanned aerial systems (UASs) equipped with portable air pollution monitors are increasingly being used for mobile monitoring (Li et al 2020).In '# 5 mobile monitoring,' the keyword 'black carbon' is an air indicator of high concern among scholars (Zhang et al 2019).By accurately identifying three clusters of measurement methods, models, and indicators using CiteSpace software, popular research methods in the field of TRAP were identified for future researchers.
(2) The impact of TRAP mainly involves '# 1 asthma,' '# 2 heart rate variability,' and '# 3 physical activity.'Among them, '# 1 asthma' is a chronic respiratory disease, and TRAP is an important cause (Tiotiu et al 2020).He et al (2023) emphasized that some particles generated by urban traffic can deposit in the respiratory system and cause serious health problems.For '# 2 heart rate variability,' decreased heart rate variability (HRV) is a sign of poor cardiac autonomic nervous function, which is related to air pollution, especially PM 2.5 (Park et al 2005).Moreover, scholars have indicated that HRV is likely caused by TRAP (Wu et al 2013).For '# 3 physical activity,' exercise increases the intake of urban air pollution and causes damage to the brain (Bos et al 2014).The keyword 'long-term exposure' in this cluster is also associated with brain damage.Bos et al (2014) proposed that exposure to TRAP during exercise may inhibit brain function.
The above analysis demonstrates the accuracy of the clusters in this study, which successfully identified keywords in the same field and provided a summary of the associated effects.By reviewing these high-cooccurrence keywords, future researchers can comprehensively and quickly determine the meaning of the keywords; thus, this finding provides valuable reference guidance for future research (Geng et al 2022).

Hotspots trend analysis
Trend analysis and change point detection in time series are commonly used analytical tools (Sharma et al 2016).Hotspot trend analysis can also determine the cutting-edge direction of a field and help to understand past and future research hotspots.In the section on keyword analysis, six clusters were successfully selected.The keywords of the six clusters were revealed over time, as shown in figure 13.This section focuses on analyzing the top three largest clusters, namely, '# 0 source apportionment,' '# 1 asthma,' '# 2 heart rate variability,' and exploring the temporal distribution of keywords.
Cluster '# 0 source apportionment' represents the core research direction in the field of TRAP.The transition of the research direction to '# 0 source apportionment' over the past 20 years can be divided into three stages.(1) From 2003 to 2009, the research hotspots mainly included 'particulate matter,' 'aerosol,' 'elements,' and 'chemical composition.'These keywords mainly focus on the analysis of chemical components.(2) From 2009 to 2016, the research hotspots included 'emission factors,' 'positive matrix factorization,' 'source identification,' and 'identification.'The hotspot 'emission factors' is used to calculate greenhouse gas emissions, such as carbon dioxide emissions, which is a parameter used in the modeling process for environmental impact assessments (Shan et al 2018).Positive matrix factorization (PMF) was successfully used in chemometric assessments by Brereton (2013) and represents a method of evaluating toxic and harmful gases by identifying potential sources of PAHs (Wang et al 2009).For the hotspot 'identification', Jiang et al (2022) obtained the vertical distribution of PM 2.5 by identifying the structure of the atmospheric boundary layer.Ma et al (2020) used machine learning technology to identify air quality.These keywords are all involved in the process of evaluating the impact of traffic on urban air pollution modeling.Therefore, during this period, the main focus was on evaluation models and methods.(3) From 2016 to 2022, the main keywords were 'seasonal variation,' 'nonmethane hydrocarbons,' and 'climate.'During this period, the research focus shifted to exploring the changes in urban traffic-air pollution levels related to climate and season.In addition, the research on chemical components also shifted from 'aerosols' to 'nonmethane hydrocarbons.' For the '# 1 asthma' cluster, the following trends were observed.Based on the analysis of keyword clustering and the hotspot trend graph, we found that development in each cluster in the field of TRAP has occurred over the past 20 years, with a shift from simple research on chemical components to more basic research on models and methods, followed by research on diseases caused by TRAP and collaborative research on multiple diseases.This means that many new directions and ideas have emerged from the published research.Thus, the findings presented here will provide a guide for other researchers in determining their research focus.

Limitations
In this study, CiteSpace was used for the bibliometric analysis of the TRAP field; however, CiteSpace itself has certain limitations.For example, the setting of the analysis algorithm can introduce differences that may affect the results (Zhang et al 2023).In addition, standard expressions may not be used for certain regions and institutions; however, this limitation did not affect the analysis results of this article.For example, the general expression P.R.China is displayed as Peoples R China in CiteSpace; however, since such expressions will not affect the readers' understanding, the originality and authenticity of the data will not be impacted.In addition, the data samples in this study were only obtained from WoS.Although WoS is considered one of the most reliable scientific publication databases, due to limitations in the scope of the database, meaningful publication information may not be available (Huang et al 2022).

