Abstract:

Aims: The development of China’s national botanical garden system has made digitalization and artificial intelligence (AI) increasingly important for enhancing the effectiveness of ex situ plant conservation. However, AI applications in many botanical gardens remain fragmented, with limited system-level design and few quantitative tools for capability assessment. This study aims to address this gap by constructing a comprehensive AI capability reference model, an associated maturity assessment framework, and an application architecture tailored to the national botanical garden system.
Methods: Grounded in practical experience at Shanghai Chenshan Botanical Garden, we developed the national botanical gardens-AI capability reference model (NBG-AICRM). The model follows a bottom-up architecture that links infrastructure support, data acquisition, data governance, intelligent analytics, and scenario-based applications. It consists of five logical layers—infrastructure, data sensing, data management, intelligent analytics, and application scenario—which together cover 25 key capability items and several extended items. Each capability item is assessed using a six-level maturity scale from Level 0 to Level 5. Based on this framework, we formulated an application-scenario-driven reverse architecture design method and a “Smart Conservation Matrix”, a nine-grid strategic framework that maps conservation priorities to AI solutions. To validate the architectural guidance of the model, we built the smart plant attendant (SPA), a large language model-based system that uses Text-to-SQL technology to support natural-language-driven, management-oriented analysis of conservation data. The system was trialed at Shanghai Chenshan Botanical Garden.
Results: The NBG-AICRM provided a structured diagnosis of AI maturity levels, identified specific capability gaps in infrastructure and algorithm application, and outlined a phased upgrade path. The SPA system converted natural language queries into executable SQL statements, enabling conservation staff to retrieve and analyze conservation-relevant information—such as living-collection status trends and the representation of endangered species—without writing code. The initial deployment demonstrated that the model can help address the planning gap in the AI-enabled transformation of botanical gardens and strengthen the linkage among heterogeneous data resources, AI technologies, and day-to-day conservation operations.
Conclusion: The NBG-AICRM and its companion design methods offer a unified and replicable reference framework for building AI capabilities across China’s national botanical garden system. By formalizing maturity assessment and application architecture, this approach provides both a theoretical paradigm and a practical pathway for integrating data, intelligence, and conservation operations, thereby supporting the digital transformation of plant ex situ conservation. It may also provide a basis for coordinating future AI investment, capability planning, and policy support within the national botanical garden system.

Key words: national botanical garden, artificial intelligence (AI), reference model, capability maturity, intelligent conservation, large language model (LLM)