Exploring the Emerging Evolution Trends of Urban Resilience Research by Scientometric Analysis

Numerous studies in urban resilience have been published in the past decade. However, only a few publications have tracked the evolution trends of urban resilience research, the findings of which can serve as a useful guide for scholars to foresee worth-effort research areas and make the best use of precious time and resources. In order to fill the research gap, this study performed a scientometric analysis on the evolution trends of urban resilience research using a versatile software package-CiteSpace. The scientomentric analysis focuses on distribution of lead authors and their institutions, high frequency categories and keywords, high influential journals, author contribution, and evolutionary trends based on co-author analysis, co-word analysis, co-citation analysis and cluster analysis of documents. This study discoveries that first, the U.S., England, Australia, Canada, China and Sweden are the countries that make the most significant contributions in the advancement of urban resilience research; second, the existing urban resilience research focuses primarily on environmental studies, geography and planning development; third, hot topics of the urban resilience research keep shifting from 1993 to 2016; fourth, the knowledge body of urban resilience research consists of five clusters: resilience exploratory analysis, disaster resilience, urban resilience, urban resilience practice, and social-ecological systems; last, the emerging trends in urban resilience research include defining urban resilience, adaptation model, case studies, analytical methods and urban social-ecological systems, resulting in cutting-edge research areas in urban resilience.


Introduction
The concept of resilience has become a very popular area of research in terms of publications [1 -3] and been adopted in various disciplines, such as physics, psychology, and social-ecological systems [4][5][6]. The urban population increased from 10% to over 50% in the past two decades with the unprecedented urbanization process [7]. Resilience is becoming attractive in urban research with cities theorized as highly complex and adaptive systems [8,9]. Urban resilience reflects the ability of urban systems to maintain and rapidly recover to the desired status after external disturbance, adapting to changes and quickly transforming systems to a new balanced status [10]. Despite great popularity among researchers, urban resilience has no commonly accepted definition: "Urban resilience reflects the degree to which cities tolerate alteration before reorganizing around a new set of structures and processes [11]." "A disaster-resilient city can be understood as a city: (a) reduce or avoid current and future hazards; (b) reduce current and future susceptibility to hazards; (c) establish functioning mechanisms only to prune the merged network [82,83]. We propose a modular Q value and an value greater than 0.5 to ensure the reliability and importance of the cluster [84, ness centrality between nodes is defined as the shortest paths between other vert through the node [86][87][88]. Moreover, the publications records contain different scien tions, such as authors, institutions, keywords, articles and terms, which represent f nodes in scientific collaboration networks. In the using process of CiteSpace, users c node type or multiple concurrent node types. The created links represent the rela different nodes, which represent co-citation or co-occurrence relationships.

a Collection for Scientometric Analysis and Visualization
is study adopts the scientometric perspective to review the emerging trends o ce research in urban studies. Three search criteria are established for publication r only academic journals were selected for scientometric analysis, in consideration positions in urban resilience domain. Secondly, different keywords related to urban r in the publication retrieval. Thirdly, the time span of publication retrieval should b ible. Based on the above three document retrieval principles, this study searched 39 to the Web of Science (WOS) core collection database for urban studies [89,90]. A ds used in our search strategy were selected based on previous urban resilience

Data Collection for Scientometric Analysis and Visualization
This study adopts the scientometric perspective to review the emerging trends of urban resilience research in urban studies. Three search criteria are established for publication retrieval. Firstly, only academic journals were selected for scientometric analysis, in consideration of their impact positions in urban resilience domain. Secondly, different keywords related to urban resilience are used in the publication retrieval. Thirdly, the time span of publication retrieval should be as long as possible. Based on the above three document retrieval principles, this study searched 39 journals related to the Web of Science (WOS) core collection database for urban studies [89,90]. All of the keywords used in our search strategy were selected based on previous urban resilience studies. Ultimately, three keywords are selected in this study include resilience, resilient and resiliency. The time span of publication retrieval is from January 1993 and December 2016, because the first paper in 39 journals related to urban studies is published in 1993 [91].
After pre-analysis and comparison, we decided to apply the following retrieval strategy: [(keywords = ("resilience" OR "resilient" OR "resiliency") AND Publication name = ("Landscape and Urban Planning" OR "Journal of Urban Planning and Development" OR "Journal of Urban Economics" OR "European Urban and Regional Studies" OR "Cities" OR "Habitat International" OR "Urban Forestry & Urban Greening" OR "Urban Studies" OR "International Journal of Urban and Regional Research" OR "Journal of Planning Literature" OR "Housing Policy Debate" OR "Environment and Urbanization" OR "Urban Geography" OR "Housing Studies" OR "Journal of Urban Technology" OR "Journal of the American Planning Association" OR "City & Community" OR "European Planning Studies" OR "Journal of Planning Education and Research" OR "Journal of Housing Economics" OR "Housing Theory & Society" OR "Regional Science and Urban Economic" OR "Economic Development Quarterly" OR "Journal of Contemporary Ethnography" OR "Urban Affairs Review" OR "International Regional Science Review" OR "Urban Policy and Research" OR "Journal of Urban Affairs" OR "Real Estate Economics" OR "Urban Education" OR "Journal of Housing and the Built Environment" OR "Journal of Real Estate Finance and Economics" OR "Education and Urban Society" OR "Journal of Architectural and Planning Research" OR "EURE-Revista Latinoamericana De Estudios Urbano Regionales" OR "Urban Design International" OR "Journal of Urban History" OR "Open House International" OR "Urban Lawyer"))]. 405 records are obtained after the publication retrieval process on 18 February 2017, and these records are downloaded and saved in TXT format.
In the WOS core collection, the publication retrieval records can be analyzed from different aspects, such as author, source journal, category, and references. After eliminating the invalid records, 355 valid document records were obtained, of which includes articles for 96.62% (343) including proceedings papers for 3.1% (11), other 3.38% (12) publications records are reviews. After the pre-analysis, 355 articles and review publications records were selected as the database used in this study during scientometric analysis. The number of these 355 articles published each year is shown in Figure 2. "Journal of Urban Affairs" OR "Real Estate Economics" OR "Urban Education" OR "Journal of Housing and the Built Environment" OR "Journal of Real Estate Finance and Economics" OR "Education and Urban Society" OR "Journal of Architectural and Planning Research" OR "EURE-Revista Latinoamericana De Estudios Urbano Regionales" OR "Urban Design International" OR "Journal of Urban History" OR "Open House International" OR "Urban Lawyer"))]. 405 records are obtained after the publication retrieval process on 18 February 2017, and these records are downloaded and saved in TXT format.
In the WOS core collection, the publication retrieval records can be analyzed from different aspects, such as author, source journal, category, and references. After eliminating the invalid records, 355 valid document records were obtained, of which includes articles for 96.62% (343) including proceedings papers for 3.1% (11), other 3.38% (12) publications records are reviews. After the preanalysis, 355 articles and review publications records were selected as the database used in this study during scientometric analysis. The number of these 355 articles published each year is shown in Figure 2.  Figure 2 shows the current state of urban resilience research and shows the distribution of 355 document records for 39 journals related to urban studies in the WOS core database from January 1, 1993 to December 31, 2016. The publication number per year related to urban resilience presents discontinuous characteristic and only has a very small number before 2000. Then, the total number of annual publications related to urban resilience in the WOS presents slight increasing and fluctuation trend between 2000 and 2010. Since then, it has grown steadily since 2010, and the number of publications reached its peak in 2015. Figure 2 details the annual number of publication on urban resilience research.  1995  1996  1997  2000  2001  2002  2003  2004  2005  2006  2007  2008  2009  2010  2011  2012  2013  2014  2015  2016 Publication number Figure 2. Publications records related to urban resilience in the WOS core collection, published from January 1993 and December 2016. Figure 2 shows the current state of urban resilience research and shows the distribution of 355 document records for 39 journals related to urban studies in the WOS core database from January 1, 1993 to December 31, 2016. The publication number per year related to urban resilience presents discontinuous characteristic and only has a very small number before 2000. Then, the total number of annual publications related to urban resilience in the WOS presents slight increasing and fluctuation trend between 2000 and 2010. Since then, it has grown steadily since 2010, and the number of publications reached its peak in 2015. Figure 2 details the annual number of publication on urban resilience research.

