Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Through 100 years of Ecological Society of America publications: development of ecological research topics and scientific collaborations

Through 100 years of Ecological Society of America publications: development of ecological... IntroductionIntegration and convergence of historical knowledge is becoming increasingly important, and new information is being discovered with retrospective analysis of old data. Since the discipline of ecology includes diverse academic fields (Cullen et al. , Thompson et al. ), there have been limited comprehensive reviews using objective data. Large bodies of text extracted from scientific publications have not been used frequently in ecological reviews. In particular, there have been limited attempts to focus on conceptual terms used in ecology (Cherrett , Lawton , Borrett et al. , Reiners et al. ) or the text itself in scientific publications. Scientific texts, as a historical record (i.e., ecology papers), contain information that reflects the academic consensus of the scientific community. Thus, the use of scientific terms is potentially important in estimating relative interest in scientific concepts within the academic community.Scientometric research (analysis of publications) and culturomics (quantitative analysis of text) have recently emerged as new tools to investigate certain phenomena in a scientific community using quantitative measurements of word frequencies in digital texts (Ladle et al. ). Words used by a community reflect the rooted ideas of its members (Pennebaker et al. ). Extending this concept, the language that scientists use in their papers reflects the current conceptual models or ideas of science. Culturomic tools have already provided new insights into conservation topics and have supported the conservation decision process by providing scientific measurements of public interest in ecological topics (Kim et al. , Do et al. ).Our study quantified the relative use of ecological concepts and research collaborations during the last 100 yr (1920–2015) using text from scientific publications. We performed a comprehensive scientometric analysis using 22,179 publications from journals published by the Ecological Society of America (ESA) since 1920 (Appendix S1). Our objectives were to (1) identify the relative volumes of certain terms in the papers, (2) compare trends in ecological topics and subjects, and (3) evaluate research collaborations in diverse ecological fields.MethodsText corpusWe established an initial article database of ESA journals (Appendix S1: Table S1) using Web of Science (webofscience.com) and Wiley Online Library (esajournals.onlinelibrary.wiley.com). These two academic databases contained all ESA journal publications from 1920 to the present. Each ESA journal was browsed for its full publication period (1920–2015), and publication records for scientific articles were collected semi‐automatically using the reference import function in a bibliography management software program (Zotero Version 4.0.27, Fairfax, Virginia, USA). Collected publication data included title, abstract, keywords, authors, institution or organization of authors, publication year, article type, and list of literature cited in each article. The complete collection constituted the text corpus (body of text) used for analysis. To check the integrity of the text corpus, all publication records were integrated into a single text file. Duplicated records in the database were merged into a single record. Corrections (e.g., errata), news, book reviews, and editorials were not included in the final text corpus. Thus, only peer‐reviewed articles were included. Natural language processing was performed before text data mining. To compile meaningful terms in the text corpus, we applied a linguistic filter (Waltman et al. ) that combined different verb tenses into a single verb and integrated different forms of nouns into a single noun.Text data miningScientometric techniques for word frequency counts and word and author network analyses were used to trace quantitative changes in ecological issues over the last 100 yr. Word frequency counts were used to identify the quantitative frequency of terms used in different time periods. Relative rank frequency of all unigrams (one‐word terms, e.g., ecosystem) and bigrams (two‐word terms, e.g., ecosystem health) was calculated using the text corpus. We applied a binary counting method to calculate the frequency of terms. In the case of binary counting, a term was counted only once per publication. As certain terms were repeatedly used in some papers, a full counting method may have overestimated the occurrence of those terms. Meaningless combinations of prepositions or adverbial phrases were excluded from analysis.Relationships between ecological terms were mapped with a word network using the text corpus. In a word network, an algorithm determines the relative position of each term on a density map by weighting the frequency of its co‐occurrence with other terms. We used VOSviewer 1.6.5 (Leiden University, Leiden, The Netherlands) and Gephi 0.9.1 (The Gephi Consortium, Compiègne, France) with binary counting methods to visualize the network and density map. Synonyms were merged into a single term before analysis. Terms in the word network were grouped into representative network clusters using the VOS clustering algorithm (Waltman et al. ).We divided the study period into five time frames (1920–1939, 1940–1959, 1960–1979, 1980–1999, and 2000–2015) so that the relative use of topic words over time could be compared. To compare the relative use of each term, its frequency of use was normalized by dividing it by the total number of publications within the selected time period. Thus, word frequency was expressed as the relative volume (%) of a term within each decade. A Mann–Kendall trend test was used to assess increasing or decreasing trends in variables (number of publications and relative volume of specific terms) over the study period. Additionally, we screened candidate terms based on their relative volume and trend slope during the study periods. Terms that rapidly increased or decreased over time were selected, and these words were further grouped by either similar topic or complementary pairs.The relative importance and influence of ecological concepts have been estimated based on questionnaire survey methods (Sutherland et al. , Reiners et al. ) and group discussions among experienced scientists. To efficiently utilize our text‐mining results, we compared the relative volume of terms from our text mining with the concept term rankings from previous survey methods. Reiners et al. () have shared the list of concept terms and associated ranking scores from their study in the University of Wyoming Research Data Repository (https://doi.org/10.15786/m2z599). Additionally, there were similar surveys conducted of the members of the British Ecological Society in 1986 (Cherrett ) and again in 2013 (Sutherland et al. ). These concept terms and ranks were integrated into a single database, and we listed overlapping terms from the different survey results. Finally, relationships among the different scores and ranks of concept terms were analyzed with the regression analysis tool in SigmaPlot (version 12.5; Systat Software, Erkrath, Germany).Research collaborationResearch collaboration relationships were analyzed through publication co‐authorship as peer‐reviewed publications are one of the core forms of media in the scientific community. We assumed that projects or other collaborative endeavors would be represented by researchers named as co‐authors on published articles. We organized author and organization data from the text corpus as described above. Based on collaborative research units (i.e., a co‐author group in a paper), an organizational research network was established and network centrality indices (e.g., eigenvector centrality, harmonic centrality, and closeness centrality) were calculated for quantitative understanding of the network structure. Isolated organizations from the main network were not included in this analysis. Organizations included in the network were geocoded onto a world map (WGS84, EPSG:4326) using the network analysis tools in ArcMap 10.1 (ESRI, Redlands, California, USA). Global or national organizations (e.g., the United States Geological Survey) were mapped at their headquarters. Organizations were clustered based on network connections using the Markov Cluster Algorithm in Gephi 0.9.1 (van Dongen and Abreu‐Goodger ).ResultsWhat was the core research topic during the last century?A total of 22,179 ecology publications were used to trace different use pattern of terms over the last 100 yr (Appendices S1, S2). The complex relationships of terms in these papers were clearly identified by a word network map (Fig. ) in which terms were clustered broadly into four groups (population, conservation and landscape, community and forest, and ecosystem ecology) based on overall network strength and the relationships between terms (Fig. a). The word network of ESA publications contained a population ecology cluster reflected by terms such as population (3578 occurrences), survival (1409), predator (1301), behavior (1118), mortality (1100), probability (879), selection (818), and movement (546). Population dynamics, competition, predation, and behavior were broadly connected research topics in this cluster. Trait, stage, and risk were more recently used terms in the population ecology cluster (Fig. b).Word network map of terms in Ecological Society of America journal publications from 1920 to 2015. Closely located points represent terms frequently used together. Size of the circle is proportional to the frequency of occurrence of the term in the text corpus (combined text of all publications evaluated). (a) Word network map; different colors represent different network clusters. Red, population ecology; green, conservation and landscape ecology; blue, community and forest ecology; yellow, ecosystem ecology. (b) Relative time scale network map; more recently used terms are red and terms used in earlier time periods are blue.The conservation and landscape ecology cluster was characterized by the terms approach (1976 occurrences), management (1314), landscape (1337), estimate (1134), biodiversity (768), framework (718), and simulation (529). Most of the terms in this cluster, including restoration, ecosystem service, policy, scenario, and network, were more frequent in recently published articles. The community and forest ecology cluster was positioned in the center of the other clusters and had close