Green Human Resource Management: A Bibliometric Analysis

: Despite the growing attention which green human resource management (GHRM) has been gaining, the ﬁeld is still remarkably young, and conceptual clarity is yet to be attained. A bibliometric review of GHRM from 2010 to 2020 was conducted on the Web of Science and analyzed using the VOSviewer software package, version 1.6.16 developed by Van Eck and Waltman (Leiden, The Netherlands). The results show the exponential growth of this topic, although there seems to be no consensus regarding its deﬁnition, conceptualization and measurement. Concerning its conceptual development, GHRM seems to currently ﬁt the second stage of development (evaluation and augmentation) of Reichers and Schneider’s three-stage model of the evolution of constructs. Future research seems to point in the direction of establishing the barriers separating GHRM from other human resource management topics, deﬁning green practices and determining the antecedents and consequences of GHRM.


Introduction
The economic growth of the last two centuries has represented prosperity and longevity for humanity. However, the impact of various industrial practices, reflected in pollution and the accumulation of waste and residues, is increasingly evident (IPCC 2018;Jackson et al. 2011). Climate change represents a new set of challenges, altering competitive and regulatory spheres, indirectly through global policy efforts to curb carbon emissions, and directly as a means of combating the unravelling consequences of extreme weather events (IPCC 2018). By invoking unprecedented changes in socio-economic systems, climate change makes it urgent for society to prepare and adapt in the face of new contingencies (Winn et al. 2011). With increasing regulatory and public pressure, more and more business entities are proactively implementing practices and strategic changes, such as investing in low-CO 2 technologies and renewable energy, in order to transition to environmental sustainability (González-Benito and González-Benito 2006;IPCC 2018).
The multidisciplinarity required to reduce organizational environmental footprints challenges the vision that this theme is restricted to environmental sciences, causing new fields of study to emerge in the areas of social research and management (Shrivastava and Berger 2010). In this sense, several related topics have been increasingly emphasized: corporate social responsibility (CSR), environmental management (EM), business ethics, socially responsible consumption and sustainability strategies, among others (Schuler et al. 2017;Shrivastava and Berger 2010).
One topic increasingly referred to in the literature in addressing environmental challenges is green human resource management (GHRM) (Ari et al. 2020;Renwick et al. 2016; The bibliometric analysis will be supported in this work by the visualization of similarities (VOS) mapping technique of the VOSviewer software package. Thus, the main objectives of this study are: (1) to identify literature streams; (2) to map the topics studied; (3) to observe and analyze the temporal evolution of the construct; and (4) to identify the stage of development of the construct as a useful tool for the professional community and not just an academic tool for bibliometricians.
The next section of this paper is the method where the steps required to collect and analyze the data are discussed. Then, the results section is divided into the data collected from the keyword search and from the bibliometric maps, describing the results obtained from both methods. Finally, the conclusion section reinforces the objectives of the paper, the main findings as well as the limitations of the study, and suggests future research directions on GHRM.