Results Analysis
In this section, we analyze a series of scientometric networks using the publications data retrieved in the section of data collection. The relevant publication retrieval records are analyzed and visualized from three aspects: co-author analysis, co-word analysis, and co-citation analysis. Through co-author analysis, we assess and analyze the academic influences of authors, countries and institutions in the field of urban resilience. The main research topics of urban resilience are discovered by co-word analysis. Additionally, the scientific structure and development trends for urban resilience are extracted and generalized by co-citation networks analysis.

Co-Author Analysis
In this section, we analyze the academic cooperation information of authors obtained from the publication retrieval records and assess leading authors, countries, and institutions on urban resilience studies. The scientometric data is also used to generate and analyze the co-author network and the author's countries and institutional networks.

Co-Authorship Network
The academic collaboration can describe the productivity level and contribution of researchers to scientific research. Through co-authorship analysis, Figure 3 shows that the author's contribution on urban resilience research presents a decentralized trend in the co-authorship network; no scholar has an absolute advantage on the publications outputs of urban resilience. The co-authorship network has 424 nodes and 315 links, including the lead authors of publications. The node represents the author, and the link between the two nodes indicates that the two authors directly collaborate through writing papers together. Figure 3 reveals the outputs and co-author relationships of leading authors in urban resilience research. The size of the authors' nodes presents a positive correlation to the numbers of authors' publications. The thicknesses of lines between authors indicate the levels of the collaborative relationships in a given year. The different colors of links represent different phases during January 1993 to December 2016. Figure 3  to scientific research. Through co-authorship analysis, Figure 3 shows that the author's contribution on urban resilience research presents a decentralized trend in the co-authorship network; no scholar has an absolute advantage on the publications outputs of urban resilience. The co-authorship network has 424 nodes and 315 links, including the lead authors of publications. The node represents the author, and the link between the two nodes indicates that the two authors directly collaborate through writing papers together.   Table 1 lists the authors who have most publications on urban resilience research in recent years.
The co-author relationships only reflect the outputs and contribution of publications on urban resilience research. The highly productive authors are not necessarily highly influential to urban resilience research. The levels of authors' influence are reflected by co-citation analysis. To solve this problem, the co-citation frequency of authors is compared and analyzed in later co-citation networks analysis. The co-author relationships only reflect the outputs and contribution of publications on urban resilience research. The highly productive authors are not necessarily highly influential to urban resilience research. The levels of authors' influence are reflected by co-citation analysis. To solve this problem, the co-citation frequency of authors is compared and analyzed in later co-citation networks analysis.

