Unveiling the Knowledge Structure and Emerging Research Trends of Google Earth Engine Applications in Remote Sensing: A Bibliometric Analysis of Scopus-Indexed Publications (2021–2026)

Abstract

The convergence of the Random Forest (RF) ensemble classifier with the Google Earth Engine (GEE) planetary-scale computing platform has become one of the dominant operational paradigms in contemporary Earth observation. Despite the rapid accumulation of literature, no systematic bibliometric account of this specific intersection has been published. This study analyses 107 Scopus-indexed journal articles published between 2021 and 2026 that jointly apply RF and GEE to remote sensing problems. Using the bibliometrix R package and VOSviewer, the corpus was examined through descriptive indicators, performance analysis, and five science-mapping techniques: co-authorship, keyword co-occurrence, co-citation, bibliographic coupling, and thematic mapping. Results show a compound annual growth rate of 51.97% between 2021 and 2025, 1,685 total citations, an h-index of 24, and a highly concentrated publication landscape in which three journals Sustainability, International Journal of Digital Earth, and Ecological Informatics account for 59.8% of output. China dominates production with 48 affiliated documents, followed by the United States (20). Co-citation analysis identifies a compact intellectual core anchored by Gorelick et al. (2017), Breiman (2001), and Belgiu and Drăguţ (2016). Thematic analysis reveals a field that is methodologically mature but conceptually narrow: RF is applied across eight application domains, yet only 10.3% of documents address model transferability, 4.7% address uncertainty quantification, and 4.7% address reproducibility. Emerging fronts include explainable AI (SHAP), SAR–optical fusion, and deep-learning hybridisation. The study concludes by articulating seven research gaps and a corresponding agenda for the next phase of cloud-based Earth observation research.