Method
To meet the stated objectives, a sample of publications from the Web of Science (WOS) produced between 2010 and 2020 was used. The WOS was selected because it brings together the journals with the highest impact in the social sciences (Crossan and Guatto 1996), namely in the SSCI (Social Sciences Citation Index) and the ESCI (Emerging Sources Citation Index).
The first step consisted of a descriptive survey of all publications of the last decade (2010-2020) of the WOS, through a keyword search with the term "green human resource* management". This was the keyword with the highest number of associated publications, by comparison with the keywords "GHRM", "green HRM" or "green hr", i.e., the sample of these last three keywords was already included in the chosen keyword "green human resource* management" sample. The choice of the selected time period (2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020) was reinforced by the fact that the first article in WOS with the selected keyword dates back to 2011 (Jackson et al. 2011). The search was restricted to SSCI and ESCI, since the former belongs to the WOS Core Collections and the latter gathers journals under final evaluation to integrate them. The search engine criterion was "TOPIC", allowing for recognizing the keyword in the (1) title, (2) abstract, (3) author's keywords and (4) "keyword plus", a tool that expands the search engine by including articles in the WOS with a significant frequency of the term being in the titles of the bibliographic references.
Then, in the second step, the sample was analyzed using VOSviewer. The visualization of bibliometric networks, often referred to as the "mapping of science", constitutes an advanced tool in the analysis of bibliometric networks. Visualization software packages, such as VOSviewer, represent an advance in the analysis of these networks by allowing a large volume of data to be processed (Van Eck and Waltman 2014). The spatial representation of bibliometric networks includes different types of analysis, from keyword co-occurrence networks to co-citation and co-authorship networks Waltman 2010, 2014). The VOSviewer maps are considered distance bibliometric maps, that is, the distance between two items reflects the strength of their relationship: a small distance indicates a strong relationship and vice versa (Van Eck and Waltman 2010). In this way, the identification of groups of related items is facilitated (Van Eck and Waltman 2010).
Similarly, to the procedure adopted by Bellucci et al. (2020), the bibliometric maps used in this study were from the keyword co-occurrence network, which depicts the proximity of the keywords in the sample, as well as bibliographic coupling, representing the relationship between the publications in the sample based on the bibliographic references they share. While the first is aimed at mapping potential lines of research in GHRM, the second is intended to identify the main topics that build them. As we will see later in the results, the two maps complement each other, allowing us to perform a joint analysis.
The analysis of the keyword co-occurrence network determines the frequency with which terms are repeated in the sample. This analysis results in a map illustrating the interrelationships between the conceptual structures of the sample under study (Zupic andČater 2015). The size of the circle and the links are proportional to the frequency in the sample and the strength of the keywords' relationship with others, respectively. More central terms show more links with different clusters. The bibliographic coupling of publications measures the similarity between two articles from the references they share (Kessler 1963). The smaller the distance between two items, the more references they share and, therefore, the more content they share (Li and Hale 2015). The size of the spheres is proportional to the citations received in WOS. Thus, the intention is to identify the intellectual structure of GHRM. Finally, to achieve the last objective of this paper-identify the development stage of the GHRM construct-the three-stage model of the evolution of constructs, developed by Reichers and Schneider (1990), was adopted as a framework. In line with this model, constructs, from their introduction to their acceptance into mainstream literature, evolve according to a predictable and developmental sequence that follows three main phases: (1) introduction and elaboration; (2) evaluation and augmentation; and (3) consolidation and accommodation (Reichers and Schneider (1990)). In the discussion section, the three stages of the model are described, as is the one that fits into the current stage of development of the GHRM field.
The step-by-step method of the current paper is described in Figure 1.

Results
The results are divided into two sections: firstly, a characterization of the keyword search sample is presented, and secondly, the VOSviewer bibliometric maps are highlighted.

1st step -Selection of database:
Web of Science.
4th step -Selection of software: VOSviewer: bibliometric network analysis and mapping .

7th step: Analysis and results:
15 papers analyzed from each cluster.
8th step: Discussion and conclusion: Identification of the GHRM current evolution stage.

Results
The results are divided into two sections: firstly, a characterization of the keyword search sample is presented, and secondly, the VOSviewer bibliometric maps are highlighted.

Keyword Search
The keyword search conducted on 19 January 2021 resulted in a total of 143 publications. Of the 143 publications, 127 (88.81%) are empirical articles, 14 (9.79%) are literature reviews, one is (0.69%) a correction article and one (0.69%) an editorial.

Evolution of the Number of Citations
The 143 publications present a total of 3445 citations (see Figure 3). Of note is the evolution curve, which depicts an almost non-existent scientific dialogue at the beginning of the decade, transitioning to a marked growth from approximately 2017. .

Authors Affiliation
The scientific production related to GHRM is mainly represented by China, Malaysia and England (Table 1). Please note that the table only includes the ten most active authors' affiliations.

Evolution of the Number of Citations
The 143 publications present a total of 3445 citations (see Figure 3). Of note is the evolution curve, which depicts an almost non-existent scientific dialogue at the beginning of the decade, transitioning to a marked growth from approximately 2017.