Network of Co-Authors' Countries and Institutions
The authors' countries and institutions form a collaborative relationship network which reflects the collaborative relationship between countries and institutions. To characterize the collaborative relationship of publications on urban resilience, CiteSpace is used to form this collaborative relationship network. According to the publication frequency, nodes represent different countries and institutions in collaborative relationship network, the links represent the collaborative relationship between different countries and institutions on urban resilience research. The academic collaborative relationships between countries and institutions on urban resilience research are shown in Figure 4. relationship of publications on urban resilience, CiteSpace is used to form this collaborative relationship network. According to the publication frequency, nodes represent different countries and institutions in collaborative relationship network, the links represent the collaborative relationship between different countries and institutions on urban resilience research. The academic collaborative relationships between countries and institutions on urban resilience research are shown in Figure 4.   Figure 4 shows that the majority of the total publication output mainly comes from the USA, England, Australia, Canada, China and Sweden. Regarding the geographical distribution of the publications output, the country with the largest publications output is the USA with 136, followed by England with 55, Australia with 24, Canada with 19, China with 15, Sweden   Figure 4 shows that the majority of the total publication output mainly comes from the USA, England, Australia, Canada, China and Sweden. Regarding the geographical distribution of the publications output, the country with the largest publications output is the USA with 136, followed by England with 55, Australia with 24, Canada with 19, China with 15, Sweden with 15, South Africa with 13, Germany with 12, Netherlands with 11 and Spain with 10. The urban resilience research is conducted mainly in countries with high urbanization levels or in the process of urbanization. Figure 4 shows that the U.S., England and Australia are the first countries to research urban resilience. Although USA is the largest publication contributor among all of the countries, England has almost the same influence as the USA in the collaborative relationship network. Interestingly, some countries that appear as major publication contributors have less influence in the collaborative relationship network, such as The Netherlands. Additionally, China as the biggest country in the process of urbanization has close collaborative relationships with almost countries contributing high publications output, including the USA, England, Australia and Canada.
The lines and circular of nodes in Figure 4 show the contribution and cooperation of research institutions in the field of urban resilience. The USA has the most active research institutions on urban resilience research, with the largest number of publications published between January 1993 and December 2016. The institution with highest number of publications is Texas A&M University (with 10 publications), followed by Arizona State University (6), UCL (6), University of Melbourne (6), King's College London (5), Stockholm University (5), University of Manchester (5), Chinese Academy of Sciences (4), University of Birmingham (4), University of Illinois (4) and University of North Carolina (4). The top 11 institutions includes four institutions from the USA, other institutions are mainly from countries with high publication outputs, such as England, Australia and China. Stockholm University is ta traditional resilience research center which has some world-renowned researchers on resilience, such as Professor Carl Folke [91].
Betweenness centrality between nodes represents the impact of nodes on the flow of information between other nodes. There are some nodes with high betweenness centrality values that countries such as the USA (0.57), England (0.57), China (0.29), Australia (0.26), Sweden (0.2), Spain (0.2), South Africa (0.15), Canada (0.12), Germany (0.09), Italy (0.07), Netherlands (0.05) and institutions such as Arizona State University (USA, 0.1), University of Manchester (England, 0.08), Gran Sasso Science Institute (Italy, 0.08), Stockholm University (Sweden, 0.07), Florida State University (USA, 0.05). These nodes play important roles in the cooperative network of urban resilience research. Therefore, the publications from countries and institutions with high betweenness centrality should be paid more attentions, including countries such as the USA, England, China, Australia, and institutions such as Arizona State University, University of Manchester, Gran Sasso Science Institute, Stockholm University, Florida State University.

Co-Word Analysis
In this section, we analyze the changes of research topics on urban resilience research from January 1993 to December 2016. The main contents include co-occurring categories network analysis and co-keywords network analyses, which are used to analyze the evolution trends and academic frontiers of urban resilience research.

Network of Co-Occurring Categories
The 355 records retrieved from the 39 journals related to urban research in the WOS Core Collection Database can be attributed to one or more categories based on the WOS categories of source journals. To analyze the subject categories distribution of publications on urban resilience research, we summarize the subject categories of 355 publications on urban resilience research in Table 2  In addition to the category of Planning Development, all the 5 main categories have the publications of urban resilience in three panels between 1993 and 2016. Other 10 categories contribute the less numbers of publications on urban resilience research during 1993 and 2016. Only the category of Education Educational Research appears in the early research of urban resilience. The three categories of Engineering Civil, Law and Sociology just appear in the research of urban resilience after 2010. In order to accurately analyze the evolution trends of urban resilience, we use CiteSpace to generate the network of co-occurring categories on urban resilience as shown in Figure 5.     Figure 5 shows that the top three categories in urban resilience research are Urban Studies, Environmental Studies and Ecology. In addition to the subject category of Urban Studies, Environmental Studies is the biggest subject category, which means that urban resilience is mainly researched from the environmental perspective. Interestingly, some categories as the major publication contributors have less influence in co-occurring categories network, such as Ecology and Geography Physical. In Figure 5, the node with the highest betweenness centrality is represented by red-ring. Some nodes with high betweenness centrality in co-occurring subject categories network are Urban Studies (0.17), Environmental Studies (0.15) and Planning Development (0.11), which represent these subject categories are the major turning nodes linking the urban resilience research in different phases and play important roles in the development of urban resilience research.

Network of Co-Occurring Keywords
Co-occurring keyword network analysis provides the basic information of core research content and helps researchers to track the development trends of research topics at different stages of urban resilience research. In co-occurring keyword analysis process using CiteSpace, keywords are composed of "author keywords (DE)" and "keyword plus (ID)". The author keyword (DE) is derived from the author's article, and the keyword plus (ID) is identified by the journal which the author's articles are published. In the process of co-occurring keywords analysis, the keywords with similar connotations are analyzed combined, such as the keywords that resilience and resiliency. Considering co-occurring keywords network with a large number of nodes, we select the first tab Top 10 per slide as the nodes selection criteria in this section. Figure 6 is the co-occurring keywords network of urban resilience research which consists of 104 nodes and 333 links. Nodes appear the shape of tree rings which have different spectrum of colors corresponding to the years of keywords' occurring. The links also represent different colors according to the years that two keywords occurring together.

