Unearthing trends in environmental science and engineering research: Insights from a probabilistic topic modeling literature analysis

Academic contributions in environmental science and engineering (ESE) research are needed to ensure a cleaner, productive, and environmentally conscious society. Hence, an understanding of the critical trends, topics, and research developments in the field is crucial towards facilitating the identification, communication, and improved research collaboration of nascent and increasingly complex environmental problems. As a deeper analysis of the broader trend evolution of ESE research is lacking in the literature, we employ a more robust content analysis approach in the form of a topic modeling computational text analysis method to unearth key temporal and regional insights in the field. As such, we apply a latent Dirichlet allocation (LDA) model on abstract metadata based on 3572 articles procured from reputable journals that published subject matter related to ESE research from 2005 to 2019 inclusive. We analyze the statistical composition of each inferred topic in the form of word clouds and uncover general research trends, such as topics related to environmental impact assessments, improved clean cookstoves, solid waste management, and environmental lead pollution. We also perform temporal analysis experiments across each respective journal and observe a high degree of consistency and variance among topic focuses. Moreover, whilst quantifying trends at the regional level, we detect that certain countries display clearly discernible patterns, suggesting that research communities in ESE from various countries tend to focus on different sub-fields. The main contribution of our work is the application of a more refined computer-assisted content analysis method in ESE trend analysis research that can serve as the foundation for future exploratory trend analysis investigations in ESE and other related fields. •Topic modeling is used to reveal trends in environmental science and engineering.•A unique corpus of high-quality relevant abstracts is formed to extract topics.•Latent Dirichlet allocation (LDA) is applied to find topics and research themes.•Temporal changes in topics' emphasis over the past fifteen years are disclosed.•Unique trends across different regions and countries are revealed and discussed.