Conclusion
In this study, 1,179 articles related to TRAP in the WoS from 2003 to 2022 were screened.Further in-depth analysis was conducted on the 158 highly cited articles selected, and academic development and trends in this field were visualized using CiteSpace software.The 'front page' and TC year filters used for data processing are relatively rarely used among other articles of the same type.A framework can be drawn based on the results presented herein, as shown in figure 14.The findings may help researchers and other personnel to clarify their research directions.
The main conclusions of this article are as follows.
(1) The number of publications has shown a tortuous upward trend, and scholars worldwide are gradually focusing on the intersection of environmental and green traffic research.
(2) The USA showed an early focus on the field of TRAP and has maintained a leading position.Canada, England and Germany are authoritative countries in this field.Academic research in China, Mexico, and Saudi Arabia is still in the developmental stage, but these countries have had a strong impact.
(3) In terms of journal circulation, Environmental Research, Atmospheric Environment, and Environmental Health Perspectives have provided significant support for TRAP research.
(4) For the core research institutions, a large amount of research has been conducted by the University of California System and Harvard University in the USA.Currently, scientific exchange and cooperation are observed between the USA and Europe, whereas Chinese institutions have shown only limited participation in scientific research exchange.
(5) Cooperation between authors is relatively dispersed, with scholars showing a greater tendency to communicate with familiar researchers.The teams of Brunekreef and Bert, Brauer and Michael, and Sunyer and Jordi have made the greatest contributions to research on TRAP.
(6) 'Particulate matter,' 'exposure,' 'health,' 'nitrogen dioxide,' and 'mortality' are the five keywords with the greatest frequency in this field, and they represent hotspots in the field of TRAP research.
(7) The main research directions and trends were subjected to the following processing: A: the main contaminants involved in pollution were explored; B: the main methodologies and modeling methods in the field of TRAP were evaluated; and C: the impact of TRAP on humans and the next generation was identified.
Compared with similar bibliometric articles, this study focused on the importance of data preprocessing.Some authors have neglected the fact that the WoS Core Collection cannot be used directly for bibliometric research.Thus, the research presented in this article is more accurate and objective.Moreover, it utilizes multiple filters and Microsoft Excel preprocessing to ensure that the selected papers are highly cited, thereby reflecting the direction of development in the field of TRAP.Owing to the limitations of CiteSpace software and the diversity of bibliometric analysis methods, the conclusions presented here may differ from that of other articles.For future research directions, scholars should pay more attention to the impact of TRAP on humans and the next generation, which will provide reference guidance for future scholars.

Figure 1 .
Figure 1.PRISMA framework for this study.
(3) The obtained knowledge network graphs were then analyzed to reveal the current research status, overall evolution of this field in recent years, and future development trends.A flowchart of the overall analysis method is shown in figure 2.
1. Temporal analysis of publications An analysis of the 1,179 articles obtained during data collection rapidly revealed the changes in the number of publications in the field of TRAP each year.The time distribution of publications from 2003 to 2022 is shown in figure 3.

Figure 5
Figure 5 clearly shows that Environmental Research had a dominant position in recent years.During the 15 years from 2003 to 2017, Atmospheric Environment, and Environmental Health Perspectives maintained the largest circulation and interest in traffic and urban air pollution.These two journals published research on TRAP during the early period and had a dominant position for a long time.From 2018 to 2022, Environmental Research, and Science of the Total Environment showed a rapid increase in the level of interest in the field of TRAP.An interesting phenomenon worth paying attention to is that the topic of TRAP appeared extensively in human health-related journals from 2008 to 2017.However, since 2018, the topic of TRAP has been published in a large number of environmental journals, and this trend has maintained an upward trajectory.Therefore, research on TRAP has attracted more attention from scholars and journals over time.

Figure 4 .
Figure 4. Key journals in the field of TRAP and their circulation.

Figure 5 .
Figure 5. Circulation trends of the top ten key journals.

Figure 6 .
Figure 6.Cooperation network between countries or regions.
cars and are engaging in research, development, and updates in an attempt to reduce the impact of traffic on the urban environment.However, China differs from other countries.In China, electric vehicles are extremely popular, and the electric vehicle brand BYD is extremely well represented in the market and rapidly expanding (Qudrat-Ullah 2022).These findings indicate that China has also focused considerable attention on green traffic and air pollution(Wang and Harvey 2015).

Figure 7 .
Figure 7. Top 15 countries with the strongest citation bursts.

Figure 9 .
Figure 9. Cooperation network between core authors.

Figure 11 .
Figure 11.Top 15 keywords with the strongest citation bursts.

( 1 )
From 2003 to 2011, the main keywords were 'cohort,' 'exposure,' and 'risk factors.'During this period, research primarily focused on exploring the causes of asthma.(2) From 2011 to 2022, the keywords evolved to 'preheating,' 'neuroimaging,' and 'social distance,' which are not directly related to asthma.During this period, research mainly focused on exploring the related effects of TRAP and diseases other than asthma.The 20-year keyword change trend under the '# 2 heart rate variability' cluster shows the flow of 'oxidative stress -cardiovascular disease -inflammation -coronary heart disease -diabetes mellitus -cardiovascular health.'The research direction has changed over the past 20 years and mainly focused on different diseases at different times.

Figure 14 .
Figure 14.Framework based on the TRAP findings.

Table 1 .
Top 15 most collaborative countries for TRAP research.

Table 2 .
Top 15 most collaborative institutions for TRAP research.

Table 3 .
Top 15 most collaborative authors for TRAP research.

Table 4 .
Top 15 co-occurrence keywords for TRAP research.