迈向人机协同的生态学知识生产新范式:青藏高原草地植被群系图绘制跨学科方法论透视OA

Artificial intelligence(AI)is increasingly transforming ecological research by enabling large-scale environmental observation,complex data analysis,and automated pattern recognition.Although numerous studies have demonstrated the technical advantages of AI in vegetation mapping,biodiversity monitoring,and ecological modeling,considerably less attention has been paid to how AI reshapes the organization and generation of scientific knowledge.Existing discussions primarily focus on algorithmic performance or ecological applications,while the underlying knowledge production process remains insufficiently understood.Using the development of the 1:500000 Vegetation Formation Map of Grasslands on the Qinghai-Tibetan Plateau as a representative case,this study investigates how ecological theory,remote sensing observations,and AI are integrated throughout the research process.Rather than evaluating AI as merely a computational tool,this paper aims to explain its role in interdisciplinary ecological research and to clarify how human researchers and intelligent systems jointly contribute to scientific knowledge generation. This study adopts a qualitative case study approach combining technical process analysis with science and technology studies.The complete workflow of vegetation mapping was reconstructed,including field investigation,vegetation classification,remote sensing data acquisition,deep learning-based classification,vegetation map production,ecological interpretation,and scientific validation.The analysis focuses on how ecological knowledge is embedded into AI-assisted mapping through vegetation classification systems,sample labeling,variable selection,and model construction.Theoretical interpretation draws upon interdisciplinary research theory,Mode 2 knowledge organization,and the Fourth Paradigm of data-intensive scientific discovery.Rather than treating these theories as alternative explanations,they are employed to analyze different dimensions of the research process,namely interdisciplinary capability integration,changes in research organization,data-driven scientific discovery,and the emergence of Human-AI collaborative knowledge production. The case demonstrates that the production of the Qinghai-Tibetan Plateau vegetation formation map represents more than a technological achievement in ecological mapping.First,the mapping task exceeded the capability of any individual discipline because it simultaneously required ecological classification knowledge,regional-scale remote sensing observations,and AI-based high-dimensional data analysis.These heterogeneous capabilities were integrated into a unified research framework to accomplish regional vegetation mapping.Second,the study reveals a distinctive mechanism of in-model knowledge integration.Ecological knowledge was not only involved in defining research questions and interpreting final results,but was continuously embedded into the AI workflow through vegetation classification systems,sample labeling,feature selection,and model constraints.Consequently,ecological expertise became part of the computational learning process rather than remaining external to it.This finding provides a concrete explanation of how interdisciplinary knowledge integration can occur within AI-supported scientific research.Third,AI functioned neither as a simple analytical tool nor as an autonomous scientific agent.Instead,it served as a knowledge intermediary connecting ecological theory with remotely sensed observations.By learning the correspondence between field survey data and multi-source environmental variables,the AI model extended localized ecological knowledge to regional spatial scales,enabling the identification of vegetation formations across the entire Qinghai-Tibetan Plateau.However,AI-generated spatial patterns did not automatically become scientific conclusions.They required subsequent ecological interpretation,comparison with historical vegetation maps,long-term observations,warming experiments,and ecological process studies before meaningful scientific explanations could be established.Finally,the case exhibits a dual-loop pathway of scientific discovery.Unlike the conventional hypothesis-driven paradigm,in which theories generate hypotheses subsequently verified by observations,important scientific findings in this study first emerged from AI-assisted analysis of massive ecological datasets.Newly identified vegetation changes subsequently stimulated ecological explanation and causal investigation.Data-driven pattern discovery generated new scientific questions,whereas theory-driven interpretation established ecological mechanisms,together forming a dual-loop pathway of scientific discovery. This study argues that AI is reshaping ecological research not by replacing scientists but by reorganizing the relationships among theory,observations,and computation.The Qinghai-Tibetan Plateau vegetation mapping project illustrates a new mode of Human-AI collaborative knowledge production characterized by interdisciplinary capability integration,in-model knowledge integration,AI functioning as a knowledge intermediary,and a dual-loop pathway linking data-driven discovery with theory-driven explanation.At the same time,this collaborative mode depends upon several essential conditions,including mature theoretical frameworks,high-quality observational datasets,and continuous human interpretation.Its significance therefore lies not in demonstrating that AI can replace scientific reasoning,but in revealing how intelligent systems can become integral participants in scientific knowledge generation while maintaining the indispensable role of human expertise.These findings contribute to a deeper understanding of methodological transformation in ecological research and provide a conceptual framework for analyzing AI-enabled scientific knowledge production in other data-intensive disciplines.