Digital Return to Nature: Mapping the evolution of AI in landscape ecology

Artificial intelligence (AI) is increasingly applied in landscape ecology (LE) to map land cover, quantify landscape patterns and configuration, model ecological connectivity, and assess disturbance regimes. However, the structural evolution of this emerging integration and its dominant research themes remain insufficiently mapped through quantitative, data-driven analyses. To address this gap, we synthesize the evolving interface between AI and LE through bibliometric analysis and transformer-based semantic topic modeling on a curated corpus of 1387 publications indexed in Scopus and the Web of Science Core Collection. Results reveal rapid growth in AI-enabled LE research in recent years, a modest but statistically significant positive association between team size and citation impact, and a steady increase in average authorship over time. Integrating evidence from keyword co-occurrence networks and transformer-based topic modeling, we observed a Digital Return to Nature pattern, indicating that despite the methodological transition toward deep learning, AI research continues to focus on core ecological themes, including forest ecosystems, biodiversity, habitat connectivity, and land-system dynamics. These findings highlight the growing role of AI in biodiversity conservation, ecosystem monitoring, urban planning, and land-system analysis, creating new opportunities for AI-driven decision support in adaptive landscape management and evidence-based environmental policy.