Evolution of the Number of Citations
The 143 publications present a total of 3445 citations (see Figure 3). Of note is the evolution curve, which depicts an almost non-existent scientific dialogue at the beginning of the decade, transitioning to a marked growth from approximately 2017. .

Authors Affiliation
The scientific production related to GHRM is mainly represented by China, Malaysia and England (Table 1). Please note that the table only includes the ten most active authors' affiliations.

Authors Affiliation
The scientific production related to GHRM is mainly represented by China, Malaysia and England (Table 1). Please note that the table only includes the ten most active authors' affiliations. According to WoS criteria "Highly cited papers", these are the most-cited publications ( Table 2).

Bibliometric Maps
Next, the VOSviewer bibliometric maps resulting from the bibliographic reference coupling analysis (see Figure 4 and Tables A1-A4 in Appendix A) and the keyword cooccurrence network analysis (see Figure 5 and Table A5 in Appendix B) are presented. The 143 papers published between 2010 and 2020 were subjected to this analysis.

Bibliometric Maps
Next, the VOSviewer bibliometric maps resulting from the bibliographic reference coupling analysis (see Figure 4 and Tables A1-A4 in Appendix A) and the keyword cooccurrence network analysis (see Figure 5 and Table A5 in Appendix B) are presented. The 143 papers published between 2010 and 2020 were subjected to this analysis.  Both maps were submitted to chronological processing between 2010 and 2020 (see Figures 6 and 7). The color gradation identifies the average date of co-occurrence of the keywords for the former, and for the latter, it identifies the year of publication. Both maps were submitted to chronological processing between 2010 and 2020 (see Figures 6 and 7). The color gradation identifies the average date of co-occurrence of the keywords for the former, and for the latter, it identifies the year of publication.   Four distinct clusters emerged in this analysis (yellow, blue, green, red), as shown in Figure 4. These represent publications which are strongly linked through the references they share, thus suggesting different streams of literature (Bellucci et al. 2020). All papers present links to each other, with the exception of the correction article by Guerci et al. (2016). Note that some links are not visible as they are rather weak.

Co-Occurrence of Keywords
Of the 683 keywords identified, 57 are presented, resulting from the standard VOSViewer criterion that selects keywords repeated at least five times in the sample (see Figure 5). Note that this analysis included keywords plus an option provided by VOSViewer, which includes terms that occur frequently in the title of the bibliographical references of the sample. This selection was based on the fact that the resulting bibliometric map is richer in information, thus allowing for a more complete analysis of the sample.

Cluster Analysis
The map of the keyword analysis reflects the map of the bibliographic linkage of references. More specifically, the clusters of the keywords (see Figure 5) correspond to the clusters of the bibliographic linkage of references (see Figure 4). For instance, the keyword "Brazil" in the blue cluster is associated with the articles in the blue cluster of the bibliographic reference coupling map. As we will see below, the four colored clusters (blue, yellow, green and red) represent four distinct research lines.
The identification and analysis of the clusters (see Figures 4 and 5) was performed, similarly to the procedure adopted by Bellucci et al. (2020), with the selection of the 15 most cited articles in each cluster of the bibliographic coupling of references for analysis. Tables A1-A4 (in Appendix A) present the selected articles of each cluster, and Table A5 (in Appendix B) describes the frequency of occurrence of the keywords per cluster. By the reading and the analyses of the 15 most cited articles in each cluster, and by correlating the main themes founded on those articles with the most prominent keywords in each cluster, it was possible to identify the literature streams that characterize the clusters.

