The authors have declared that no competing interests exist.
Current address: Centre for Environment and Climate Research, Lund University, Lund, Sweden
We analyze how the content of ecosystem service research has evolved since the early 1990s. Conducting a computational bibliometric content analysis we process a corpus of 14,118 peer-reviewed scientific article abstracts on ecosystem services (ES) from Web of Science records. To provide a comprehensive content analysis of ES research literature, we employ a latent Dirichlet allocation algorithm. For three different time periods (1990–2000, 2001–2010, 2011–2016), we derive nine main ES topics arising from content analysis and elaborate on how they are related over time. The results show that natural science-based ES research analyzes oceanic, freshwater, agricultural, forest, and soil ecosystems. Pollination and land cover emerge as traceable standalone topics around 2001. Social science ES literature demonstrates a reflexive and critical lens on the role of ES research and includes critiques of market-oriented perspectives. The area where social and natural science converge most is about land use systems such as agriculture. Overall, we provide evidence of the strong natural science foundation, the highly interdisciplinary nature of ES research, and a shift in social ES research towards integrated assessments and governance approaches. Furthermore, we discuss potential reasons for observable topic developments.
The ecosystem services (ES) concept was developed in the 1970s and 1980s by conservation biologists and ecological economists to encourage decision-makers to recognize and attend to socio-ecological linkages [
The ES concept thus encompasses and bridges several disciplines under the umbrella of sustainability science. Research stemming from the natural and social sciences to the humanities contribute to a myriad of interpretations and applications of the ES concept [
The highly interdisciplinary and socially constructed nature of the ES framework invites a series of ontological and epistemological challenges [
This methodological challenge provides the impetus for our analysis, which aims to capture the internal diversity of ES research over time. Our research question is:
The structure of the article is the following. We present related literature in section 2, expound our methodology in section 3, provide the results in section 4, and discuss topic developments in section 5. In section 6, we conclude briefly.
Several reviews on ES exist, and we employ several of their insights to structure our analyses and relate our results to their findings. Gómez-Baggethun et al. [
Abson et al. [
Chaudhary et al. [
Furthermore, there is an ongoing conceptual development in ES research. Chan et al. [
We conducted a bibliometric analysis of the scientific literature dealing with the concept of ES. We retrieved Web of Science (WoS) core collection (all years, by topic) data using the string “ecosystem service*” (
In order to outline the evolution of the ES literature over time, we downloaded the WoS entries for three periods 1990–2000, 2001–2010, and 2011–2016. This choice corresponds broadly to the phases of ES research development from Chaudhary et al. [
Years | Records available | Final dataset |
---|---|---|
1990–2000 | 136 | 108 |
2001–2010 | 2,719 | 2,521 |
2011–2016 | 12,183 | 11,489 |
All years | 15,038 | 14,118 |
The final data set excludes records with empty abstracts and double entries.
The following analyses were performed within the
In the unsupervised process of learning the set of topics LDA employs a multinomial dirichlet distribution and infers a posterior distribution through a variational Bayes approximation [
The results of the LDA analysis can be explored interactively with a web-based visualization (
The descriptive statistics plots in this section are based on an analysis of the entire data set.
Colour scales represent the count of author affiliation locations. Note the different fixed width logarithmic colour scale breaks.
Note the different x-axis-scale of the plots for each period.
For each of the periods, the LDA algorithm provides nine topics with a probability distribution over words. Here, we describe the topics in more generic terms. A graphical overview of the topics for each period, their relative size, and their development can be found in
The height of the topic boxes represents the relative topic proportion within each period. Note that the number of assessed articles increases over time (1990–2000: N = 108, 2001–2010: N = 2,521, and 2011–2016: N = 11,489). The links between periods have interpretatively been deducted from LDA results. There is no particular order of topics except for clarity of linkage exposition.
