“坤元”是地理科学迈向智能时代的重要里程碑，为中国构建自主可控的地理认知系统奠定了基础。

The rapid development of artificial intelligence (AI) is propelling scientific research toward the fifth paradigm, characterized by intelligence−driven inquiry, following the earlier paradigms of experiment, theory, computation, and data−driven science. This paper reviews the historical evolution of research paradigms, analyzes the logic of transformation from "human−centered" to "human–AI collaborative" research, discusses the development and current state of the large model era, and examines the applications of AI in research management, hypothesis generation, and academic writing and projects. The findings indicate that AI not only reshapes research processes and improves efficiency but also facilitates cross−disciplinary knowledge recombination and innovation, gradually forming an intelligent research ecosystem centered on large models and generative AI. At the same time, challenges such as hallucinations, poor interpretability, and ethical risks cannot be overlooked. Accordingly, this paper proposes building a trustworthy AI system for scientific research, advancing innovative models of human–AI collaboration, improving governance and policy frameworks, strengthening interdisciplinary integration, and enhancing mechanisms of research ethics and social responsibility, in order to ensure the sustainable development and orderly evolution of AI−driven research paradigms.

With the development of wireless communication and artificial intelligence technology, the number of small mobile devices is increasing rapidly, and the traditional wired power supply model can no longer meet people's requirements for portability and mobility. Radio frequency and microwave wireless energy transfer technology can get rid of the limitation of wired power, and has a greater application potential in wireless devices. Metasurface is a new kind of artificial material, which has unique advantages like regulating electromagnetic parameters and two−dimensional compact conformal. It is expected to play a key role in radio frequency and microwave wireless energy transfer. In wireless energy transfer systems, metasurface can generate energy beam at the transmitting terminal, enhance coupling resonance in the transmission path, and improve AC−DC conversion efficiency at the receiving terminal. So this paper took stock of the hot spots of this technology, and introduced the advanced scientific progress of wireless energy transfer integration wireless communication, wireless sensing, reconfigurable intelligent surface and target recognition & localization. The hot spots of wireless energy transfer metasurface technology, including wireless power transfer, wireless energy harvesting, simultaneous wireless information and power transfer, and reconfigurable wireless energy transfer technology.

Geographic Intelligence System (GeoIS), an emerging technological framework that integrates geographic science and artificial intelligence, is rapidly becoming a key driver for the reconstruction of spatial cognition and intelligent spatiotemporal decision−making. This paper systematically reviews the development trajectory of GeoIS and the latest international research advances. Focusing on the three core capabilities of perception, analysis, and decision−making, it identifies major shortcomings in China's GeoIS ecosystem—particularly in sensor development, algorithmic foundations, platform engines, and data governance. Based on this analysis, the study proposes a development pathway centered on "core technological breakthroughs–cross−domain integration–application−driven scenarios", and offers policy recommendations including standards development, security governance, and institutional support. The goal is to provide a reference for building an autonomous, secure, and co−evolving GeoIS system.

In recent years, large models have made breakthroughs in natural language processing and computer vision by virtue of their powerful sequence modeling capability, excellent representation learning potential, and flexible pre-training–fine-tuning paradigm, which also bring new development opportunities for time-series and spatio-temporal data intelligent analysis and are expected to revolutionize the analysis paradigm. This paper provides a systematic review of research progress on large models for time series and spatio-temporal data analysis, focusing on two major directions: empowering large language models and building dedicated foundation models. The former leverages prompt engineering, tokenization, and parameter-efficient fine-tuning to adapt large models to time series and spatio-temporal tasks, while the latter employs large-scale cross-domain pre-training to establish unified dynamic representations. The path of empowering large language models offers advantages such as low development costs and flexible zero-shot/few-shot transfer learning, while specialized foundational models demonstrate superior cross-domain generalization capabilities. At the same time, both approaches still face challenges including insufficient interpretability and difficulties in multimodal semantic alignment. Future research urgently requires breakthroughs in enhancing interpretability, advancing multimodal joint modeling, and innovating model architectures.