relationships with other overall topics. Ecological terms including forest (2282 occurrences), diversity (1466), tree (1301), disturbance (1175), plot (1050), composition (1045), and climate (919) were used more than other words in this cluster. Climate change, fire regime, community composition, functional group, and invasion were more recent terms in this cluster. Finally, the ecosystem ecology cluster was represented by terms like ecosystem (2188 occurrences), biomass (1420), production (1222), soil (1170), treatment (1152), concentration (1018), productivity (933), nitrogen (862), and carbon (724). Species richness, flux, food web, and trophic cascade were more recent terms in the ecosystem ecology cluster.What issues increased in importance during the last century?Term frequency analysis of publication titles revealed that forest (8.9% of all publications included this term) was the overall highest‐frequency term in titles of papers in ESA journals since the 1920s (Table ). Unigrams species (6.7%), population (6.7%), plant (6.6%), and community (6.5%) followed. The relative frequency of terms used in papers changed over time. In the period 1920–1939, plant (9.2%), soil (7.1%), vegetation (5.7%), and distribution (5.0%) constituted a large proportion of all terms used. An increased use of population (8.2%), growth (3.6%), and range (2.0%) was observed in the 1940–1959 period. Ecosystem (2.1%) entered the top 20 terms in the 1960–1979 period, and competition (2.9%), habitat (2.9%), and behavior (2.5%) were also high‐ranking terms in this period. In the 1980–1999 period, stream (6.3%) occurred frequently as did forest (8.6%) and community (6.9%). Interaction (4.2%), partitioning (4.2%), size (3.7%), predation (3.5%), and foraging (2.9%) also emerged as popular terms in this period. Lastly, model (5.6%), climate (4.4%), diversity (4.4%), landscape (4.3%), and fire (3.1%) appeared on the top 20 list during the 2000–2015 period.Frequency rank of unigrams (i.e., one‐word terms) in Ecological Society of America publication titles (number of publications) from 1920 to 2015Frequency rank1920–1930s (n = 1266)1940–1950s (n = 2209)1960–1970s (n = 3475)1980–1990s (n = 5428)2000–2010s (n = 9801)1Forest0.101Forest0.094Population0.089Forest0.086Forest0.0982Plant0.092Population0.082Forest0.067Community0.069Species0.0953Soil0.071Vegetation0.082Community0.055Stream0.063Plant0.0804Vegetation0.057Plant0.064Species0.054Competition0.056Community0.0795Distribution0.050Lake0.043Plant0.049Species0.056Population0.0716Water0.044Distribution0.040Vegetation0.049Population0.054Ecosystem0.0717Animal0.038Community0.037Soil0.040Plant0.049Model0.0568Lake0.036Growth0.036Distribution0.036Ecosystem0.043Habitat0.0549Temperature0.036Species0.034Growth0.034Habitat0.043Climate0.04410Community0.034Soil0.033Water0.030Structure0.043Diversity0.04411Insect0.034Water0.031Competition0.029Interaction0.042Landscape0.04312Pine0.028Temperature0.030Habitat0.029Partitioning0.042Spatial0.04013Life0.026Mountain0.022Temperature0.028Pattern0.041Tree0.03914Growth0.023Prairie0.021Lake0.027Model0.038Soil0.03915Population0.022Animal0.020Behavior0.025Size0.037Pattern0.03616Fish0.022Range0.020Pattern0.025Predation0.035Interaction0.03517Bog0.021History0.020Structure0.022Tropical0.034Tropical0.03118Prairie0.021Pine0.020Desert0.021Growth0.032Fire0.03119Succession0.021Fish0.019Dynamics0.021Tree0.031Marine0.03020Tree0.021Bird0.019Ecosystem0.021Foraging0.029Structure0.030Total no. of terms25034438743111,70622,699NoteNumbers in the table indicate the ratio of term frequency to the total number of publications in each decade.The relative use of complementary terms was compared (Fig. ) with diversity exhibiting the highest relative volume (%) among biodiversity terms, and its use increased rapidly since the 1960s (Fig. a). Use of prey and host peaked during the 1980s and were used with a similar average relative volume (Fig. b). Predator also increased in use since the 1960s. More publications used source than sink (Fig. c). In the 1950s, interspecific was used more frequently than intraspecific, but these terms were used at similar rates by the 2010s (Fig. d). Latitude and vertical appeared more frequently than longitude and horizontal (Fig. e), and there was a notable increase in the use of latitude (Mann–Kendall trend S = 27, Z = 0.34, P = 0.0083) and decrease in the use of vertical (S = −33, Z = −0.37, P = 0.0011) from the 1940s to the 1960s. Indirect effects and relationships in heterogeneous spaces were used more frequently than their opposites (Fig. f).Comparison of relative use of complementary ecological terms in Ecological Society of America journal publications during the last 100 yr. The frequency of each word is represented by the relative volume (%) of articles including the word out of within the total number of publications per decade.Six groups of terms were further compared over time to evaluate changes in use by decade (Fig. ). Relative volumes of terms indicating ecological level (e.g., individual, population, and ecosystem) increased since the 1920s (Fig. a). Use of landscape and global showed notable increases in the 1990s. Environmental media terms (i.e., water, soil, air, and light) were common in the 1920s, but they became less common after the 1940s (Fig. b). Vegetation constituted more than 6% of total usage until the 1960s, but this rapidly decreased after that time (Fig. c). Succession (3.2% in the 1920s, 0.8% in the 2010s) showed a similar pattern. Model and climate increased sharply in the 1960s (Fig. d). The use of restoration and Bayesian in publications began to increase in the 1990s. Studies using structure and competition peaked in the 1980s, but those using the term function steadily increased since that time (Fig. e). Predation, forage, and behavior peaked in the 1970s or 1980s (Fig. f). Among the various types of ecological disturbances, the relative volume of fire in ESA publications exhibited a rapid increase since the 1960s (Fig. g). Relative use of the terms invasive and genetic has skyrocketed in recent decades, but their relative contribution to the total volume remained low (approximately 2–3%; Fig. h). Carbon and nitrogen rose around the 1950–1960s, but an increasing trend was experienced only by carbon. The frequency of occurrence of phosphorous did not fluctuate much in the text.Decadal change in the relative volumes of ecological terms from Ecological Society of America journal publications over the last 100 yr with notable trends. The frequency of each word is represented by the relative volume (%) of articles including the word within the total number of publications per decade.How well did relative volumes of concept terms match with the consensus among ecologists?Of the 131 ecological concepts presented by Reiners et al. (), we were able to match 93 of them (i.e., unigram or bigram) with those from our text corpus, and they reasonably matched the mean concept scores determined from ecologist responses to the previous questionnaire surveys (Fig. a; adjusted R2 = 0.37, P < 0.0001). Relative volumes of terms calculated from the text corpus in our study that were about 20 times different were approximately 0.5 points different in terms of mean concept score from the questionnaire survey results. A similar positive relationship was observed with ranking scores (Fig. b; adjusted R2 = 0.34, P < 0.0001). The general rank pattern of ecological concepts was well matched even among the different questionnaire results (Fig. c; concept rankings in 1986 and 2014 determined from questionnaire survey results). Relative changes in term use identified in our text‐mining results were also comparable with the previous questionnaire survey results. Changes in the rank of ecological concepts as voted on by ecologists in the surveys exhibited a similar pattern of increase/decrease (i.e., succession was rank 2 in the 1980s and rank 24 in the 2010s; competition changed from rank 5 to 9; disturbance changed from rank 26 to 5) to the actual frequency of occurrence in ESA publications (i.e., relative volume; refer to Fig. ).Relationships between the relative volume (RV) of ecological concept terms appearing in publications from Ecological Society of America journals and voted rankings determined from questionnaire surveys of ecologists. Mean concept score and ranks of ecological concepts were referenced from Reiners et al. () and Cherrett (). (a) Comparison between average RVs and mean concept scores (Ecological Society of America [ESA] survey in 2014); (b) comparison between average RVs and concept rankings (ESA survey in 2014); (c) relationship of concept rankings from different questionnaire surveys (British Ecological Society [BES] in 1986 and ESA in 2014).How did ESA ecologists collaborate at the national and global scales?We identified 4055 organizations and 29,638 research collaborations in the ESA publication network (Fig. ). More than 9100 authors (n = 9126; 32.46% of all authors) had more than two publications. Only 686 authors (2.44%) had more than 10 publications in the study corpus. About 40% (40.71%; n = 1651) of all organizations had more than one paper in ESA journals. The average number of institutions per paper gradually increased since the 1980s (Fig. a), and the average number of institutions co‐authoring a paper increased since 2008 (1980s, 1.05 ± 0.27 SD; 1990s, 1.28 ± 1.11; 2000s, 2.02 ± 1.94; 2010s, 4.03 ± 3.60). The average degree of this network was 14.61, which means that each organization had about 14 or 15 collaborators on their scientific publications.Number of organizational co‐authors and organizational network map from Ecological Society of America publications from 1920 to 2015. (a) Number of organizations co‐authoring each paper and (b) global, (c) North American, (d) and European distributions of collaborations. Relationships between the number of publications and (e) network degree and (f) average number of citations. The 400 organizations (node) with the highest network degrees are shown on the map. Node size is proportional to the number of degrees in the full network, and different colors represent different cluster groups.Also notable was that the majority of organizational networks were tightly linked within national or other geographical boundaries. Globally, North America, Europe, and Australia represented the majority of collaborations (Fig. b). Within the United States, the majority of collaborations were within either western (Fig. c) or eastern groups (Fig. c). These three clusters constituted about 36.17% of the total number of organizations and had the highest average network degrees (Fig. e), which were linearly related to the