References to the 20 most representative, as in most probable, articles for each of the topics and a table of the assigned topics for all articles per period can be found in the supporting information (
Topic | Thematic focus | Methods |
---|---|---|
Agriculture | Food production, provisioning services, conservation, payments for ecosystem services, carbon sequestration, policy measures | Economic valuation (willingness to pay / accept), science and technology studies, trade-off analysis, policy evaluation, scenario methods |
Assessment | Ecosystem services assessments, science-policy interfaces of national ecosystem service assessments and planning procedures | Mapping techniques (spatially explicit) ecosystem service flow models, participatory (scenario) planning, Bayesian belief networks, |
Conservation | Biodiversity loss, (regional) environmental change, climate change, agricultural landscapes, agroforestry, management strategies | Impact / effectiveness studies, (financing) strategy evaluations, driver and impact analyses, spatio-temporal and functional assessments |
Ecosystem functions | Ecosystem composition and fragmentation, species communities and richness, terrestrial ecosystems above and below ground, | Experiments, field study sampling, meta studies, species distribution studies, functional assessments, process and (material) flow analyses |
Forests | Management practices, land use / cover change, species composition, forest fires, carbon storage, urban green space, alien species | Disturbance recovery studies, impact evaluations, land cover assessments, longitudinal studies, satellite imagery |
Freshwater | Hydrology, retention services, flow regimes, sediment streams, wetland and marsh management, material discharge and uptake, hydropower | Hydrological models, GIS analyses, physical, chemical and biological analyses, blue, green, gray water assessments, morphological studies |
Global awareness | Total economic value, nitrogen retention and drainage, (sea)food production, reefs, human demand, ecological thresholds, natural capital | Economic valuation, GIS analyses, large scale biophysical-monetary assessments, quantitative evaluations, |
Governance | Conservation, management, sustainable development, policies, research, social change, challenges, adaptation, environment | Network analyses, legitimacy and accountability studies, system analyses, transdisciplinary research, policy evaluation |
Land cover | Land use change / cover models and scenarios, sensors, surrogates, accuracy, resolution, spatial data, classification, patterns | Satellite imagery and remote sensing, spatially explicit analyses, geo-spatial assessments, maps, simulations, lidar |
Marine | Habitats, coral reefs, mangroves, fishery, climate change, functional diversity, species composition, extinctions | Functional assessments, genetic studies, species distribution analyses, diversity assessments, global change analyses |
Pollination | Supporting services, inter-insect relations, predators, biological and chemical pest-control, habitat structure, plant communities | Field samples, management intensity and type studies, crop / harvest assessments, diversity / composition evaluations |
Risk management | Ecological processes, ecosystem services, stressors, man-made disturbance, ecological resilience, response management | Spatio-temporal ecosystem models, risk assessments, toxicity tests, management studies, system-wide driver-impact evaluations |
Role of science | Research, policy, sustainability, frameworks, social change, systemic analyses, knowledge, practitioners, communication, civic engagement | Transdisciplinary research, socio-ecological models, qualitative studies, knowledge classifications, conceptual studies |
Soils | Soil structure, species community, functional diversity, carbon storage, crop production, bioenergy, feedstock, micro-biota, fertilizer | Bio-geo-chemical analyses, diversity assessments, functional assessments, cropping system analyses, land management studies |
Sustainability management | Development, restoration, humans, socio-ecological systems, capital, planetary perspectives, future, life support, paradigms | Conceptual studies, system boundary assessments, history to future extrapolations, capital accounting, response formulation |
Urban | Human livelihood, urban sprawls, green spaces, parks, recreation, health, planning, economic value, rural dwellings, households, tourism | Land cover assessments, GIS analyses, micro-climate models, economic valuation, urban planning, surveys, cultural evaluations |
Valuation | Methods, (non-)market goods and services, value concepts, aggregation, discounting, accounting, benefits, costs | Revealed preference methods, contingent valuation methods, cost-benefits analyses, total economic value aggregation, surveys |
Source: Author’s elaboration based on from Latent Dirichlet Allocation analysis results of WoS data.