Network of Co-Occurring Keywords
Co-occurring keyword network analysis provides the basic information of core research content and helps researchers to track the development trends of research topics at different stages of urban resilience research. In co-occurring keyword analysis process using CiteSpace, keywords are composed of "author keywords (DE)" and "keyword plus (ID)". The author keyword (DE) is derived from the author's article, and the keyword plus (ID) is identified by the journal which the author's articles are published. In the process of co-occurring keywords analysis, the keywords with similar connotations are analyzed combined, such as the keywords that resilience and resiliency. Considering co-occurring keywords network with a large number of nodes, we select the first tab Top 10 per slide as the nodes selection criteria in this section. Figure 6 is the co-occurring keywords network of urban resilience research which consists of 104 nodes and 333 links. Nodes appear the shape of tree rings which have different spectrum of colors corresponding to the years of keywords' occurring. The links also represent different colors according to the years that two keywords occurring together.  Figure 6 shows that "city OR cities OR urban'' with a frequency of 112 is the most important keywords. Other keywords with high frequency are "resilience'' (111), "adaptation'' (65), "climate change" (62), "management" (41), "biodiversity" (35), "vulnerability" (33), "systems" (28), "ecosystem services" (28), "communities" (28), "sustainability" (26), "risk" (24), "social-ecological systems" (22) and "governance" (21). The main keywords can be divided into four types. The  Figure 6 shows that "city OR cities OR urban" with a frequency of 112 is the most important keywords. Other keywords with high frequency are "resilience" (111), "adaptation" (65), "climate change" (62), "management" (41), "biodiversity" (35), "vulnerability" (33), "systems" (28), "ecosystem services" (28), "communities" (28), "sustainability" (26), "risk" (24), "social-ecological systems" (22) and "governance" (21). The main keywords can be divided into four types. The keywords "city OR cities OR urban", "social-ecological systems", "systems", "communities" and "ecosystem services" represent the research objects of urban resilience. The keywords "resilience", "climate change" and "risk" represent the research backgrounds of urban resilience, which means that urban resilience is mainly researched on the background of environmental uncertainty. The keywords "adaptation", "vulnerability" and "biodiversity" represent the research contents of urban resilience, which refer to adaptive theory, vulnerability analysis and biodiversity characteristics. The keywords "management", "sustainability" and "governance" represent that the goals of urban resilience research are achieving sustainable development through the necessary management measures.
Some keywords with close connection relationships can be identified in Figure 6. These keywords include "resilience", "management", "vulnerability", "biodiversity" and "conservation". Considering the betweenness centrality of nodes, the nodes of keywords with close connection relationships also have high betweenness centrality in co-occurring keywords network. The keywords "resilience" has the highest betweenness centrality which is 0.23, followed by "biodiversity" (0.16), "management" (0.11), "vulnerability" (0.11) and "conservation" (0.1). These keywords are the major nodes linking the urban resilience research in different phases and play important roles in the development of urban resilience research.
The keywords' evolution trends of urban resilience research are show in Table 3. The evolution trends timeline keywords is divided into four periods according to the frequency of keywords, these four periods are 1993-2004, 2005-2010, 2011-2013 and 2014-2016. Table 3 shows the evolution trends of major focuses on urban resilience research between 1993 and 2016. As shown in Table 3, urban resilience research mainly focused on a few topics between 1993 and 2004, such as biodiversity, environment conservation, urban systems and management measures to urban resilience. Some new topics begin to appear in urban resilience research between 2005-2010, such as adaptive theory, vulnerability analysis, ecosystem services, urbanization and land use. During 2005-2010, Resilience Alliance begins to research urban resilience from the perspective of social-ecological systems. Urban resilience started to get attention in the USA between 2011 and 2013. At the same time, researchers continue to be concerned about the topics of social-ecological systems, climate change and adaptation during this period. Each topic keep the state of prosperity between 2014-2016, management measures and biodiversity get great attentions to the urban resilience researchers with the unprecedented urbanization process and the growing number of natural disasters.

Co-Citation Analysis
Co-citation analysis provides a unique insight into the structure and dynamics of the scientific paradigm. As a way to measure the proximity between documents, co-citation analysis is a good way to show the relationship between source articles in a dataset and corresponding citations in a wide range of external reference records [92]. Co-citation analysis includes co-citation analysis of journals, co-citation analysis by authors, and co-citation analysis based on citations [93]. In addition to the above three co-citation analysis, cluster analysis is also used to detect and identify changes in development trends in different periods, and to clarify the relationship between important turning points and urban resilience research trends.

Journal Co-Citation Network
The function of retrieval records analysis is used to indicate the major journals source for urban resilience research. Table 4 summarizes the analysis results of journals source, the top 11 major journals source for urban resilience research is indicated identified by 355 retrieval records from 39 journals related to urban studies in the WOS core collection database.  Landscape and Urban Planning also published the first article on urban resilience research among the 355 retrieved records from 39 journals related to urban studies in the WOS core collection database [91].
The 355 retrieval records cite references from different journals as the knowledge foundation of urban resilience research. In this section, CiteSpace is used to generate journal co-citation network. Considering journal co-citation network with a large number of nodes, we select the first tab Top 10 per slide as the nodes selection criteria in this section. The journal co-citation network is used to analyze and evaluate the impact and co-citation models of cited journals. The journal co-citation network is shown in Figure 7. The size of each node in journal co-citation network represents the co-citation frequency of corresponding journal. The influence of cited journals is primarily assessed by its citation frequency. Figure 7 shows that the journal co-citation network includes 125 nodes and 437 links derived from 1084 cited journals. Landscape and Urban Planning is the most important cited journals with the highest cited frequency (with 100 citations), which is also the journal with the highest number of publications for urban resilience research. Other journals with high cited frequency are Urban Studies According to the co-citation frequency, most of journals with high influence are journals from among the 39 journals related to urban studies in the WOS core collection database, including Landscape and Urban Planning, Urban Studies, Journal of the American Planning Association, Environment and Urbanization, International Journal of Urban and Regional Research, Cities and Habitat International. These journals all belong to the top 11 journals source listed in Table 4. However, some journals have fewer cited frequency and show lower influences than the top 11 journals listed in Table 4, such as European Planning Studies, Urban Forestry & Urban Greening, Education and Urban Society and Urban Education. Science and Nature are also journals with high influence on urban resilience research, which means that urban resilience has caused widespread concern in international academic fields. Other journals with high influence focus on general resilience research, such as Global Environmental Change and Ecology and Society. fewer cited frequency and show lower influences than the top 11 journals listed in Table 4, such as European Planning Studies, Urban Forestry & Urban Greening, Education and Urban Society and Urban Education. Science and Nature are also journals with high influence on urban resilience research, which means that urban resilience has caused widespread concern in international academic fields. Other journals with high influence focus on general resilience research, such as Global Environmental Change and Ecology and Society.

