Data science in ecosystem-service research: A CICES-based bibliometric mapping and decision-support synthesis

Data science (DS) methods have attracted growing attention in ecosystem-services (ES) research, yet their bibliographic visibility across service categories and the implications of this pattern for environmental management remain insufficiently understood. This study combines bibliometric analysis, CICES-based classification, association-rule analysis, and decision-oriented synthesis to examine how DS methods are represented across ES domains. Bibliographic records were retrieved from the Web of Science Core Collection for 2015-2025. After LLM-assisted relevance screening, CICES-based classification, and manual validation, 53,227 ES publications were retained. We analyze temporal publication trends, geographical and institutional patterns, collaboration networks, service-domain distributions, and the bibliographic visibility of DS terms in titles, abstracts, author keywords, and Keywords Plus. Association-rule analysis was used to identify recurring ES theme-DS method associations using appearance, lift, and confidence. The results show sustained growth in ES publications and geographically uneven research output. DS-related terms were identified in 9.56% of publications under the broad bibliographic-field indicator and in 3.17% under the keyword-field indicator. Their visibility increased over time but remained concentrated in established methods. Random forest and principal component analysis were prominent among the highest-scoring theme-method associations, whereas deep learning, generative models, and explainable artificial intelligence appeared less prominently. The results suggest that established DS can support prediction, classification, and dimensionality reduction, but effective decision support requires validation, uncertainty characterization, interpretability, representativeness assessment, and accountability safeguards. The findings highlight the need to align ES questions, analytical tasks, data structures, validation strategies, management scales, and intended uses when designing DS applications for environmental management.