Object−oriented modeling is a fundamental paradigm in the field of geographic information system (GIS), offering distinct advantages in representing spatial structures, semantic relationships, and dynamic behaviors. This paper first provides a systematic review of the theoretical foundations and developmental trajectory of object−oriented spatial modeling, highlighting its central role and evolution within spatial cognition modeling frameworks. It then analyzes the technical progress and practical applications of this approach in key scenarios such as 3D urban modeling, watershed management, and landslide analysis. In response to the spatial intelligence demands of the foundation model era, this study proposes a pathway to empower geographic cognition through object−oriented spatial modeling, focusing on four key aspects: structured spatial semantic embedding, computable spatial relational logic, graph−based cognitive execution frameworks, and unified intermediate representation languages. The findings suggest that integrating object−oriented cognitive structures and dynamic process models can significantly enhance the capabilities of pre−trained models in spatial semantic understanding, causal reasoning, and complex task execution. This research provides a theoretical perspective and key element analysis for integrating GIS modeling with large model technologies, offering conceptual support and an exploratory framework for building spatially intelligent foundation models.

This article provides an overview of the domestic and international development of remote sensing multidimensional data formats, especially multidimensional spatiotemporal spectral data, the latest research results, current technical difficulties, and future development directions. This paper mainly introduces the multi−dimensional data format MDD (Multi Dimensional Dataset) proposed by the Aerospace Information Innovation Institute of the Chinese Academy of Sciences for the first time in the world, as well as the theory and technical system of multi−dimensional remote sensing data synthesis and representation, which fills the original gap in data organization in China and has a positive impact on international research. With the rapid development of remote sensing technology, multidimensional spatiotemporal spectral remote sensing data continues to emerge. This article aims to provide an overview of the domestic and international development of remote sensing multidimensional data formats, expound on the latest research achievements, analyze current technical difficulties, and look forward to future development directions, providing comprehensive references for researchers in related fields.

Due to technological limitations in early archaeological work, a large number of sites were documented only with vague textual descriptions, lacking precise geographic coordinates. This has posed significant challenges for subsequent archaeological investigations, site protection, and research. Traditional field survey methods are time−consuming and labor−intensive, making them unsuitable for large−scale or high−throughput site localization tasks. To address this issue, this study proposes an intelligent localization framework for archaeological sites based on large language models (LLMs). By designing tailored natural language prompts, the framework guides LLMs to automatically extract geographic descriptions—such as landmarks, directions, and distances—from online archaeological literature and records. It then integrates this information with high−resolution satellite imagery analysis to infer and calculate the precise geographic coordinates of the target site. This pipeline achieves full−process automation, covering text−based information extraction, remote sensing data retrieval, and spatial reasoning, thereby significantly improving the efficiency and intelligence level of site localization. Validation experiments conducted on the Laohushan Site, Sanxingdui Site, and Liao Zhongjing Site show that the deviation between the automatically inferred locations and actual surveyed coordinates is within one kilometer, with the best accuracy reaching approximately 10 meters—meeting the basic precision requirements of archaeological applications. Localization errors primarily stem from ambiguities in textual descriptions and uncertainties in image interpretation. Compared with traditional manual methods, this approach greatly reduces labor costs and offers enhanced scalability and application potential. The proposed method provides a novel technical pathway and tool support for the digitalization of archaeological information, site conservation, and further research.

The construction of well−facilitated farmland in China has imposed higher requirements on the digital and refined management of farmland, posing new challenges for the automatic extraction of multi−element information from remote sensing data. This study proposes a framework for the automatic extraction of multiple key farmland elements by integrating prompt engineering with remote sensing feature knowledge. Leveraging the general segmentation capabilities of the vision foundation model, combined with open−source data and feature−driven specialized algorithms, the proposed approach enables efficient automatic identification of critical farmland elements, including plots, field roads, shelterbelts, and irrigation and drainage facilities. Taking the well−facilitated farmland construction area in Shouguang City, Shandong Province as a case study, we conducted technical validation using high−resolution domestic satellite remote sensing imagery. Experimental results demonstrate that the proposed method can achieve batch automatic processing within farmland project areas, significantly reducing dependence on large amounts of high−quality training samples and manual workload. Moreover, it shows good generalization capabilities across images with different acquisition times, spatial resolutions, and regions, thus greatly improving data production efficiency and practical application potential. This research provides a novel approach for the intelligent interpretation of digitalized farmland areas and offers strong technical support for the application of remote sensing in post−construction supervision of agricultural engineering projects.