Author Co-Citation Network
Author co-citation analysis is an efficient method to measure relationships and connections between authors and to describe mainstream, or at least the leading edge, in the collaborative structure of urban resilience research. In this section, we identify and analyze the evolutionary trends of authors whose scientific papers are cited in the same paper and community of science. Consider a co-citation network with multiple nodes, selecting the first ten tabs of the first slide as the node selection criteria in this section. The author co-citation network is shown in Figure 8, which contains 173 author nodes and 412 co-citation links. In author co-citation network, the size of each node reflects co-citation frequency between authors, and the link represents an indirect collaborative relationship between authors. Some scientific communities have significant co-citation relationships in author cocitation network.
According to the analysis results of author co-citation network, we identify and analysis the authors with highly cited frequency and their countries source. Figure 8 shows that the author with the highest cited frequency is Carl Folke (52 citations, Sweden), followed by Crawford Stanley Some journals with close connection relationships can be identified in Figure 7. These journals include Landscape and Urban Planning, Urban Studies, Environment and Urbanization, International Journal of Urban and Regional Research and Nature. These journals also have high betweenness centrality in journal co-citation network. Environment and Urbanization has the highest betweenness centrality which is 0.34, followed by Urban Studies

Author Co-Citation Network
Author co-citation analysis is an efficient method to measure relationships and connections between authors and to describe mainstream, or at least the leading edge, in the collaborative structure of urban resilience research. In this section, we identify and analyze the evolutionary trends of authors whose scientific papers are cited in the same paper and community of science. Consider a co-citation network with multiple nodes, selecting the first ten tabs of the first slide as the node selection criteria in this section. The author co-citation network is shown in Figure 8, which contains 173 author nodes and 412 co-citation links. In author co-citation network, the size of each node reflects co-citation frequency between authors, and the link represents an indirect collaborative relationship between authors. Some scientific communities have significant co-citation relationships in author co-citation network.
According to the analysis results of author co-citation network, we identify and analysis the authors with highly cited frequency and their countries source. Figure 8 shows that the author with the highest cited frequency is Carl Folke (52 citations, Sweden), followed by Crawford Stanley Holling  (7), Sweden (2), Canada (2) and Australia (1). The distribution of highly cited authors' countries indicates that urban resilience research has become a hot topic in major countries around the world. As the highest cited author, Carl Folke, Professor at the Stockholm Resilience Centre at Stockholm University, focuses on the role that social-ecological systems at different scales play in social and economic development, and how to govern and manage integrated social-ecological systems towards resilience [94]. Henrik Ernstson also comes from Stockholm University and studies resilience under the background of climate change [33], which manifests that Stockholm University is an important research center for resilience. Crawford Stanley Holling is the pioneer of resilience science and first proposed the concept of resilience in the social and ecology domain [1]. Urban resilience researchers hold the largest proportion of top 20 authors with highly cited frequency (6/20); these authors include William Neil Adger, David Harvey, Philip R. Berke, Jon Coaffee, Erik Swyngedouw and Marina Alberti.

Document Co-Citation Network
Document co-citation analysis is used to measure the dependency relationships of previous urban resilience research. The document co-citation analysis assumes that the two documents cited in the same article are similar from a theoretical point of view. Documents with similarities are commonly cited together, and the similarities of these documents depend on the frequency of cocitations. This method reveals the structure of academic research and predicts future research directions. Clusters are generated during the process of document co-citation analysis and used for evolution trends analysis.
In the process of generating a document co-citation network, the node selection criteria include three sets of thresholds: citation threshold (c), co-citation threshold (cc), and co-citation coefficient As the highest cited author, Carl Folke, Professor at the Stockholm Resilience Centre at Stockholm University, focuses on the role that social-ecological systems at different scales play in social and economic development, and how to govern and manage integrated social-ecological systems towards resilience [94]. Henrik Ernstson also comes from Stockholm University and studies resilience under the background of climate change [33], which manifests that Stockholm University is an important research center for resilience. Crawford Stanley Holling is the pioneer of resilience science and first proposed the concept of resilience in the social and ecology domain [1]. Urban resilience researchers hold the largest proportion of top 20 authors with highly cited frequency (6/20); these authors include William Neil Adger, David Harvey, Philip R. Berke, Jon Coaffee, Erik Swyngedouw and Marina Alberti.

Document Co-Citation Network
Document co-citation analysis is used to measure the dependency relationships of previous urban resilience research. The document co-citation analysis assumes that the two documents cited in the same article are similar from a theoretical point of view. Documents with similarities are commonly cited together, and the similarities of these documents depend on the frequency of co-citations. This method reveals the structure of academic research and predicts future research directions. Clusters are generated during the process of document co-citation analysis and used for evolution trends analysis.
In the process of generating a document co-citation network, the node selection criteria include three sets of thresholds: citation threshold (c), co-citation threshold (cc), and co-citation coefficient threshold (ccv). Referencing to previous research on document co-citation by Chen in 2006 [80], these three sets of threshold levels are set as follows: (2, 1, 10), (3, 1, 0), and (3, 2, 10). The time span of 24-years between 1993 and 2016 is divided into twenty four 1-year time slices. Table 5 summarizes the construction configuration for document co-citation network.  There is no article related urban resilience research in three 1-year slices of this study dataset, these three 1-year slices are 1994-1994, 1998-1998 and 1999-1999. Thus, no co-citation document exists in above three 1-year slices. Document co-citation network begins to appear the network nodes and links together after 2000. The numbers of nodes and links in the last four years account for the most of total nodes and links' numbers, which manifests that the document co-citation network in this study can reflects the emerging development trends of urban resilience research.