average number of publications (Fig. e, regression slope: 0.66, R2 = 0.67, P < 0.005). The average number of citations, however, did not have a simple linear relationship with the number of publications across clusters (Fig. f).DiscussionA broad spectrum of terms was identified from the 100 yr of scientific texts we analyzed, and dynamic changes in ecological issues were outlined. It was difficult to separate a stand‐alone topic from the texts as diverse topics were mixed within ecological research themes. We would, however, like to point out a few specific issues that were examples of common issues overall in the ecological word network. Readers should keep in mind that the difference in the relative frequency of term use is not necessarily reflective of the degree of development in the field, but only reflects the relative degree of interest or practical use within the scientific community.Issues of population‐level approach and forest studiesPopulation‐level research frequently involves structure, function, and relationships with environmental conditions. We assume population ecology had comparable outcomes (Chesson and Kuang ) as diverse experimental designs are possible under either natural or manipulated conditions. Comparatively speaking, large‐scale studies with ecosystem‐ or landscape‐level approaches lead to a high proportion of articles focusing on conservation and management. More efforts to integrate the hypotheses and results from population‐ and community‐level studies into knowledge at the ecosystem level would be a benefit. The high frequency of forest‐related research topics was also notable. As ESA was created when a relatively high proportion of members were studying plant and forest ecology (plant ecology 28.7%, animal ecology 28.0%, forestry 14.0%, entomology 12.7%, marine ecology 4.6%, agriculture 3.9%, and others 8.1%; from Shelford , Egerton ), this initial composition of topics may have remained and was thus reflected by the relative frequencies of these topics.Transition to a new topicWe identified gradual and concurrent changes in the use of terms in ecological publications. Changes in relative frequencies of groups of terms represented transitions in research in the subject area. In the 1920s, a large number of publications used terms associated with characteristics of environmental media (e.g., soil, water, air, and light), such as temperature, vegetation, and succession. Structure, competition, and disturbance peaked during the 1980s. Climate, function, fire, invasive, and genetic rose rapidly in the 2010s. Long‐term, flux, and restoration also increased since the 1980s. It is notable that the biological traits of species and their distribution, productivity, and ecological interactions seemed to be the main focus during the early–mid years of ecological publications; however, more recently, use of human‐related terms (e.g., management, conservation, policy, restoration, land use, and human activity) appeared in the text corpus. This pattern matches with recent strong demands for sustainability, an interest in increased human impacts on the earth (Palmer et al. ), and the development of urban ecology (Pickett et al. ).Changes in the use of terms related to these research topics have been influenced by the diversity of ecological themes resulting from the accumulation of research data and the increasing number of researchers. However, there was also an increase in the number of subjects due to new approaches and the development of new technologies from different disciplines. For example, recent studies of ecological disturbances such as hurricanes, flooding, droughts, and wildfires have improved accuracy in disturbance frequency determination and probability information from newly developed climate prediction models (Rind et al. ). A recent systematic review of ecology journals also detected rapid changes in the frequency of climate change and biodiversity topics (Carmel et al. ). Increased social awareness and ecosystem conservation and recovery efforts could be important factors (Perring et al. ) driving the increasing frequency of terms related to restoration. Knowledge of populations and ecosystems was core contributions to the development of restoration techniques such as bioremediation and biological control. More recently, molecular techniques have been widely applied to study invasive species, regional biodiversity of microorganisms, and food web structure (Huver et al. , Shelton et al. ). Big data analytics are also being applied to ecosystem complexity studies. It seems clear that the convergence of multiple disciplines and application of new techniques will provide us with new information (Moe et al. ), but we would like to emphasize that existing ecological knowledge will be a good reference to interpret emerging results using new tools.Expansion of the research unitAs assessed by terms related to ecological hierarchy, population‐level approaches have generally predominated over the last 70 yr. This hierarchical structure was thought to be the traditional basic unit of biological response in ecological disciplines, and many studies have focused on the population level. Population‐level approaches notably include specific characteristics of species including abundance, mortality, dispersal, adaptation, reproduction, interaction, and predation. Our results showed that the use of community in ecological papers decreased in the 1980s, and there was long discussion about its practical application during this time (Grubb and Whittaker ). Ecosystem‐ and landscape‐scale approaches have increased rapidly since the 1970s. Ecosystem and landscape ecology are more focused on energy flow, nutrient cycling, habitat composition, and complex patterns on a large scale. These large‐scale and systematic data have been used to support management and conservation of related ecosystems (Dale and Beyeler ). However, Carmel et al. () noted that community and ecosystem studies constituted only a minor proportion (17%, 25%) of all studies evaluated, even considering their roles as two of the major concepts in ecology, and in the same systematic review of major ecology journals, they found that more than 60% of ecological studies still focused on a single species.It may be that research scales used by ecologists will be expanded as techniques and research networks grow. It is also evident that the relative frequency of larger‐scale research has increased, but this was not a transitional increase from the small scale to the large scale. A gap has arisen from the use of different basic units of scale, and the directions that researchers of these two groups (small‐ and large‐scale studies) pursued. Population and community ecology involves study of the structural laws that make up an ecosystem. Thus, these subfields of ecology detail demographic structure and involve many manipulated experiments using model systems, whereas ecosystem and landscape ecology are more focused on the systematic structure of biological and environmental components.To understand the structure and processes of the global ecosystem, it is necessary to integrate knowledge from both research scales. Kingsland () stressed the importance of diverse scales in studies aiming to solve ecological problems and expand our understanding of the entire biosphere. A deeper understanding of the responses of contiguous local populations and communities has accumulated over decades (Wootton ), and it is contributing to a more accurate evaluation and prediction of the impacts of global ecosystems and global change. In addition, many researchers are looking in new directions regarding community studies based on functional traits rather than individual species (Simberloff , McGill et al. ). Quantitative comparisons of the relative intensities of relationships between species and environmental factors in changing environmental contexts will enable us to coordinate with higher‐level system studies (Agrawal et al. ). Larger‐scale research topics will also require more effort to reduce the gap with population‐level research. For example, landscape ecology is particularly limited by the application of existing experimental design approaches, and new approaches are needed (Wu and Hobbs , Jenerette and Shen ). Ecological studies using individual‐based models (Shugart et al. ) and macroecology are also parts of the effort to reduce these gaps.Potential application of text mining of ecological concept studiesIn this study, we demonstrated the compatibility of text mining and questionnaire survey methods to evaluate the use of ecological concept terms. Generally, a limited number of experienced researchers can be considered as a tentative pool for the survey method (Sutherland et al. ). From this perspective, a text‐mining‐based approach has some merit as it can benefit from the rich legacy of ecological literature that has accumulated for the last 100 yr. Further, this approach can include a broader range of perspectives from the ecologist pool as peer‐reviewed papers are a standard tool for communication in the scientific community.For effective evaluation and categorization of ecological concepts, however, these two approaches should be well coordinated. A text‐mining approach would be suitable for extracting candidate ecological terms based on their frequency of occurrence and the network influence of the associated scientific publications. However, candidate terms will definitely need screening and discussion by ecologists to evaluate their relevance to ecological concepts or theories. Reiners et al. () suggested possible application of text mining to help understand the intellectual structure of the ecological conceptual pool (i.e., using term network structure for thematic classification and tracing historical use of a concept term over time for temporal trends). To further extend application of text mining of concept studies, it will be necessary to categorize and link specific ecological concepts and related terms used in the publications. As described above, it is thought that a network analysis of terms in ecological publications can be useful toward this end. In future studies, it seems it will be important to analyze differences in the use of ecological concept terms by researcher age class or geographical locality. Previous studies have also noted differences in concept ranking among age classes (Reiners et al. ). This difference in the tendency of use of ecological concepts may be reflected in the term pools from papers published by different generations of ecologists.Geographical barriers in organization networksEcological Society of America was first established with