Land resource type is the basis for evaluating land suitability and development potential. The geomorphic pattern of "three mountains sandwiching two basins" and the significant landscape distribution characteristics of "mountain−oasis−desert" pose great challenges to the classification and utilization of land resources in Xinjiang. The paper aims to propose a grid−based fuzzy self−organizing feature maps (GF−SOFM) coupling classification method through the following steps: 1) A four−tier classification system was established based on dominant factors (climate+topography), stable factors (soil), relatively stable factors (vegetation), and dynamic factors (land use), comprising five factors with multiple indicators. 2) The study area was partitioned into 1 km×1 km grid units, where all indicators were spatially quantified. After fuzzy processing, the indicator data were input into SOFM model with dominant factors as control boundaries, enabling automated land resource type identification. 3) Area consistency test was conducted to compare the classification results of GF−SOFM method with those of the traditional thematic overlay method. The results indicate that Xinjiang was classified into 133 dominant factor types, 1906 stable factor types, 6054 relatively stable factor types, and 38493 dynamic factor types, achieving an average overall accuracy of 86%. The classification results of GF−SOFM method exhibited high spatial consistency with those of the traditional hierarchical overlay method, with 128 dominant factor types showing area consistency exceeding 92.35%. By refining topography classification and land use status, the GF−SOFM method effectively enables fine−scale land resource classification, accurately capturing the spatial patterns of climate−landform−soil−vegetation−land use. This approach serves as an ideal integrative classification method for physical geography, effectively indicating regional differentiation in ecological and geographical studies across varying climatic and landform regions. This research can provide a scientific basis for rational land development and utilization.

Rapid precise point positioning (PPP) over wide areas breaks through the dependency of traditional differential techniques on dense reference network, and serves as one of the key techniques to establish and maintain autonomous, wide-area, high-precision spatiotemporal framework using Global Navigation Satellite Systems (GNSS). This work provides a comprehensive review of the current development of satellite constellations, ground station infrastructure, and associated precise satellite products. Key technical advances are summarized in four respects, including multi-frequency and multi−GNSS integration, ambiguity resolution, atmospheric augmentation, and Low Earth Orbit (LEO) augmentation. All the analysis concentrate on their contributions to improving PPP positioning accuracy and convergence speed. Recent progress in PPP commercialization is discussed, alongside the latest developments in satellite-based PPP services, including BDS PPP-B2b. Finally, we outline persistent challenges and future research directions confronting wide-area rapid PPP, highlighting deep integration of high and low orbit constellations, multi-source fusion for positioning enhancement, and the evolution of next-generation satellite-based PPP service architectures. The study aims to support the large-scale deployment and service system development of high-precision positioning based on BDS/GNSS.

This article empirically studies the spatiotemporal differentiation and dynamic transfer characteristics of the open innovation level of 147 national high−tech zones from 2015 to 2020. Research has found that: the level of openness and innovation in world−class high−tech parks is higher than that in innovative technology parks, and innovative technology parks are better than innovative characteristic parks; the overall regional gap in the level of open innovation in the eight major national high−tech zones is widening, the gap in the level of open innovation in the southern coastal, northern coastal, and eastern coastal national high−tech zones is the largest, while the gap in the northwest national high−tech zone is the smallest; the regional gap in the level of open innovation in the eight major national high−tech zones has shown an expanding trend, and the growth rate of the regional gap has obvious regional heterogeneity, the regional gap has gradually become the main source of regional gap; without considering spatial correlation effects, its spatial transfer characteristics exhibit trends such as "asymmetric upward", "club convergence", and "stage differences".

In the early stages of Sino−American high−energy physics collaboration, gaps in cooperation demands, talent resources, and goal alignment posed significant challenges. Addressing these required effective communication channels. Due to his own characteristics, Tsung−Dao Lee has the control advantage of guiding the flow of elements of Sino−US scientific and technological cooperation. During the Beijing Electron−Positron Collider project, he acted as a trust, talent, and knowledge intermediary, facilitating alignment in policy−making, goal−setting and talent development between the two countries, thus effectively advancing Sino−American high−energy physics collaboration. This paper emphasizes the significance of intermediary roles in international scientific and technological cooperation under the situation of asymmetric and offers insights for future collaborations.