Blue Cluster: Implementation of GHRM as a Facet of Environmental Management
This cluster gathers 32 articles. The keywords "companies", "environmental management", "implementation", "green management", "strategy", "sustainable operations", "industry" and "organizations" are highlighted: the cluster highlights the implementation of environmental sustainability in HRM in light of environmental management (EM), (Jabbour et al. 2012;Teixeira et al. 2016). EM entails a tool in the pursuit of environmental goals that advocate sustainable development. One of the first empirical contributions reveals that when EM is an emerging goal, HRM practices tend to incorporate this environmental goal (Jabbour et al. 2012), such as training ("green training"), Nejati et al. 2017;Teixeira et al. 2016) or practices concerning GSC (Teixeira et al. 2016). Of note is the highlighting of the keyword "Brazil", since authors, such as Jabbour, have Brazilian companies as the sample of their studies.
Importantly, the cluster includes articles that bring EM closer to employees, highlighting their influence on environmental change. Pinzone et al. (2016) view EM as an organizational change which is dependent on employee support. The authors support the effectiveness of green practices to foster voluntary green behaviors. On the other hand, Obeidat et al. (2020) support the mediating role of GHRM in the relationship between management support, internal environmental orientation and environmental performance. With the intention of expanding the relationship of GHRM with other stakeholders, Guerci et al. (2016) conclude that green practices act as mediators in the relationship between environmental pressures and environmental performance.
Finally, the proximity of this cluster to the green cluster, which is also connected to EM, is highlighted, as we shall see below.

Red Cluster: Consequences for Employees
This cluster, composed of 50 articles, is dedicated to the study of HRM consequences for employees, mainly in the workplace. The attitudinal approach is represented by the terms "organizational citizenship behavior", "commitment", "knowledge", "proenvironmental behavior", "motivation", "employee green behavior" and "empowerment".
The analysis of the impact of green practices on employees is often carried out by performance outcomes. As an example, Kim et al. (2019) highlight that GHRM increases employees' organizational commitment in the hospitality industry, their green behavior and their environmental performance. Focusing on this industry as well, Pham et al. (2019b) observe that GHRM directly influences environmental organizational citizenship behaviors (OCB), i.e., behaviors not required by the organization or not listed as a job requirement and which impact environmental performance.
This cluster also contains multiple articles focusing on GHRM without exploring the environmental performance aspect. For example, the study by Saeed et al. (2019) suggests that green practices increase employees' pro-environmental behaviors through pro-environmental psychological capital. Two terms with high co-occurrence are notably "environmental sustainability" and "corporate social responsibility". The former is highlighted in the sample as the ultimate goal of GHRM (Chaudhary 2020), while the latter is related to GHRM in the sense that this HRM model "motivates employees to participate in an important aspect of CSR, green activities and initiatives" (Shen et al. 2018, p. 595).

Green Cluster: Consequences for Organizations
This cluster groups 26 articles associating organizational-level GHRM outcomes above all, mirrored in the map of keywords by the high occurrence of terms such as "sustainability", "environmental performance", "impact", "financial performance", "firm performance" and "innovation" (see Table A5 in Appendix B). The focus on the consequences of GHRM is particularly dedicated to organizational environmental performance; however, the financial facet is not neglected (Yong et al. 2020). It is relevant to point out that this cluster is also associated with EM, reflected by the proximity of "environmental performance" and "environmental management" to the blue cluster (see Figure 5). The keyword "sustainability" is also operationalized as an organizational outcome (Mousa and Othman 2020).
The empirical approach to GHRM implies measuring its implementation through green practices. Gupta (2018), after identifying the most important green practices through a literature review and consultation with experts, contributed to the field with an organizational environmental performance measurement grid.
From the articles analyzed (see Tables A1-A4 in Appendix A), it was found that there are also publications in this cluster focusing on the employees' attitudinal sphere ("attitudes"). Yong et al. (2019c), for example, suggest that green intellectual capital (GIC) and green relational capital (CRV) influence HRM. According to Chen (2008), green intellectual capital agglomerates the "totality of knowledge, skills, relationships and intangible assets related to environmental protection and green innovation" (p. 277), while green relational capital is the "totality of interactive relationships with customers, suppliers, network members and partners in corporate environmental management and green innovation" (p. 278).
Lastly, the state of the art according to Amrutha and Geetha (2020), which contributes to future directions of research on green practices, stands out. The authors explore a theoretical model that proposes a mediating role of green workplace behaviors in the relationship between HRM and social sustainability.