core memberships in the United States of America (96.74% of total members; Shelford ). Subsequently, individuals from more than 4000 international organizations have joined during the last 100 yr. Thus, we estimated how researchers have collaborated with others based on co‐authorship networks. Analysis of research collaboration through joint participation of organizations revealed that the majority of ecological research cooperation was geographically concentrated between limited institutions. These biased and concentrated distributions also matched with the previous findings of a study of highly cited environmental scientists (Parker et al. ). Accumulative geographic advantage of academic resources (McIntosh , Pasterkamp et al. ) and advanced infrastructure in academic fields (Basu ) may have contributed to the current unbalanced distribution of the collaboration network. However, it should be considered that our results are likely an underestimation of actual collaboration in the field of ecology. The actual collaboration network is greater than what our results imply as the current analysis only focused on the authorship network of peer‐reviewed publications in ESA journals.Collaboration of ecologists and researchers from a wide range of other disciplines (e.g., physics, chemistry, biology, geography, and many others) is essential to understanding the complexity of communities, ecosystems, evolution, and the core structures of global ecosystems. Although the specific objectives are different, LTER (Long‐Term Ecological Research), GLEON (Global Lake Ecological Observatory Network), and NEON (National Ecological Observatory Network) are examples of projects that emphasize the importance of collaboration to find insight into ecological questions (Vanderbilt et al. , Springer et al. ). In addition, the amount of ecological data (e.g., from an ecological sensor network) that is currently accumulating is too massive for any one particular group to cover and assimilate (Weathers et al. ). Studies of global ecosystems and future changes will require further international collaboration by ecologists from diverse areas. Taking this idea one step forward, this international collaboration of ecologists should be expanded to include diverse stakeholders including policy makers, decision makers, and the public (Palmer et al. ).ConclusionsWe recommend more in‐depth review of the accumulated historical data and ideas in the ecological literature. Digital archives (i.e., historical scientific data) are accessible at a low cost, and they contain massive amounts of unused information. Tangible or intangible forms of historical records from ecological fields should be conserved for future integration and in‐depth analysis. This work highlights the importance of continued and safe storage of historical ecological data. Further review of current ecological archives would provide helpful information for deciding the future direction of ecology and could be used to refine ecological concept models from a different perspective.AcknowledgmentsWe sincerely thank the enthusiastic ESA members who attended our presentation for fruitful discussion and their comments at the 100th and 101st ESA annual meetings. Ji Yoon Kim is a JSPS (Japan Society for the Promotion of Science) International Research Fellow (ID No: P17387). JSPS is a nonprofit organization.Literature CitedAgrawal, A. A., et al. 2007. Filling key gaps in population and community ecology. Frontiers in Ecology and the Environment 5:145–152.Basu, A. 2006. Using ISI's ‘Highly Cited Researchers’ to obtain a country level indicator of citation excellence. Scientometrics 68:361–375.Borrett, S. R., J. Moody, and A. Edelmann. 2014. The rise of network ecology: maps of the topic diversity and scientific collaboration. Ecological Modelling 293:111–127.Carmel, Y., R. Kent, A. Bar‐Massada, L. Blank, J. Liberzon, O. Nezer, G. Sapir, and R. Federman. 2013. Trends in ecological research during the last three decades – a systematic review. PLoS ONE 8:e59813.Cherrett, J. M. 1989. 1. Key concepts: the results of a survey of our members' opinions. Pages 1–16 in J. M. Cherrett, A. D. Bradshaw, F. B. Goldsmith, P. G. Grubb, and J. R. Krebs, editors. Ecological concepts: The contribution of ecology to an understanding of the natural world. Blackwell Scientific Publications, Oxford, UK.Chesson, P., and J. J. Kuang. 2008. The interaction between predation and competition. Nature 456:235–238.Cullen, P. W., R. H. Norris, V. H. Resh, T. B. Reynoldson, D. M. Rosenberg, and M. T. Barbour. 1999. Collaboration in scientific research: a critical need for freshwater ecology. Freshwater Biology 42:131–142.Dale, V. H., and S. C. Beyeler. 2001. Challenges in the development and use of ecological indicators. Ecological Indicators 1:3–10.Do, Y., J. Y. Kim, M. Lineman, D. K. Kim, and G. J. Joo. 2015. Using internet search behavior to assess public awareness of protected wetlands. Conservation Biology 29:271–279.Egerton, F. N. 2015. A centennial history of the Ecological Society of America. CRC Press, Boca Raton, Florida, USA.Grubb, P. J., and J. B. Whittaker. 1988. Toward a more exact ecology. Blackwell Scientific Publication, London, UK.Huver, J. R., J. Koprivnikar, P. T. J. Johnson, and S. Whyard. 2015. Development and application of an eDNA method to detect and quantify a pathogenic parasite in aquatic ecosystems. Ecological Applications 25:991–1002.Jenerette, G. D., and W. Shen. 2012. Experimental landscape ecology. Landscape Ecology 27:1237–1248.Kim, J. Y., Y. Do, R. Y. Im, G. Y. Kim, and G. J. Joo. 2014. Use of large web‐based data to identify public interest and trends related to endangered species. Biodiversity and Conservation 12:1–24.Kingsland, S. 2004. Conveying the intellectual challenge of ecology: an historical perspective. Frontiers in Ecology and the Environment 2:367–374.Ladle, R. J., R. A. Correia, Y. Do, G. J. Joo, A. Malhado, R. Proulx, J. M. Roberge, and P. Jepson. 2016. Conservation culturomics. Frontiers in Ecology and the Environment 14:269–275.Lawton, J. H. 1999. Are there general laws in ecology? Oikos 84:177–192.McGill, B. J., B. J. Enquist, E. Weiher, and M. Westoby. 2006. Rebuilding community ecology from functional traits. Trends in Ecology & Evolution 21:178–185.McIntosh, R. P. 1989. Citations classics of ecology. Quarterly Review of Biology 64:31–49.Moe, S. J., R. S. Stelzer, M. R. Forman, W. S. Harpole, T. Daufresne, and T. Yoshida. 2005. Recent advances in ecological stoichiometry: insights for population and community ecology. Oikos 109:29–39.Palmer, M. A., et al. 2005. Ecological science and sustainability for the 21st century. Frontiers in Ecology and the Environment 3:4–11.Parker, J. N., C. Lortie, and S. Allesina. 2010. Characterizing a scientific elite: the social characteristics of the world's most highly cited scientists in environmental science and ecology. Scientometrics 85:129–143.Pasterkamp, G., J. I. Rotmans, D. V. P. De Kleun, and C. Borst. 2007. Citation frequency: a biased measure of research impact significantly influenced by the geographical origin of research articles. Scientometrics 70:153–165.Pennebaker, J. W., M. R. Mehl, and K. G. Niederhoffer. 2003. Psychological aspects of natural language use: our words, our selves. Annual Review of Psychology 54:547–577.Perring, M. P., R. J. Standish, J. N. Price, M. D. Craig, T. E. Erickson, K. X. Ruthrof, A. S. Whiteley, L. E. Valentine, and R. J. Hobbs. 2015. Advances in restoration ecology: rising to the challenges of the coming decades. Ecosphere 6:1–25.Pickett, S. T. A., M. L. Cadenasso, D. L. Childers, M. J. McDonnell, and W. Zhou. 2016. Evolution and future of urban ecological science: ecology in, of, and for the city. Ecosystem Health and Sustainability 2:e01229.Reiners, W. A., J. A. Lockwood, S. D. Prager, and J. C. Mulroy. 2015. Ecological concepts: What are they, what is their value, and for whom? Bulletin of the Ecological Society of America 96:64–69.Reiners, W. A., J. A. Lockwood, D. S. Reiners, and S. D. Prager. 2017. 100 years of ecology: What are our concepts and are they useful? Ecological Monographs 87:260–277.Rind, D., R. Goldberg, and R. Ruedy. 1989. Change in climate variability in the 21st century. Climatic Change 14:5–37.Shelford, V. E. 1917. The ideals and aims of the Ecological Society of America. Bulletin of the Ecological Society of America 1:1–8.Shelton, A. O., J. L. O'Donnell, J. F. Samhouri, N. Lowell, G. D. Williams, and R. P. Kelly. 2016. A framework for inferring biological communities from environmental DNA. Ecological Application 26:1645–1659.Shugart, H. H., G. P. Asner, R. Fischer, A. Huth, N. Knapp, T. Le Toan, and J. K. Shuman. 2015. Computer and remote‐sensing infrastructure to enhance large‐scale testing of individual‐based forest models. Frontiers in Ecology and the Environment 13:503–511.Simberloff, D. 2004. Community ecology: Is it time to move on? (An American Society of Naturalists presidential address). American Naturalist 163:787–799.Springer, Y. P., et al. 2016. Tick‐, mosquito‐, and rodent‐borne parasite sampling designs for the National Ecological Observatory Network. Ecosphere 7:e01271.Sutherland, W. J., et al. 2013. Identification of 100 fundamental ecological questions. Journal of Ecology 101:58–67.Thompson, J. N., et al. 2001. Frontiers of ecology. BioScience 51:15–24.van Dongen, S., and C. Abreu‐Goodger. 2012. Using MCL to extract clusters from networks. Methods in Molecular Biology 804:281–295.Vanderbilt, K. L., C. C. Lin, S. S. Lu, A. R. Kassim, H. He, X. Guo, I. S. Gil, D. Blankman, and J. H. Porter. 2015. Fostering ecological data sharing: collaborations in the International Long Term Ecological Research Network. Ecosphere 6:1–18.Waltman, L. R., N. J. P. Eck, and E. C. M. van Noyons. 2010. A unified approach to mapping and clustering of bibliometric networks. Center for Science and Technology Studies, Leiden, The Netherlands.Weathers, K. C., et al. 2016. Frontiers in ecosystem ecology from a community perspective: The future is boundless and bright. Ecosystems 19:753–770.Wootton, J. T. 1996. The nature and consequences of indirect effects in ecological communities. Annual Review of Ecology, Evolution and Systematics 25:443–466.Wu, J., and R. Hobbs. 2002. Key issues and research priorities in landscape ecology: an idiosyncratic synthesis. Landscape Ecology 17:355–365. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Ecosphere Wiley

Through 100 years of Ecological Society of America publications: development of ecological research topics and scientific collaborations

Kim, Ji Yoon; Joo, Gea‐Jae; Do, Yuno
Ecosphere , Volume 9 (2) – Jan 1, 2018
Download PDF
15 pages

Loading next page...