1-Year
This study generates the document co-citation network based on 17,283 references cited by 355 retrieval records. The clusters of document co-citation network are identified by labels and extract noun phrases from the title terms (T) of documents. Each node represents a document for urban resilience studies, with the label showing the author's name and year of publication. Each link between nodes reflects the co-citation relationships of two documents in the same article. The document co-citation network is shown in Figure 9 and includes 73 documents' nodes and 306 links. Figure 9 illustrates this document co-citation network comprising five representative clusters with closely located nodes. The labels for these five clusters are generated by the inverted document frequency (tf * idf) algorithm. The modularity Q value of 0.4518 is close to 0.5, which means that the division of clusters is relatively reasonable and can be accepted. The average silhouette value is 0.5147, which indicates that the homogeneity of these clusters is at a reasonable level.
In Figure 9, the document co-citation network is divided into different clusters which are described by different colors. The nodes with close co-citation relationship are marked the same color and concentrated in one cluster which has the same background color as the nodes. The size of the cluster represents the number of nodes in each cluster. As shown in Figure 9, cluster #0 is larger than other clusters and has 24 nodes, and there is a close co-occurrence relationship between the nodes of cluster #0 in 2005. The links in cluster #0 have four kinds of colors from light yellow to deep red, which indicates that the nodes of cluster #0 close co-citation relationships between 2011 and 2016. document co-citation network is shown in Figure 9 and includes 73 documents' nodes and 306 links. Figure 9 illustrates this document co-citation network comprising five representative clusters with closely located nodes. The labels for these five clusters are generated by the inverted document frequency (tf * idf) algorithm. The modularity Q value of 0.4518 is close to 0.5, which means that the division of clusters is relatively reasonable and can be accepted. The average silhouette value is 0.5147, which indicates that the homogeneity of these clusters is at a reasonable level. Figure 9. Cluster view of document co-citation network.
In Figure 9, the document co-citation network is divided into different clusters which are described by different colors. The nodes with close co-citation relationship are marked the same color and concentrated in one cluster which has the same background color as the nodes. The size of the cluster represents the number of nodes in each cluster. As shown in Figure 9, cluster #0 is larger than other clusters and has 24 nodes, and there is a close co-occurrence relationship between the nodes of cluster #0 in 2005. The links in cluster #0 have four kinds of colors from light yellow to deep red, which indicates that the nodes of cluster #0 close co-citation relationships between 2011 and 2016. The clusters #1 and #2 have 15 numbers of nodes and forms in 2005 and 2006. The newest cluster is cluster #3 with 12 nodes which forms in 2007. Although the cluster #4 forms in 2003, it only contains 7 nodes. Cluster #1 has the most connections with clusters #2 and #3. The most frequently referenced document in cluster #0 is Folke (2006)#1, which connects cluster #0 and cluster #1 2, #3 and #4 [94].
We used CiteSpace to analyze 17,283 references cited in 355 retrieval records from 39 journals related to urban studies in WOS core collection database city. Table 6 summarizes the top 19 commonly cited references. In the process of selecting top 19 co-cited documents, we first select 21 documents that have been co-cited more than 10 times. Three documents of between centrality close to 0 are deleted from high co-cited documents. Another document with high betweenness centrality is also selected as high co-cited document. Finally, this study selects 19 documents as high co-cited documents. Table 4 provides detailed information of the top 19 co-cited documents.
The top 19 co-cited documents is presented in Table 4 can be divided into four topics: the basic concepts and theories of resilience (8/19), the conceptual models and pathways of urban resilience (6/19), the adaptive capacity of social-ecological systems (3/19) and resilience analysis in the scenarios of climate change and natural hazards (2/19). As the research foundation of urban resilience, the documents related to the basic concepts and theories of resilience accounts for the largest proportion of top 19 co-cited documents. All these 8 documents belong to cluster #0 which includes No. 1, No. 2,No. 5,No. 6,No. 8,No. 10,No. 11 and No. 15. The concept of "resilience" is first proposed in the social and ecology domain [1]. Human-natural systems are abstracted into a theoretical resilience We used CiteSpace to analyze 17,283 references cited in 355 retrieval records from 39 journals related to urban studies in WOS core collection database city. Table 6 summarizes the top 19 commonly cited references. In the process of selecting top 19 co-cited documents, we first select 21 documents that have been co-cited more than 10 times. Three documents of between centrality close to 0 are deleted from high co-cited documents. Another document with high betweenness centrality is also selected as high co-cited document. Finally, this study selects 19 documents as high co-cited documents. Table 4 provides detailed information of the top 19 co-cited documents.
The top 19 co-cited documents is presented in Table 4 can be divided into four topics: the basic concepts and theories of resilience (8/19), the conceptual models and pathways of urban resilience (6/19), the adaptive capacity of social-ecological systems (3/19) and resilience analysis in the scenarios of climate change and natural hazards (2/19). As the research foundation of urban resilience, the documents related to the basic concepts and theories of resilience accounts for the largest proportion of top 19 co-cited documents. All these 8 documents belong to cluster #0 which includes No. 1, No. 2,No. 5,No. 6,No. 8,No. 10,No. 11 and No. 15. The concept of "resilience" is first proposed in the social and ecology domain [1]. Human-natural systems are abstracted into a theoretical resilience model used to understand some methods that can help researchers develop effective policies for environmental management [95,96]. The resilience approach emphasizes nonlinear dynamics, thresholds and uncertainties, and focus on the effects of dynamic interactions on human-natural systems at spatiotemporal scales [94,97]. As a valuable concept for climate change, resilience is applied to the planning theory and practice of human society and urban society [16,98,99]. Cluster #2 and cluster #3 are combined as the top co-cited documents focuses on the conceptual models and pathways of urban resilience. Urban resilience is a promising new tool to promote further urban design and proposes a conceptual framework for promoting the sustainable development of urban ecosystems and human health [100,101]. Under the background of climate change and energy shortages, Cities present various problems and resilience provides a new way to above challenges in increasing urbanization world [33,102]. Transition is an important pathway of urban resilience and should be considered as an evolutionary approach [103,104]. Cluster #4 consists of 3 top co-cited documents relating to the adaptive capacity of social-ecological systems which strongly focus on the thresholds of change and upgrading the adaptive capacity by appropriate strategies [95][96][97][98][99][100][101][102][103][104][105][106][107]. Cluster #1 only includes 2 top co-cited documents which refer to resilience analysis in the scenarios of climate change and natural hazards; resilience is an important transformation to protect core assets or functions from the risks of climate change [108,109].