Yellow Cluster: Reviews, Models and Measurement
Bringing together 35 articles, this cluster is marked by bringing together reviews, models and measures of GHRM. Note the overlap of the blue cluster's "of-the-art" term and proximity to the green cluster, as both also include literature reviews. Of the 15 articles targeted for analysis, there are four literature reviews of GHRM (Ahmad 2015;Jackson et al. 2011;Ren et al. 2018;Renwick et al. 2013) and one literature review of green employee empowerment in GHRM (Tariq et al. 2016). It is also worth mentioning the measurement of GHRM in the work of Tang et al. (2018), in which they contributed to the development and validation of a GHRM measurement instrument.
EM is also relevant in this cluster (see Table A5 in Appendix B), reflected in the proximity and overlap of the keyword "environmental management" in the blue cluster with this one. This is the least represented cluster in the map of keywords (see Figure 5). This is probably due to the fact that its content includes keywords from the other clusters. In other words, literature reviews, by recognizing different perspectives and literature streams, represent a more comprehensive view of the literature field, neutralizing the tendency in the reviews, models and measurement cluster to emphasize certain keywords associated with specific literature streams.

Evolution over Time
From the chronological bibliometric maps presented in Figures 6 and 7, it can be concluded that most of the keywords are present in more recent publications. The keywords circling 2016 and 2018 approximately are "Brazil", "environmental management" and "implementation". These keywords predominate in the blue cluster, resulting from works by Jabbour (2013, Jabbour and Jabbour (2016), Nejati et al. (2017) and Teixeira et al. (2016). These works connect GHRM to CFV, green product development and the environmental maturity of companies.
In older publications, we observed some references to the yellow cluster, such as the works of Jackson et al. (2011), Renwick et al. (2013Renwick et al. ( , 2016 and Paillé et al. (2014). The bibliographic coupling map allows us to see that the blue and yellow clusters are less recent compared to the red and green clusters. Note that the latter two explore the consequences of GHRM on environmental performance and employees. A temporal transition centered on the organizational system to a more employee-focused approach can also be observed. The most represented keywords between 2019 and 2020 are "responsibility", "attitudes", "motivation", "impact", "workplace", "employee green behavior" and "pro-environmental behavior". It should also be noted that environmental training, despite being a highlighted term in the blue cluster, continues to be represented in more recent publications.