 
/lp/wiley/through-100-years-of-ecological-society-of-america-publications-um0Go4C74j

References (43)

Publisher
Wiley
Copyright
© 2018 The Ecological Society of America
ISSN
2150-8925
eISSN
2150-8925
DOI
10.1002/ecs2.2109
Publisher site
See Article on Publisher Site

Abstract

IntroductionIntegration and convergence of historical knowledge is becoming increasingly important, and new information is being discovered with retrospective analysis of old data. Since the discipline of ecology includes diverse academic fields (Cullen et al. , Thompson et al. ), there have been limited comprehensive reviews using objective data. Large bodies of text extracted from scientific publications have not been used frequently in ecological reviews. In particular, there have been limited attempts to focus on conceptual terms used in ecology (Cherrett , Lawton , Borrett et al. , Reiners et al. ) or the text itself in scientific publications. Scientific texts, as a historical record (i.e., ecology papers), contain information that reflects the academic consensus of the scientific community. Thus, the use of scientific terms is potentially important in estimating relative interest in scientific concepts within the academic community.Scientometric research (analysis of publications) and culturomics (quantitative analysis of text) have recently emerged as new tools to investigate certain phenomena in a scientific community using quantitative measurements of word frequencies in digital texts (Ladle et al. ). Words used by a community reflect the rooted ideas of its members (Pennebaker et al. ). Extending this concept, the language that scientists use in their papers reflects the current conceptual models or ideas of science. Culturomic tools have already provided new insights into conservation topics and have supported the conservation decision process by providing scientific measurements of public interest in ecological topics (Kim et al. , Do et al. ).Our study quantified the relative use of ecological concepts and research collaborations during the last 100 yr (1920–2015) using text from scientific publications. We performed a comprehensive scientometric analysis using 22,179 publications from journals published by the Ecological Society of America (ESA) since 1920 (Appendix S1). Our objectives were to (1) identify the relative volumes of certain terms in the papers, (2) compare trends in ecological topics and subjects, and (3) evaluate research collaborations in diverse ecological fields.MethodsText corpusWe established an initial article database of ESA journals (Appendix S1: Table S1) using Web of Science (webofscience.com) and Wiley Online Library (esajournals.onlinelibrary.wiley.com). These two academic databases contained all ESA journal publications from 1920 to the present. Each ESA journal was browsed for its full publication period (1920–2015), and publication records for scientific articles were collected semi‐automatically using the reference import function in a bibliography management software program (Zotero Version 4.0.27, Fairfax, Virginia, USA). Collected publication data included title, abstract, keywords, authors, institution or organization of authors, publication year, article type, and list of literature cited in each article. The complete collection constituted the text corpus (body of text) used for analysis. To check the integrity of the text corpus, all publication records were integrated into a single text file. Duplicated records in the database were merged into a single record. Corrections (e.g., errata), news, book reviews, and editorials were not included in the final text corpus. Thus, only peer‐reviewed articles were included. Natural language processing was performed before text data mining. To compile meaningful terms in the text corpus, we applied a linguistic filter (Waltman et al. ) that combined different verb tenses into a single verb and integrated different forms of nouns into a single noun.Text data miningScientometric techniques for word frequency counts and word and author network analyses were used to trace quantitative changes in ecological issues over the last 100 yr. Word frequency counts were used to identify the quantitative frequency of terms used in different time periods. Relative rank frequency of all unigrams (one‐word terms, e.g., ecosystem) and bigrams (two‐word terms, e.g., ecosystem health) was calculated using the text corpus. We applied a binary counting method to calculate the frequency of terms. In the case of binary counting, a term was counted only once per publication. As certain terms were repeatedly used in some papers, a full counting method may have overestimated the occurrence of those terms. Meaningless combinations of prepositions or adverbial phrases were excluded from analysis.Relationships between ecological terms were mapped with a word network using the text corpus. In a word network, an algorithm determines the relative position of each term on a density map by weighting the frequency of its co‐occurrence with other terms. We used VOSviewer 1.6.5 (Leiden University, Leiden, The Netherlands) and Gephi 0.9.1 (The Gephi Consortium, Compiègne, France) with binary counting methods to visualize the network and density map. Synonyms were merged into a single term before analysis. Terms in the word network were grouped into representative network clusters using the VOS clustering algorithm (Waltman et al. ).We divided the study period into five time frames (1920–1939, 1940–1959, 1960–1979, 1980–1999, and 2000–2015) so that the relative use of topic words over time could be compared. To compare the relative use of each term, its frequency of use was normalized by dividing it by the total number of publications within the selected time period. Thus, word frequency was expressed as the relative volume (%) of a term within each decade. A Mann–Kendall trend test was used to assess increasing or decreasing trends in variables (number of publications and relative volume of specific terms) over the study period. Additionally, we screened candidate terms based on their relative volume and trend slope during the study periods. Terms that rapidly increased or decreased over time were selected, and these words were further grouped by either similar topic or complementary pairs.The relative importance and influence of ecological concepts have been estimated based on questionnaire survey methods (Sutherland et al. , Reiners et al. ) and group discussions among experienced scientists. To efficiently utilize our text‐mining results, we compared the relative volume of terms from our text mining with the concept term rankings from previous survey methods. Reiners et al. () have shared the list of concept terms and associated ranking scores from their study in the University of Wyoming Research Data Repository (https://doi.org/10.15786/m2z599). Additionally, there were similar surveys conducted of the members of the British Ecological Society in 1986 (Cherrett ) and again in 2013 (Sutherland et al. ). These concept terms and ranks were integrated into a single database, and we listed overlapping terms from the different survey results. Finally, relationships among the different scores and ranks of concept terms were analyzed with the regression analysis tool in SigmaPlot (version 12.5; Systat Software, Erkrath, Germany).Research collaborationResearch collaboration relationships were analyzed through publication co‐authorship as peer‐reviewed publications are one of the core forms of media in the scientific community. We assumed that projects or other collaborative endeavors would be represented by researchers named as co‐authors on published articles. We organized author and organization data from the text corpus as described above. Based on collaborative research units (i.e., a co‐author group in a paper), an organizational research network was established and network centrality indices (e.g., eigenvector centrality, harmonic centrality, and closeness centrality) were calculated for quantitative understanding of the network structure. Isolated organizations from the main network were not included in this analysis. Organizations included in the network were geocoded onto a world map (WGS84, EPSG:4326) using the network analysis tools in ArcMap 10.1 (ESRI, Redlands, California, USA). Global or national organizations (e.g., the United States Geological Survey) were mapped at their headquarters. Organizations were clustered based on network connections using the Markov Cluster Algorithm in Gephi 0.9.1 (van Dongen and Abreu‐Goodger ).ResultsWhat was the core research topic during the last century?A total of 22,179 ecology publications were used to trace different use pattern of terms over the last 100 yr (Appendices S1, S2). The complex relationships of terms in these papers were clearly identified by a word network map (Fig. ) in which terms were clustered broadly into four groups (population, conservation and landscape, community and forest, and ecosystem ecology) based on overall network strength and the relationships between terms (Fig. a). The word network of ESA publications contained a population ecology cluster reflected by terms such as population (3578 occurrences), survival (1409), predator (1301), behavior (1118), mortality (1100), probability (879), selection (818), and movement (546). Population dynamics, competition, predation, and behavior were broadly connected research topics in this cluster. Trait, stage, and risk were more recently used terms in the population ecology cluster (Fig. b).Word network map of terms in Ecological Society of America journal publications from 1920 to 2015. Closely located points represent terms frequently used together. Size of the circle is proportional to the frequency of occurrence of the term in the text corpus (combined text of all publications evaluated). (a) Word network map; different colors represent different network clusters. Red, population ecology; green, conservation and landscape ecology; blue, community and forest ecology; yellow, ecosystem ecology. (b) Relative time scale network map; more recently used terms are red and terms used in earlier time periods are blue.The conservation and landscape ecology cluster was characterized by the terms approach (1976 occurrences), management (1314), landscape (1337), estimate (1134), biodiversity (768), framework (718), and simulation (529). Most of the terms in this cluster, including restoration, ecosystem service, policy, scenario, and network, were more frequent in recently published articles. The community and forest ecology cluster was positioned