Evolution Trends Analysis
In this section, we use representative documents in document co-citation network to identify the main research topics in each cluster. The main research topics in each cluster represent the key focuses of urban resilience research. The academic structure of urban resilience research is described based on the analysis of the main research topics of each cluster, including the relationship between the research content of different clusters. Table 7 summarizes five document co-citation clusters by document co-citation analysis. Each row of Table 7 contains cluster ID, size, silhouette value, cluster label (TFIDF), representative documents in the cluster, the most representative citing document and cluster label (MI). The size of cluster represents the number of co-cited references, and the representative cited references reflect the main research topics of cluster. The cluster label (TFIDF) is the label for the different components in each cluster, and the cluster label (MI) is the combined label for each cluster.
As the largest cluster, Cluster #0 contains 24 documents in document co-citation network, its silhouette value is 0.609. Cluster #0 is labeled as "flood-prone area", the cluster label (MI) is combined as "resilience exploratory analysis" through reviewing relevant citing and co-cited documents. This cluster focuses on the basic theories and methods of resilience; the concept is applied to urban planning and land use analysis. To understand the dynamics of social development, the resilience perspective is increasingly used to develop the adaptive management approach for responding to regional change [89,94]. The most representative document in cluster #0 is Schmidt and Garland [110], which define resilience as the ability to reduce vulnerability and addresses how resilience thinking can assist planners and their communities in disaster-prone regions.
Cluster #1 is the second largest cluster, which contains 15 documents in document co-citation network, its silhouette value is 0.924. Cluster #1 is labeled as "disaster", the cluster label (MI) is combined as "disaster resilience" through reviewing relevant citing and co-cited documents. The top 2 representative documents in cluster #1 are Blaikie et al. [108] and Pelling [109]. These researchers analyzed the impact of climate change on human development and used the concept of resilience to guide human society's vulnerability to natural disasters. The most representative document in cluster #1 is Nguyen and Salvesen, which reveals that culture is essential for an effective disaster recovery process and more appropriate and effective disaster recovery plans can be developed by analyzing resilience prior to a disaster [111]. Cluster #2 contains 15 documents in its document co-citation network; its silhouette value is 0.507. Cluster #2 is labeled as "residential neighborhood", the cluster label (MI) is combined as "urban resilience" through reviewing relevant citing and co-cited documents. The most representative documents in cluster #2 are published by Bolunda and Hunhammar [112] and Pickett et al. [113]. These researchers have proposed an emerging urban resilience framework that considers ecosystem services having significant impacts on the quality of life in urban areas. The representative document in cluster #2 is Hunter and Brown [114], in which the potential value of social contagion has been discussed as a mechanism to expand sustainable behavior that supports ecological resilience in urban areas.
Cluster #3 contains 12 documents in document co-citation network, its silhouette value is 0.736. It is labeled as "co-existence", the cluster label (MI) is combined as "urban resilience practice". The most representative documents in cluster #3 are published by Simmie and Martin [104] and Evans [115]. These researchers focus on adaptive cycle theory and make assumptions about urban and regional resilience. The representative citing document of cluster #3 is that of Ozuduru et al. [116], which think that the sustainability of cities depends on the viability of city center and the resilience of shopping district.
Cluster #4 containins 7 documents in document co-citation network, its silhouette value is 0.827. It is labeled as "ecological wisdom", the cluster label (MI) is combined as "social-ecological systems". The most representative documents in cluster #4 are published by Berkes et al. [105] and Folke et al. [107]. These documents research urban resilience from the perspective of social-ecological systems and investigate the reorganization of urban systems during periods of abrupt change. The representative citing document of cluster #4 is that of Patten [117], in which ecological wisdom as a comprehensive management approach play an important role in creating sustainable and resilient urban and natural ecosystems.
To explore the evolution trends of urban resilience research, some important documents with strong citation bursts are selected from the document co-citation network. These documents provide high potential guides for future urban resilience research. The development and evolution trends of urban resilience research are identified by analyzing these important documents with strong citation bursts. Future directions for urban resilience research can be forecasted and further explored by analyzing the existing major research topics. We think that the topics with significant increase in a short time will represent a developing trend in future. This study selects documents and topics with strong citation bursts in recent years to reflect the newest trends of urban resilience research. We used CiteSpace to generate a summary list of documents associated with the referenced burst. Table 8 shows which document has the strongest reference burst and which period has the strongest burst occurrence.   16,94], the adaptation of urban resilience [109,118], urban resilience model [16,100], the case studies of urban resilience [100,119], urban resilience and vulnerability analysis [109] and urban social-ecological systems [1,94].