Discussion
By performing bibliometric mapping of the keyword search sample of publications, this study explored the evolution of GHRM during the 2010s, focusing mainly on the investigation of literature streams and the most recurring topics.
Then, the main contributions of the keyword search and VOSviewer are discussed, and finally, the stage of evolution of GHRM will be characterized.
The keyword search allowed the recent boom of the construct to be ascertained, namely from 2018 onwards (see Figures 2 and 3). GHRM is a topic attracting increasing attention from researchers from various scientific domains (e.g., management, environmental engineering and psychology). The fact that it is often described as a strategic element of EM applied to HRM leads to the conclusion that this is already a consolidated aspect in the definition of the construct. Despite this, there is still no consensual definition of GHRM (Benevene and Buonomo 2020; Ren et al. 2018). While, on one hand, there are proposals to redefine the construct suggesting a more holistic consideration that takes into account other facets of sustainability, such as the circular economy, on the other hand, a definition that focuses on the ecological dimension is proposed (Ren et al. 2018).
Regarding the results of the analyses carried out in VOSviewer, we highlight the fact that the map for the bibliographic coupling of references represents an almost perfect circular figure. This suggests that GHRM is a multidisciplinary research area and does not yet present totally clear and differentiated literature streams. Nevertheless, four groups of items related to each other emerged from the analyses, according to the similarity criterion assumed by the software (Van Eck and Waltman 2010, 2014). It was possible to identify a differentiating theme in each of the four clusters: blue (implementation of GHRM as a facet of EM), red (consequences on employees), green (consequences at the organizational level) and yellow (reviews, models and measures). It should be noted, however, that all clusters present some publications that deviate from their main topic. As an example, the article by Moktadir et al. (2019), exploring the antecedents of GHRM, belongs to the green cluster and is mostly devoted to the organizational-level consequences of GHRM. This may be justified by the fact that there is still no representativeness of the study of antecedents in the GHRM literature; conversely, the exploration of GHRM consequences at the organizational and individual levels predominates (Benevene and Buonomo 2020). Although all clusters show some overlap, the yellow cluster is particularly ambiguous. While, on one hand, it brings together literature reviews (Jackson et al. 2011;Ren et al. 2018;Renwick et al. 2013;Tariq et al. 2016) and theoretical developments (Renwick et al. 2016), on the other hand, it also includes quantitative or qualitative empirical studies. These focus on the validation of HRM measurement scales ) on the assessment of implemented green practices (Masri and Jaaron 2017) and on green behaviors in the workplace (Rayner and Morgan 2018) or outside the workplace (Bombiak and Marciniuk-Kluska 2018).
The temporal evolution maps reveal that the publications referring to the red and green clusters are the most recent ones, compared to the blue and yellow clusters. Although the map of keywords does not reveal a very clear temporal transition, it does show that the latest articles tend to focus on the attitudinal dimension, while the "older" terms (e.g., implementation, EM, sustainable operations) represent the initial conception of GHRM, which is more restricted to its implementation in organizations as an environmental management system or tool. Furthermore, it is important to note that by the middle of the decade, authors such as Jabbour, Renwick and Teixeira dominated the literature on GHRM, viewing it as a tool to achieve environmental sustainability. Thus, the efforts made were mainly focused on the analysis and understanding of the benefits of GHRM practices in relation to environmental performance. Contrastingly, authors who have recently stood out belong to a line of research that recognizes GHRM as an end in itself, capable of promoting attitudinal and behavioral changes, which also impact the performance of organizations. Authors such as Yusliza, Pham, Singh and Dumont represent this line of research. Another relevant aspect is the fact that most empirical studies are conducted in emerging economies, such as Malaysia and India. GHRM thus seems to acquire a more evident role in these contexts, since the large-scale industrial sector predominant in these countries requires a huge demand for natural resources and generates high CO 2 emissions. The implementation of GHRM will contribute to natural resource conservation, a reduction in greenhouse gas emissions and protection of the ecosystem (Yong et al. 2019a). Ren et al.'s (2018) GHRM definition, which reinforces HRM as a relevant phenomenon to understand the relationships between the environmental impact of organizational activities and HRM systems, has, in our view, contributed to the delimitation of boundaries of the construct vis-à-vis other HRM specialties. These include Sustainable HRM, which refers to the triple bottom line approach and moves away from the environmental dimension (Ahmad 2015).
The delimitation of boundaries in relation to other HRM specialties, such as Sustainable HRM, seems to be a relevant path for future research. In this sense, it can be seen that the lack of a consensual definition of the construct is mirrored in the plurality of HRM measures. For example, we observe authors who resort to strategic HRM measurement instruments (Paillé et al. 2014). Let us note, in this scope, the contribution of the scale by Tang et al. (2018), focused on the ecological component of HRM practices. However, it will be important in the future to validate the instrument outside the Chinese context.