in the center of the other clusters and had close relationships with other overall topics. Ecological terms including forest (2282 occurrences), diversity (1466), tree (1301), disturbance (1175), plot (1050), composition (1045), and climate (919) were used more than other words in this cluster. Climate change, fire regime, community composition, functional group, and invasion were more recent terms in this cluster. Finally, the ecosystem ecology cluster was represented by terms like ecosystem (2188 occurrences), biomass (1420), production (1222), soil (1170), treatment (1152), concentration (1018), productivity (933), nitrogen (862), and carbon (724). Species richness, flux, food web, and trophic cascade were more recent terms in the ecosystem ecology cluster.What issues increased in importance during the last century?Term frequency analysis of publication titles revealed that forest (8.9% of all publications included this term) was the overall highest‐frequency term in titles of papers in ESA journals since the 1920s (Table ). Unigrams species (6.7%), population (6.7%), plant (6.6%), and community (6.5%) followed. The relative frequency of terms used in papers changed over time. In the period 1920–1939, plant (9.2%), soil (7.1%), vegetation (5.7%), and distribution (5.0%) constituted a large proportion of all terms used. An increased use of population (8.2%), growth (3.6%), and range (2.0%) was observed in the 1940–1959 period. Ecosystem (2.1%) entered the top 20 terms in the 1960–1979 period, and competition (2.9%), habitat (2.9%), and behavior (2.5%) were also high‐ranking terms in this period. In the 1980–1999 period, stream (6.3%) occurred frequently as did forest (8.6%) and community (6.9%). Interaction (4.2%), partitioning (4.2%), size (3.7%), predation (3.5%), and foraging (2.9%) also emerged as popular terms in this period. Lastly, model (5.6%), climate (4.4%), diversity (4.4%), landscape (4.3%), and fire (3.1%) appeared on the top 20 list during the 2000–2015 period.Frequency rank of unigrams (i.e., one‐word terms) in Ecological Society of America publication titles (number of publications) from 1920 to 2015Frequency rank1920–1930s (n = 1266)1940–1950s (n = 2209)1960–1970s (n = 3475)1980–1990s (n = 5428)2000–2010s (n = 9801)1Forest0.101Forest0.094Population0.089Forest0.086Forest0.0982Plant0.092Population0.082Forest0.067Community0.069Species0.0953Soil0.071Vegetation0.082Community0.055Stream0.063Plant0.0804Vegetation0.057Plant0.064Species0.054Competition0.056Community0.0795Distribution0.050Lake0.043Plant0.049Species0.056Population0.0716Water0.044Distribution0.040Vegetation0.049Population0.054Ecosystem0.0717Animal0.038Community0.037Soil0.040Plant0.049Model0.0568Lake0.036Growth0.036Distribution0.036Ecosystem0.043Habitat0.0549Temperature0.036Species0.034Growth0.034Habitat0.043Climate0.04410Community0.034Soil0.033Water0.030Structure0.043Diversity0.04411Insect0.034Water0.031Competition0.029Interaction0.042Landscape0.04312Pine0.028Temperature0.030Habitat0.029Partitioning0.042Spatial0.04013Life0.026Mountain0.022Temperature0.028Pattern0.041Tree0.03914Growth0.023Prairie0.021Lake0.027Model0.038Soil0.03915Population0.022Animal0.020Behavior0.025Size0.037Pattern0.03616Fish0.022Range0.020Pattern0.025Predation0.035Interaction0.03517Bog0.021History0.020Structure0.022Tropical0.034Tropical0.03118Prairie0.021Pine0.020Desert0.021Growth0.032Fire0.03119Succession0.021Fish0.019Dynamics0.021Tree0.031Marine0.03020Tree0.021Bird0.019Ecosystem0.021Foraging0.029Structure0.030Total no. of terms25034438743111,70622,699NoteNumbers in the table indicate the ratio of term frequency to the total number of publications in each decade.The relative use of complementary terms was compared (Fig. ) with diversity exhibiting the highest relative volume (%) among biodiversity terms, and its use increased rapidly since the 1960s (Fig. a). Use of prey and host peaked during the 1980s and were used with a similar average relative volume (Fig. b). Predator also increased in use since the 1960s. More publications used source than sink (Fig. c). In the 1950s, interspecific was used more frequently than intraspecific, but these terms were used at similar rates by the 2010s (Fig. d). Latitude and vertical appeared more frequently than longitude and horizontal (Fig. e), and there was a notable increase in the use of latitude (Mann–Kendall trend S = 27, Z = 0.34, P = 0.0083) and decrease in the use of vertical (S = −33, Z = −0.37, P = 0.0011) from the 1940s to the 1960s. Indirect effects and relationships in heterogeneous spaces were used more frequently than their opposites (Fig. f).Comparison of relative use of complementary ecological terms in Ecological Society of America journal publications during the last 100 yr. The frequency of each word is represented by the relative volume (%) of articles including the word out of within the total number of publications per decade.Six groups of terms were further compared over time to evaluate changes in use by decade (Fig. ). Relative volumes of terms indicating ecological level (e.g., individual, population, and ecosystem) increased since the 1920s (Fig. a). Use of landscape and global showed notable increases in the 1990s. Environmental media terms (i.e., water, soil, air, and light) were common in the 1920s, but they became less common after the 1940s (Fig. b). Vegetation constituted more than 6% of total usage until the 1960s, but this rapidly decreased after that time (Fig. c). Succession (3.2% in the 1920s, 0.8% in the 2010s) showed a similar pattern. Model and climate increased sharply in the 1960s (Fig. d). The use of restoration and Bayesian in publications began to increase in the 1990s. Studies using structure and competition peaked in the 1980s, but those using the term function steadily increased since that time (Fig. e). Predation, forage, and behavior peaked in the 1970s or 1980s (Fig. f). Among the various types of ecological disturbances, the relative volume of fire in ESA publications exhibited a rapid increase since the 1960s (Fig. g). Relative use of the terms invasive and genetic has skyrocketed in recent decades, but their relative contribution to the total volume remained low (approximately 2–3%; Fig. h). Carbon and nitrogen rose around the 1950–1960s, but an increasing trend was experienced only by carbon. The frequency of occurrence of phosphorous did not fluctuate much in the text.Decadal change in the relative volumes of ecological terms from Ecological Society of America journal publications over the last 100 yr with notable trends. The frequency of each word is represented by the relative volume (%) of articles including the word within the total number of publications per decade.How well did relative volumes of concept terms match with the consensus among ecologists?Of the 131 ecological concepts presented by Reiners et al. (), we were able to match 93 of them (i.e., unigram or bigram) with those from our text corpus, and they reasonably matched the mean concept scores determined from ecologist responses to the previous questionnaire surveys (Fig. a; adjusted R2 = 0.37, P < 0.0001). Relative volumes of terms calculated from the text corpus in our study that were about 20 times different were approximately 0.5 points different in terms of mean concept score from the questionnaire survey results. A similar positive relationship was observed with ranking scores (Fig. b; adjusted R2 = 0.34, P < 0.0001). The general rank pattern of ecological concepts was well matched even among the different questionnaire results (Fig. c; concept rankings in 1986 and 2014 determined from questionnaire survey results). Relative changes in term use identified in our text‐mining results were also comparable with the previous questionnaire survey results. Changes in the rank of ecological concepts as voted on by ecologists in the surveys exhibited a similar pattern of increase/decrease (i.e., succession was rank 2 in the 1980s and rank 24 in the 2010s; competition changed from rank 5 to 9; disturbance changed from rank 26 to 5) to the actual frequency of occurrence in ESA publications (i.e., relative volume; refer to Fig. ).Relationships between the relative volume (RV) of ecological concept terms appearing in publications from Ecological Society of America journals and voted rankings determined from questionnaire surveys of ecologists. Mean concept score and ranks of ecological concepts were referenced from Reiners et al. () and Cherrett (). (a) Comparison between average RVs and mean concept scores (Ecological Society of America [ESA] survey in 2014); (b) comparison between average RVs and concept rankings (ESA survey in 2014); (c) relationship of concept rankings from different questionnaire surveys (British Ecological Society [BES] in 1986 and ESA in 2014).How did ESA ecologists collaborate at the national and global scales?We identified 4055 organizations and 29,638 research collaborations in the ESA publication network (Fig. ). More than 9100 authors (n = 9126; 32.46% of all authors) had more than two publications. Only 686 authors (2.44%) had more than 10 publications in the study corpus. About 40% (40.71%; n = 1651) of all organizations had more than one paper in ESA journals. The average number of institutions per paper gradually increased since the 1980s (Fig. a), and the average number of institutions co‐authoring a paper increased since 2008 (1980s, 1.05 ± 0.27 SD; 1990s, 1.28 ± 1.11; 2000s, 2.02 ± 1.94; 2010s, 4.03 ± 3.60). The average degree of this network was 14.61, which means that each organization had about 14 or 15 collaborators on their scientific publications.Number of organizational co‐authors and organizational network map from Ecological Society of America publications from 1920 to 2015. (a) Number of organizations co‐authoring each paper and (b) global, (c) North American, (d) and European distributions of collaborations. Relationships between the number of publications and (e) network degree and (f) average number of citations. The 400 organizations (node) with the highest network degrees are shown on the map. Node size is proportional to the number of degrees in the full network, and different colors represent different cluster groups.Also notable was that the majority of organizational networks were tightly linked within national or other geographical boundaries. Globally, North America, Europe, and Australia represented the majority of collaborations (Fig. b). Within the United States, the majority of collaborations were within either western (Fig. c) or eastern groups (Fig. c). These three clusters constituted about 36.17% of the total number of