Rapid global urbanization has generated a vast interdisciplinary literature on urban expansion and green space dynamics, creating challenges for integrated knowledge synthesis and understanding of urban sustainability transitions. Traditional keyword co-occurrence-based bibliometric reviews are limited in capturing semantic structure and temporal dynamics within scientific literature. To address this challenge, we conducted a semantic-temporal scientometric analysis of 13,001 urban expansion and green space publications indexed between 2000 and 2025. Twelve transformer-based embedding models were benchmarked within a hybrid semantic clustering framework, with BGE-large-en-v1.5 demonstrating the strongest overall performance across four diagnostic metrics. The resulting keyword network reveals a polycentric knowledge structure organized around three cores: ecological foundations, urban planning, and observation and modelling. Twenty-three semantic knowledge clusters were identified and aggregated into five higher-order meta-themes. While foundational ecology remains structurally stable, the field has undergone a post-2015 dual transformation marked by rapid expansion in data-driven observation and modelling and a normative reorientation toward nature-based solutions, human well-being, and environmental justice, reflecting broader international sustainability priorities. Rather than displacing existing domains, the field evolves through semantic reweighting and the integration of established concepts, with normative agendas exhibiting substantially longer incubation periods than methodological clusters. These findings provide a lifecycle-oriented analytical framework for urban planners, environmental agencies, and research funding bodies — with targeted implications for revitalizing saturating foundational clusters, building coordinated open-data infrastructures for data-intensive domains, and structurally embedding equity metrics within nature-based solution planning from the outset, in direct support of SDGs 10, 11, and 15.
Suman et al. (Wed,) studied this question.