Discussion
Urban resilience gained great popularity in both academia and industry with the unprecedented urbanization process. Great efforts have been made to enhance urban resilience in the process of urbanization. The concept of resilience has been applied to enhance cities' ability to resist disasters. Urban resilience is an important research topic in urban resilience research. Meerow et al. summarized different definitions from 25 previous publications, the results manifest that urban resilience is an inconsistent and ambiguous concept [10]. Because numerous disciplines engage in the field of urban resilience research, Da Silva et al. think that there are great challenges associated with defining and characterizing "urban" and "resilience" separately [28]. These conceptual inconsistencies make it difficult to apply some resilience metrics to many scenarios. The results of this study also show that the definition of urban resilience is still an important research topic in future urban resilience research. A common definition of urban resilience should be proposed in future urban resilience research. Some scholars provide a common definition of urban resilience. Meerow et al. for instance, defines urban resilience as the ability of an urban system and its constituent socio-ecological and socio-technical networks across temporal and spatial scales to maintain or rapidly return to desired functions in the face of a disturbance, to adapt to changes, and to quickly transform systems that limit current or future adaptive capacity [10].
Scientometric analysis becomes an important research method for reviewing research status in different academic domains. Scientometric methods include qualitative, quantitative and computational approaches, which focus on institutional productivity comparisons, institutional research rankings, journal rankings establishing faculty productivity and tenure standards, assessing the influence of top scholarly articles, and developing profiles of top authors and institutions in terms of research performance [72][73][74][75]. Previous similar studies mainly use a single scientometric indicator to analyze the research status of a discipline [76]. These research results only reflect the research status of the discipline from one specific aspect, such as co-authors analysis only analyze authors and their institutions, research social structures and collaborative networks [77]. Thus, the research results of single scientometric indicator cannot reveal the research status of a discipline. This study adopts three types of scientometric indicators, including co-author analysis, co-word analysis and co-citation analysis, which show influential authors, primary countries and research institutions, core categories and topics of focus, core journals, and evolution trends on urban resilience research. The research results reveal the emerging evolutionary trends and a sufficiently large and high-quality body of research that accurately reflects the global picture of urban resilience research.
Co-citation analysis is used to analyze the academic structure of urban resilience in previous studies. Fröhlich and Hassink use the page rank indicator to analyze the citation network of regional resilience, and the results manifest that "Urban Ecology and Policies" is the largest community in co-citation network [66]. Resilience on city-scales is the most important research topic which concerns environmental shocks and hazards, as well as policy about urban or regional governance, as well as resilience on an individual level with a focus on psychology and society. Fröhlich and Hassink identify common keywords, including "climate change/risks/disaster(s), governance/community/social capital", as well as "adaptive capacity/regional adaptation/sustainability", but do not give the applicable fields of these keywords. In the section "Network of Co-occurring Keywords", we divide the keywords of urban resilience research into four types and analyze the connotations of these keywords. The keywords "city OR cities OR urban", "social-ecological systems", "systems", "communities" and "ecosystem services" represent the research objects of urban resilience. The keywords "resilience", "climate change" and "risk" represent the research backgrounds of urban resilience, suggesting that urban resilience mainly studies the background of environmental uncertainty. The keywords "adaptation", "vulnerability" and "biodiversity" represent the research contents of urban resilience, which refer to adaptive theory, vulnerability analysis and biodiversity characteristics. The keywords "management", "sustainability" and "governance" represent that the goal of urban resilience research is to achieve sustainable development by taking necessary management measures.
Fröhlich and Hassink offer five key urban resilience documents by qualitative analysis, these five key documents include Pike et al. [98], Christopherson et al.
[120], Hassink [121], Pendall et al. [99] and Folke [94]. We summarize the top 19 co-cited documents in the document co-citation network of urban resilience. These top 19 co-cited documents already include the documents Pike et al. [98], Pendall et al. [99] and Folke [94]. The documents Christopherson et al. [120] and Hassink [121] present a significant regional resilience perspective and do not be included in top 19 co-cited documents. We divide the top 19 co-cited documents into four topics: the basic concepts and theories of resilience (8/19), the conceptual models and pathways of urban resilience (6/19), the adaptive capacity of social-ecological systems (3/19) and resilience analysis in the scenarios of climate change and natural hazards (2/19). Compared with Fröhlich and Hassink's paper, the most significant contribution of this study is to reveal the latest trends of urban resilience research. This study summarizes the top seven documents with strong citation dramatic increase between 2011 and 2016. These seven documents reflect the newest trends of urban resilience: the definition of urban resilience, the adaptation of urban resilience, urban resilience model, the case studies of urban resilience, urban resilience and vulnerability analysis and urban social-ecological systems.

Conclusions
Urban resilience has become a very popular area of research among academics and practitioners with the unprecedented urbanization. Great efforts have been made to enhance urban resilience in the process of urbanization. This study explores the emerging trends of urban resilience research by scientometric analysis from 1993 to 2016. Some highlights in this study are listed as follows: (1) Identifying the lead authors with significant contributions and influences based on co-authorship and author co-citation analysis. The results indicate that authors with a large number of publications do not necessarily have significant impacts on urban resilience research. Instead, some less productive authors maybe have larger influences on urban resilience research. (2) Discovering the evolutionary trend of geographic distribution on urban resilience research by analyzing co-authors' countries. The result shows that urban resilience research is conducted mainly in countries with high urbanization levels or in the process of urbanization (e.g., USA, England, Australia, Canada, China and Sweden). As the biggest country in the process of urbanization, China has close collaborative relationships with most countries of high publications output, including the US, England, Australia and Canada. The results show that document co-citation network includes five typical regarding following themes: the basic concepts and theories of resilience, the conceptual models and pathways of urban resilience, the adaptive capacity of social-ecological systems, and resilience analysis in the scenarios of climate change and natural hazards. The emerging research trends are summarized by representative documents references with greatest citation increase in each cluster. These emerging research trends in urban resilience contain urban resilience definition, urban resilience