organizations and had the highest average network degrees (Fig. e), which were linearly related to the average number of publications (Fig. e, regression slope: 0.66, R2 = 0.67, P < 0.005). The average number of citations, however, did not have a simple linear relationship with the number of publications across clusters (Fig. f).DiscussionA broad spectrum of terms was identified from the 100 yr of scientific texts we analyzed, and dynamic changes in ecological issues were outlined. It was difficult to separate a stand‐alone topic from the texts as diverse topics were mixed within ecological research themes. We would, however, like to point out a few specific issues that were examples of common issues overall in the ecological word network. Readers should keep in mind that the difference in the relative frequency of term use is not necessarily reflective of the degree of development in the field, but only reflects the relative degree of interest or practical use within the scientific community.Issues of population‐level approach and forest studiesPopulation‐level research frequently involves structure, function, and relationships with environmental conditions. We assume population ecology had comparable outcomes (Chesson and Kuang ) as diverse experimental designs are possible under either natural or manipulated conditions. Comparatively speaking, large‐scale studies with ecosystem‐ or landscape‐level approaches lead to a high proportion of articles focusing on conservation and management. More efforts to integrate the hypotheses and results from population‐ and community‐level studies into knowledge at the ecosystem level would be a benefit. The high frequency of forest‐related research topics was also notable. As ESA was created when a relatively high proportion of members were studying plant and forest ecology (plant ecology 28.7%, animal ecology 28.0%, forestry 14.0%, entomology 12.7%, marine ecology 4.6%, agriculture 3.9%, and others 8.1%; from Shelford , Egerton ), this initial composition of topics may have remained and was thus reflected by the relative frequencies of these topics.Transition to a new topicWe identified gradual and concurrent changes in the use of terms in ecological publications. Changes in relative frequencies of groups of terms represented transitions in research in the subject area. In the 1920s, a large number of publications used terms associated with characteristics of environmental media (e.g., soil, water, air, and light), such as temperature, vegetation, and succession. Structure, competition, and disturbance peaked during the 1980s. Climate, function, fire, invasive, and genetic rose rapidly in the 2010s. Long‐term, flux, and restoration also increased since the 1980s. It is notable that the biological traits of species and their distribution, productivity, and ecological interactions seemed to be the main focus during the early–mid years of ecological publications; however, more recently, use of human‐related terms (e.g., management, conservation, policy, restoration, land use, and human activity) appeared in the text corpus. This pattern matches with recent strong demands for sustainability, an interest in increased human impacts on the earth (Palmer et al. ), and the development of urban ecology (Pickett et al. ).Changes in the use of terms related to these research topics have been influenced by the diversity of ecological themes resulting from the accumulation of research data and the increasing number of researchers. However, there was also an increase in the number of subjects due to new approaches and the development of new technologies from different disciplines. For example, recent studies of ecological disturbances such as hurricanes, flooding, droughts, and wildfires have improved accuracy in disturbance frequency determination and probability information from newly developed climate prediction models (Rind et al. ). A recent systematic review of ecology journals also detected rapid changes in the frequency of climate change and biodiversity topics (Carmel et al. ). Increased social awareness and ecosystem conservation and recovery efforts could be important factors (Perring et al. ) driving the increasing frequency of terms related to restoration. Knowledge of populations and ecosystems was core contributions to the development of restoration techniques such as bioremediation and biological control. More recently, molecular techniques have been widely applied to study invasive species, regional biodiversity of microorganisms, and food web structure (Huver et al. , Shelton et al. ). Big data analytics are also being applied to ecosystem complexity studies. It seems clear that the convergence of multiple disciplines and application of new techniques will provide us with new information (Moe et al. ), but we would like to emphasize that existing ecological knowledge will be a good reference to interpret emerging results using new tools.Expansion of the research unitAs assessed by terms related to ecological hierarchy, population‐level approaches have generally predominated over the last 70 yr. This hierarchical structure was thought to be the traditional basic unit of biological response in ecological disciplines, and many studies have focused on the population level. Population‐level approaches notably include specific characteristics of species including abundance, mortality, dispersal, adaptation, reproduction, interaction, and predation. Our results showed that the use of community in ecological papers decreased in the 1980s, and there was long discussion about its practical application during this time (Grubb and Whittaker ). Ecosystem‐ and landscape‐scale approaches have increased rapidly since the 1970s. Ecosystem and landscape ecology are more focused on energy flow, nutrient cycling, habitat composition, and complex patterns on a large scale. These large‐scale and systematic data have been used to support management and conservation of related ecosystems (Dale and Beyeler ). However, Carmel et al. () noted that community and ecosystem studies constituted only a minor proportion (17%, 25%) of all studies evaluated, even considering their roles as two of the major concepts in ecology, and in the same systematic review of major ecology journals, they found that more than 60% of ecological studies still focused on a single species.It may be that research scales used by ecologists will be expanded as techniques and research networks grow. It is also evident that the relative frequency of larger‐scale research has increased, but this was not a transitional increase from the small scale to the large scale. A gap has arisen from the use of different basic units of scale, and the directions that researchers of these two groups (small‐ and large‐scale studies) pursued. Population and community ecology involves study of the structural laws that make up an ecosystem. Thus, these subfields of ecology detail demographic structure and involve many manipulated experiments using model systems, whereas ecosystem and landscape ecology are more focused on the systematic structure of biological and environmental components.To understand the structure and processes of the global ecosystem, it is necessary to integrate knowledge from both research scales. Kingsland () stressed the importance of diverse scales in studies aiming to solve ecological problems and expand our understanding of the entire biosphere. A deeper understanding of the responses of contiguous local populations and communities has accumulated over decades (Wootton ), and it is contributing to a more accurate evaluation and prediction of the impacts of global ecosystems and global change. In addition, many researchers are looking in new directions regarding community studies based on functional traits rather than individual species (Simberloff , McGill et al. ). Quantitative comparisons of the relative intensities of relationships between species and environmental factors in changing environmental contexts will enable us to coordinate with higher‐level system studies (Agrawal et al. ). Larger‐scale research topics will also require more effort to reduce the gap with population‐level research. For example, landscape ecology is particularly limited by the application of existing experimental design approaches, and new approaches are needed (Wu and Hobbs , Jenerette and Shen ). Ecological studies using individual‐based models (Shugart et al. ) and macroecology are also parts of the effort to reduce these gaps.Potential application of text mining of ecological concept studiesIn this study, we demonstrated the compatibility of text mining and questionnaire survey methods to evaluate the use of ecological concept terms. Generally, a limited number of experienced researchers can be considered as a tentative pool for the survey method (Sutherland et al. ). From this perspective, a text‐mining‐based approach has some merit as it can benefit from the rich legacy of ecological literature that has accumulated for the last 100 yr. Further, this approach can include a broader range of perspectives from the ecologist pool as peer‐reviewed papers are a standard tool for communication in the scientific community.For effective evaluation and categorization of ecological concepts, however, these two approaches should be well coordinated. A text‐mining approach would be suitable for extracting candidate ecological terms based on their frequency of occurrence and the network influence of the associated scientific publications. However, candidate terms will definitely need screening and discussion by ecologists to evaluate their relevance to ecological concepts or theories. Reiners et al. () suggested possible application of text mining to help understand the intellectual structure of the ecological conceptual pool (i.e., using term network structure for thematic classification and tracing historical use of a concept term over time for temporal trends). To further extend application of text mining of concept studies, it will be necessary to categorize and link specific ecological concepts and related terms used in the publications. As described above, it is thought that a network analysis of terms in ecological publications can be useful toward this end. In future studies, it seems it will be important to analyze differences in the use of ecological concept terms by researcher age class or geographical locality. Previous studies have also noted differences in concept ranking among age classes (Reiners et al. ). This difference in the tendency of use of ecological concepts may be reflected in the term pools from