From our analysis, we could observe that there is no consensual definition within each cluster, but there is a strong tendency to integrate GA into the definition. It is prevalent for clusters to use the definition of Renwick et al. (2013), particularly in the red cluster (Dumont et al. 2017;Kim et al. 2019;Pham et al. 2019bPham et al. , 2020Saeed et al. 2019;Singh et al. 2020). However, in the yellow (Roscoe et al. 2019;Tang et al. 2018), blue (Obeidat et al. 2020;Yu et al. 2020), and green clusters (Longoni et al. 2018), this definition is also used. Based on our study we could not conclude that GHRM practices are exclusively associated with a particular cluster. For example, green recruitment and selection practices are present in the yellow cluster (Masri and Jaaron 2017;Tang et al. 2018;Ren et al. 2018;Renwick et al. 2016;Roscoe et al. 2019), blue cluster Nejati et al. 2017), green cluster (Yong et al. 2020;Yusliza et al. 2017Yusliza et al. , 2019 and red cluster (Saeed et al. 2019;Singh et al. 2020). Other practices, such as green training and development, performance appraisal, reward and compensation, and employee empowerment, are also present in the different clusters. Rather than the green practices associated with specific clusters, we emphasize what appears to be the main differentiator of these clusters: the study of the influence of organizational antecedents and consequents on green practices and GHRM implementation. Concerning the methods used, empirical articles are mainly distributed across the blue, red and green clusters, with quantitative studies prevailing over qualitative studies. Mixed-method papers such as Paillé et al. (2014) from the yellow cluster, Pham et al. (2019b) from the red cluster and Yu et al. (2020) from the blue cluster make use of data from questionnaires and also from interviews. Regarding review articles, based on our literature review, only one bibliometric analysis paper on GHRM was found (Khan and Muktar 2020), and it also uses VOSviewer bibliometric maps.
In agreement with our findings, Benevene and Buonomo (2020) state that most studies are published from 2018 onwards, and they report a geographical trend, namely, an increasing number of publications in emerging countries over time. An interesting aspect highlighted by other papers is the fact that the number of articles addressing GHRM over time, as well as their geographical distribution, have been strongly influenced by regulations on sustainability and green standards across organizations (Amrutha and Geetha 2020; Benevene and Buonomo 2020). Comparing our results with Yong et al.'s (2019a) work, which identifies the streams of the GHRM literature from a sample of publications from 2007 to 2019, we observe that the themes identified in the present analysis are not far from their classification: (1) Reviews, models and concepts; (2) Implementation of GHRM; (3) Determinants of GHRM adoption (at the organizational level); (4) Consequences of GHRM adoption (at the organizational level); and (5) Outcomes of GHRM adoption (at the individual level). The yellow cluster (Reviews, models and measurement) coincides with theme 1 (Reviews, models and concepts), the red cluster (Consequences for employees) with theme 5 (Outcomes of GHRM adoption at the individual level), the green cluster (Consequences for organizations) with theme 4 (Consequences of GHRM adoption at the organizational level) and the blue cluster with theme 2 (Implementation of Green HRM). That leaves, of course, theme 3 (Determinants of GHRM adoption at the organizational level) without a cluster fit. This theme includes the articles presented in the blue cluster (Obeidat et al. 2020;Teixeira et al. 2016) and the green cluster (Yusliza et al. 2017). It is worth noting that the authors analyzed a sample of 70 articles, published in a distinct time span of the present review, which naturally limits our analysis when comparing the two results. Our findings also converge with Khan and Muktar's (2020) review that focused on the period 2008-2020 and used the Scopus database. This convergence is reflected in the focus of empirical studies on emerging economies and on the industrial sector, and it can be seen in the transition from the emphasis on environmental performance resulting from an EM at the organizational level, to an approach that returns importance to the agents of organizations, the employees, reflected in the emergence of studies focused on attitudinal variables.
According to the objectives proposed for this work, based on the literature reviewed and analyzed, and using Reichers and Schneider's (1990) model as a framework, it can be concluded that the GHRM construct is currently at the second stage of development, i.e., the assessment and augmentation stage. While it is true that there are studies focused on its antecedents and consequents, a condition of the first stage of development-introduction and elaboration-by contrast, there are already literature reviews, such as those by Benevene and Buonomo (2020), Ren et al. (2018) and Yong et al. (2019a), focused on the research gaps, both in terms of GHRM's definition and conceptualization, and in terms of its measurement. These authors also suggest future research agendas in order to advance the topic at the level of the theoretical and research field. As the concept does not present an unambiguous definition and there is ambiguity in the distinction between GHRM and other HRM specialties, the contributions and suggestions of various authors, in an attempt to clarify, delimit, distinguish and link the construct to others that are close to it, are very relevant contributions to the development of the GHRM literature. Thus, the current debate on what GHRM is and how it is conceptualized and assessed, in our view, justifies the characterization of the construct at the assessment and augmentation stage. The increasing research focus on the topic of GHRM suggests that the next stage of development of the construct-consolidation and accommodation (Reichers and Schneider 1990)-may be reached in the medium term. In this phase, a clear prevalence of a reduced number of definitions and a predominant way of operationalizing GHRM, as well as the establishment of construct boundaries and the recognition of associated antecedents and consequents, will be observed. The emergence of meta-analyses that portray the results and compare contradictions between works are also expected.