We used the results of the LDA analyses from each period to assess the content of each topic descriptively. We then tracked topic development through collaborative and iterative qualitative analyses of the content (see section 3.2). In order to provide transparency about the corresponding qualitative process,
Linkages | Shared stemmed key terms across different λ | ||
---|---|---|---|
Agriculture | → | Governance | agricultur, agrobiodivers, biodivers, benefit, conserv, develop econom, ecosystem, environment, incent, landhold, manag, market, payment, pes, polici, rancher, resourc, servic, sustain, system |
→ | Pollination | agricultur, biodivers, increas, farm, manag, provid, servic | |
→ | Soils | agricultur, bionergi, carbon, ecosystem, emiss, energi, farm, ghg, land, product, | |
Conservation | → | Agriculture | agricultur, biodivers, conserv, ecosystem, land, polici |
→ | Role of science | biodivers, chang, conserv, ecosystem, natur, process, polici | |
Ecosystem functions | → | Pollination | abund, comuniti, plant, pollin, rich, speci |
→ | Soils | biomass, function, grassland, plant | |
Forests | → | Urban | area, citi, develop, ecosystem, green, landscap, natur, park, protect, urban, space |
Global awareness | → | Freshwater | freshwat, wetland |
→ | Marine | coral, fish, global, marin, popul, ocean | |
→ | Role of science | conserv, ecosystem, function, human, natur, servic | |
→ | Valuation | account, costanza, estim, reserv, servic, valu, yuan | |
Land cover | → | Assessment | approach, assess, base, ecosystem, indic, inform, method, model, spatial, studi, |
→ | Urban | area, ecosystem, result, servic, studi, | |
Risk management | → | Role of science | address, develop, ecolog, ecosystem, human, integr, manag, problem, process, scientist, servic, societi, system |
Role of science | → | Assessment | approach, ecolog, ecosystem, framework, integr, provid, servic, system |
→ | Governance | approach, biodivers, challeng, conserv, develop, ecolog, ecosystem, environment, human, knowledg, local, manag, natur, neoliber, polici, research, resourc, scienc, servic, social, sustain, system | |
Sustainability management | → | Role of science | develop, ecolog, manag, natur, servic, sustain, system |
Valuation | → | Assessment | assess, base, benefit, ecolog, ecosystem, emergi, environment, evalu, method, provid, servic, studi, valu, valuat, |
Source: Author’s elaboration based on LDA analysis of WoS data.
In the first period, from 1990–2000, five out of nine topics mainly deal with ecology and land use (conservation, forests, ecosystem functioning, freshwater, and marine) while four topics at address social issues and practices (valuation, global awareness, risk management, and sustainability management). A core topic in this time period is dedicated to larger-scale societal impacts on ecosystems and economies, which we thus labeled global awareness. Concepts in the global awareness topic branch into different topics in later periods, such as valuation, the role of science, and freshwater ecosystems. We furthermore find that the stream of research with a focus on sustainability, risk management and global awareness precedes the reflective role of science for sustainability transformations topic. The second period (2001 to 2010) shows an increase in the share of natural science-related topics, with foci on forests, soils, pollination, freshwater and marine ecosystems. Pollination occurs as an individual topic in the second period and becomes an important (policy) issue globally, as reflected by the recent assessment by Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services [
This article is based on a reproducible quantitative analysis of the abstracts of ES literature available in the Web of Science, from which we qualitatively evaluate the evolution of research topics over time. We aimed to capture and display the internal, thematic diversity of ES research through an innovative method, and thus we focus our discussion on both the content and development of topics in ES research (section 5.1) and the value and limitations of the LDA method (section 5.2).