papers published by different generations of ecologists.Geographical barriers in organization networksEcological Society of America was first established with core memberships in the United States of America (96.74% of total members; Shelford ). Subsequently, individuals from more than 4000 international organizations have joined during the last 100 yr. Thus, we estimated how researchers have collaborated with others based on co‐authorship networks. Analysis of research collaboration through joint participation of organizations revealed that the majority of ecological research cooperation was geographically concentrated between limited institutions. These biased and concentrated distributions also matched with the previous findings of a study of highly cited environmental scientists (Parker et al. ). Accumulative geographic advantage of academic resources (McIntosh , Pasterkamp et al. ) and advanced infrastructure in academic fields (Basu ) may have contributed to the current unbalanced distribution of the collaboration network. However, it should be considered that our results are likely an underestimation of actual collaboration in the field of ecology. The actual collaboration network is greater than what our results imply as the current analysis only focused on the authorship network of peer‐reviewed publications in ESA journals.Collaboration of ecologists and researchers from a wide range of other disciplines (e.g., physics, chemistry, biology, geography, and many others) is essential to understanding the complexity of communities, ecosystems, evolution, and the core structures of global ecosystems. Although the specific objectives are different, LTER (Long‐Term Ecological Research), GLEON (Global Lake Ecological Observatory Network), and NEON (National Ecological Observatory Network) are examples of projects that emphasize the importance of collaboration to find insight into ecological questions (Vanderbilt et al. , Springer et al. ). In addition, the amount of ecological data (e.g., from an ecological sensor network) that is currently accumulating is too massive for any one particular group to cover and assimilate (Weathers et al. ). Studies of global ecosystems and future changes will require further international collaboration by ecologists from diverse areas. Taking this idea one step forward, this international collaboration of ecologists should be expanded to include diverse stakeholders including policy makers, decision makers, and the public (Palmer et al. ).ConclusionsWe recommend more in‐depth review of the accumulated historical data and ideas in the ecological literature. Digital archives (i.e., historical scientific data) are accessible at a low cost, and they contain massive amounts of unused information. Tangible or intangible forms of historical records from ecological fields should be conserved for future integration and in‐depth analysis. This work highlights the importance of continued and safe storage of historical ecological data. Further review of current ecological archives would provide helpful information for deciding the future direction of ecology and could be used to refine ecological concept models from a different perspective.AcknowledgmentsWe sincerely thank the enthusiastic ESA members who attended our presentation for fruitful discussion and their comments at the 100th and 101st ESA annual meetings. Ji Yoon Kim is a JSPS (Japan Society for the Promotion of Science) International Research Fellow (ID No: P17387). JSPS is a nonprofit organization.Literature CitedAgrawal, A. A., et al. 2007. Filling key gaps in population and community ecology. Frontiers in Ecology and the Environment 5:145–152.Basu, A. 2006. Using ISI's ‘Highly Cited Researchers’ to obtain a country level indicator of citation excellence. Scientometrics 68:361–375.Borrett, S. R., J. Moody, and A. Edelmann. 2014. The rise of network ecology: maps of the topic diversity and scientific collaboration. Ecological Modelling 293:111–127.Carmel, Y., R. Kent, A. Bar‐Massada, L. Blank, J. Liberzon, O. Nezer, G. Sapir, and R. Federman. 2013. Trends in ecological research during the last three decades – a systematic review. PLoS ONE 8:e59813.Cherrett, J. M. 1989. 1. Key concepts: the results of a survey of our members' opinions. Pages 1–16 in J. M. Cherrett, A. D. Bradshaw, F. B. Goldsmith, P. G. Grubb, and J. R. Krebs, editors. Ecological concepts: The contribution of ecology to an understanding of the natural world. Blackwell Scientific Publications, Oxford, UK.Chesson, P., and J. J. Kuang. 2008. The interaction between predation and competition. Nature 456:235–238.Cullen, P. W., R. H. Norris, V. H. Resh, T. B. Reynoldson, D. M. Rosenberg, and M. T. Barbour. 1999. Collaboration in scientific research: a critical need for freshwater ecology. Freshwater Biology 42:131–142.Dale, V. H., and S. C. Beyeler. 2001. Challenges in the development and use of ecological indicators. Ecological Indicators 1:3–10.Do, Y., J. Y. Kim, M. Lineman, D. K. Kim, and G. J. Joo. 2015. Using internet search behavior to assess public awareness of protected wetlands. Conservation Biology 29:271–279.Egerton, F. N. 2015. A centennial history of the Ecological Society of America. CRC Press, Boca Raton, Florida, USA.Grubb, P. J., and J. B. Whittaker. 1988. Toward a more exact ecology. Blackwell Scientific Publication, London, UK.Huver, J. R., J. Koprivnikar, P. T. J. Johnson, and S. Whyard. 2015. Development and application of an eDNA method to detect and quantify a pathogenic parasite in aquatic ecosystems. Ecological Applications 25:991–1002.Jenerette, G. D., and W. Shen. 2012. Experimental landscape ecology. Landscape Ecology 27:1237–1248.Kim, J. Y., Y. Do, R. Y. Im, G. Y. Kim, and G. J. Joo. 2014. Use of large web‐based data to identify public interest and trends related to endangered species. Biodiversity and Conservation 12:1–24.Kingsland, S. 2004. Conveying the intellectual challenge of ecology: an historical perspective. Frontiers in Ecology and the Environment 2:367–374.Ladle, R. J., R. A. Correia, Y. Do, G. J. Joo, A. Malhado, R. Proulx, J. M. Roberge, and P. Jepson. 2016. Conservation culturomics. Frontiers in Ecology and the Environment 14:269–275.Lawton, J. H. 1999. Are there general laws in ecology? Oikos 84:177–192.McGill, B. J., B. J. Enquist, E. Weiher, and M. Westoby. 2006. Rebuilding community ecology from functional traits. Trends in Ecology & Evolution 21:178–185.McIntosh, R. P. 1989. Citations classics of ecology. Quarterly Review of Biology 64:31–49.Moe, S. J., R. S. Stelzer, M. R. Forman, W. S. Harpole, T. Daufresne, and T. Yoshida. 2005. Recent advances in ecological stoichiometry: insights for population and community ecology. Oikos 109:29–39.Palmer, M. A., et al. 2005. Ecological science and sustainability for the 21st century. Frontiers in Ecology and the Environment 3:4–11.Parker, J. N., C. Lortie, and S. Allesina. 2010. Characterizing a scientific elite: the social characteristics of the world's most highly cited scientists in environmental science and ecology. Scientometrics 85:129–143.Pasterkamp, G., J. I. Rotmans, D. V. P. De Kleun, and C. Borst. 2007. Citation frequency: a biased measure of research impact significantly influenced by the geographical origin of research articles. Scientometrics 70:153–165.Pennebaker, J. W., M. R. Mehl, and K. G. Niederhoffer. 2003. Psychological aspects of natural language use: our words, our selves. Annual Review of Psychology 54:547–577.Perring, M. P., R. J. Standish, J. N. Price, M. D. Craig, T. E. Erickson, K. X. Ruthrof, A. S. Whiteley, L. E. Valentine, and R. J. Hobbs. 2015. Advances in restoration ecology: rising to the challenges of the coming decades. Ecosphere 6:1–25.Pickett, S. T. A., M. L. Cadenasso, D. L. Childers, M. J. McDonnell, and W. Zhou. 2016. Evolution and future of urban ecological science: ecology in, of, and for the city. Ecosystem Health and Sustainability 2:e01229.Reiners, W. A., J. A. Lockwood, S. D. Prager, and J. C. Mulroy. 2015. Ecological concepts: What are they, what is their value, and for whom? Bulletin of the Ecological Society of America 96:64–69.Reiners, W. A., J. A. Lockwood, D. S. Reiners, and S. D. Prager. 2017. 100 years of ecology: What are our concepts and are they useful? Ecological Monographs 87:260–277.Rind, D., R. Goldberg, and R. Ruedy. 1989. Change in climate variability in the 21st century. Climatic Change 14:5–37.Shelford, V. E. 1917. The ideals and aims of the Ecological Society of America. Bulletin of the Ecological Society of America 1:1–8.Shelton, A. O., J. L. O'Donnell, J. F. Samhouri, N. Lowell, G. D. Williams, and R. P. Kelly. 2016. A framework for inferring biological communities from environmental DNA. Ecological Application 26:1645–1659.Shugart, H. H., G. P. Asner, R. Fischer, A. Huth, N. Knapp, T. Le Toan, and J. K. Shuman. 2015. Computer and remote‐sensing infrastructure to enhance large‐scale testing of individual‐based forest models. Frontiers in Ecology and the Environment 13:503–511.Simberloff, D. 2004. Community ecology: Is it time to move on? (An American Society of Naturalists presidential address). American Naturalist 163:787–799.Springer, Y. P., et al. 2016. Tick‐, mosquito‐, and rodent‐borne parasite sampling designs for the National Ecological Observatory Network. Ecosphere 7:e01271.Sutherland, W. J., et al. 2013. Identification of 100 fundamental ecological questions. Journal of Ecology 101:58–67.Thompson, J. N., et al. 2001. Frontiers of ecology. BioScience 51:15–24.van Dongen, S., and C. Abreu‐Goodger. 2012. Using MCL to extract clusters from networks. Methods in Molecular Biology 804:281–295.Vanderbilt, K. L., C. C. Lin, S. S. Lu, A. R. Kassim, H. He, X. Guo, I. S. Gil, D. Blankman, and J. H. Porter. 2015. Fostering ecological data sharing: collaborations in the International Long Term Ecological Research Network. Ecosphere 6:1–18.Waltman, L. R., N. J. P. Eck, and E. C. M. van Noyons. 2010. A unified approach to mapping and clustering of bibliometric networks. Center for Science and Technology Studies, Leiden, The Netherlands.Weathers, K. C., et al. 2016. Frontiers in ecosystem ecology from a community perspective: The future is boundless and bright. Ecosystems 19:753–770.Wootton, J. T. 1996. The nature and consequences of indirect effects in ecological communities. Annual Review of Ecology, Evolution and Systematics 25:443–466.Wu, J., and R. Hobbs. 2002. Key issues and research priorities in landscape ecology: an idiosyncratic synthesis. Landscape Ecology 17:355–365.

Journal

Ecosphere – Wiley

Published: Jan 1, 2018

Keywords: ; ; ; ;

There are no references for this article.