Conclusions and Future Directions
The bibliometric analysis of the literature produced between 2010 and 2020 highlights the importance of GHRM today as an expanding field, particularly from the second half of the decade. The purpose of GHRM is to actively participate in the creation of greener societal paths and respond to environmental and ecological issues from the organizational context. This HRM specialty refers to one of the most important challenges faced by humanity.
By mapping the existing literature streams-implementation of GHRM as a strand of EM, consequences at the organizational level and consequences at the employee leveltracing their contributions and development in the last decade, and identifying the construct development stage, this paper has met its objectives. In short, the results obtained suggest that the implementation of green practices in organizations will be reflected in lower levels of environmental pollution and contamination. Despite the emphasis on the repercussions of GHRM at work, it goes beyond the labor sphere, since green environmental values are transferable from context to context, i.e., green behaviors applied at work are reflected in personal life and vice versa. The fact that the evolution of the construct has been going beyond a concept focused only on EM demonstrates the consideration of employees endowed with unique strategic value and their decisive role in organizational sustainability. The resource-based view (Barney 1991) is a theoretical contribution that points in this direction (Yusliza et al. 2017).
Despite the contributions of this work, it is also important to highlight some limitations and the attempts to overcome them. By using only one database, the WOS, we limited access to the universe of publications on GHRM, and therefore, the information that we extrapolated may contain some biases. However, in the category of bibliometric reviews, we found the work of Khan and Muktar (2020), which used the Scopus database, revealed results which are convergent with ours, a fact that suggests the reliability of the results obtained. In addition, we consider that the present bibliometric review extends the work of Khan and Muktar (2020), insofar as it crosses two similarity maps that allow the lines of research in GHRM to be mapped and the topics that build them to be identified, as it also aims to identify the stage of evolution of the construct. The last limitation refers to the fact that only the 15 most cited articles of each cluster obtained through VOSviewer were analyzed, resulting in a total of 60 articles compared to the 143 of the sample obtained. Although Bellucci et al. (2020) defend the credibility of this method, we feel we should mention this limitation.
Despite the growing interest of the scientific community in studying the topic of GHRM, we found that a high research potential remains. Thus, based on the discussion of the results, we suggest some recommendations for future work. It is important to achieve a more consensual definition of GHRM (Ari et al. 2020) and to develop a more universal and unambiguous language for the construct. Thus, emphasis is placed on the need to study the relationship of GHRM with other related concepts, accessing the different approaches that have been proposed so that commonalities and differentiating aspects of the various constructs can be more accurately observed. We recall something which was already mentioned by Renwick et al. (2013), who produced one of the first systematic reviews of the construct. They highlight GHRM as an approach that allows for going beyond the strict conception of the economic side and includes the wellbeing of the various stakeholders at the heart of HRM. By delimiting boundaries with other HRM specialties, the definition, conceptualization and measurement of GHRM will be more approachable. On the other hand, it is relevant to highlight that the need for further studies on the contextualization of GHRM is important, as it is not only influenced by the most obvious organizational actors, such as leaders and employees, but by multiple stakeholders, such as teams, municipalities, regulators and legislators, which exert pressures and constraints on the implementation of GHRM (Ren et al. 2018). We also highlight the need to delve deeper into the antecedents of GHRM, in particular the antecedents at the employee level (Obeidat et al. 2020). The study of GHRM in developed countries should also receive attention, as they are less represented by current studies focusing on emerging economies.
Pursuit of the analysis of the relationship between GHRM and other organizational features, such as GSC and environmental certification systems, is also an important way forward. In organizational practices, the existence of complex performance agendas and objectives that transcend HRM objectives must be assumed. In this sense, understanding how GHRM relates and responds to the different aspects of the organization will allow us to study the impact of this specific way of managing human resources in real organizational circumstances. Finally, we highlight the fact that artificial intelligence and big data have come to dominate the way we work, and the benefit of the intersection of these areas constitutes an interesting direction in the study of GHRM. Funding: This work is financed by portuguese national funds through FCT-Fundação para a Ciência e Tecnologia, under the projects UIDB/05037/2020 and UIDB/05422/2020.