We find that most of the topics emerging from our LDA analysis of ES literature since 1990 align with the topics identified by Abson et al. [
With regard to the topic development over time, we structured periods broadly among the phases identified by Chaudhary et al. [
Our analysis corresponds broadly to observations already made, suggesting the potential for LDA to complement other review-efforts as research publications grow and the ES topic continues to expand across disciplines. We contribute two important insights that are supported by quantitative evidence through our LDA analysis. We observe:
a substantial natural science foundation of ES research with a high volume of publications as well as a highly interdisciplinary ES community with important contributions from social science. However, we do not find evidence that there is a major shift towards market based instruments in the greater ES literature corpus. We also note some critical and self-reflective perspectives among the social ES science community, which perhaps contributed to the waning persistence of monetary valuation research for ES;
a convergence of several disciplinary perspectives (including economic valuation and land use topics) into an integrated assessment topic that combines various approaches and (e-)valuation methods. Similarly, several topics (including agricultural and role of science topics) influence a broad governance topic that includes analyses of market-based, governmental and bottom-up participatory approaches for ES management.
We would thus argue that the combined contributions of natural and social science make ES research a highly interdisciplinary field which relies on cross-fertilization and a corresponds to an evolutionary change in topics among research communities.
The main value added of the LDA technique is that key terms and topics within a literature’s entire corpus emerge based on the data itself rather than the a priori postulations formulated by researcher(s) through their analytical processes. The unsupervised learning algorithm clusters documents among topics based on their conditional distribution of words. This allows the data to “speak” for itself through conditional word occurrence probabilities. Our own contribution beyond applying the method to ES research has been to track topic development over time though interpretative linkages while holding the number of topics constant. To our knowledge, this is the first fully reproducible unsupervised machine learning algorithm analysis on ES research and link this to an interpretive time series analysis.
Previous review papers summarizing the state or evolution of ecosystem service literature have focused primarily on either top cited publications, e.g. papers with > 15 publications [
There are limits to our strategy. Our data only includes scientific literature published in English, thus excluding, for instance, grey literature and policy documents, or publications written in languages other than English. Furthermore, we were unable to perform any preliminary screening of the collected articles given the size of the dataset to verify for relevance and adherence with the ES concept. It is thus possible that an undefined portion of the literature mentions “ecosystem service(s)” as a buzzword or post-hoc justification for research, as suggested by Abson et al. [
The linkages across periods were developed through interpretative analysis by all three authors. While the LDA analysis of topics for each period is fully reproducible, the interpretation of how these topics link and develop over time may not be, although we provided a list of key terms that we used to track topic development for transparency.
A next step in topic modeling regarding ES research could be the application of dynamic topic models and the influence model [
We have analyzed the Web of Science core collection for the search term “ecosystem service*”, resulting in a dataset of over 14,000 articles We used a computational science method, latent Dirichlet allocation (LDA) analysis to derive main topics from the articles’ abstracts. We analyzed three periods of ES research, from 1990 to 2016 and qualitatively linked the topics between the periods in order to display research (dis-)continuities.
LDA analysis allowed us a broad and reproducible exploration into ES research content. Through a combination of LDA with qualitative interpretation of topic linkages across periods, we evaluated the question: what are the main topics in ES research and how do they evolve over time? A slight majority of topics contain natural science research on different ecosystems such as oceans, freshwater, soil, pollination or forests. Topics such as pollination, land cover, and an urban topic appear over time. Some topics are at the junction of socio-ecological land use systems (land cover, agriculture, forests, urban spaces). A smaller share of topics is based on social science approaches such as sustainable management, the role of science, valuation, and governance. This provides a counterpoint to former analyses who find a dominant share of economic and social science research in the most cited literature [
Similar to an analysis by Gómez-Baggethun et al. [
Overall, we observe a strong natural science foundation, a growingly interdisciplinary agenda, and a notable policy-orientation of ES research. By employing a novel combination of both quantitative and qualitative content analysis methods, we contribute comprehensive evidence of an increasingly multi-faceted and integrated assessment methodology, an evolving recognition of multiple types of values in the valuation area of ES topics, and thus indications of a growing diversity of responses and instruments to meet conservation needs among ES research communities.
(DOCX)
(XLS)
(XLS)
(XLS)
(ZIP)
(ZIP)
(ZIP)
All three authors are grateful to Arild Vatn and his Thor Heyerdahl Summer School for forging the initial collaborations, and to the academic editor and the anonymous reviewers who constructively helped to improve the manuscript.