[Objectives] Classical statistical inference typically relies on the assumptions of large sample sizes and independent, identically distributed (i.i.d.) observations, conditions that spatio-temporal data frequently violate, leading to inherent theoretical limitations in conventional approaches. In contrast, Bayesian spatio-temporal statistical methods integrate prior knowledge and treat all model parameters as random variables, thereby forming a unified probabilistic inference framework. This enables the incorporation of a broader range of uncertainties and offers robustness in modelling small samples and dependent structures, making Bayesian methods highly advantageous and increasingly influential in spatio-temporal analysis. [Progress] From the perspective of methodological evolution, this paper systematically reviews mainstream Bayesian spatio-temporal statistical models from two complementary perspectives: traditional Bayesian statistics and the Bayesian machine learning. The former includes Bayesian Spatio-temporal Evolutionary Hierarchical Models, Bayesian Spatio-temporal Regression Hierarchical Models, Bayesian Spatial Panel Data Models, Bayesian Geographically Weighted Spatio-temporal Regression Models, Bayesian Spatio-temporal Varying Coefficient Models, and Bayesian Spatio-temporal Meshed Gaussian Process Model. The latter includes Bayesian Causal Forest Models, Bayesian Spatio-temporal Neural Networks, and Bayesian Graph Convolutional Neural Networks. In terms of application, the review highlights representative studies across domains such as public health, environmental sciences, socio-economic and public safety, as well as energy and engineering. [Prospect] Bayesian spatio-temporal statistical methods need to achieve breakthroughs in multi-source heterogeneous data modeling, integration with deep learning, incorporation of causal inference mechanisms, and optimization of high-performance computing. These advances are essential to balance theoretical rigor with practical adaptability and to promote the development of a next-generation spatio-temporal modeling paradigm characterized by causal inference, adaptive generalization, and intelligent analysis.

[Significance] Human mobility is closely tied to transportation, infectious disease spread, and public safety, making trajectory analysis and modeling a long-standing research focus. While numerous specialized trajectory models, such as interpolation, prediction, and classification models, have been developed using machine learning or deep learning, most are task-specific and trained on localized datasets, limiting their generalizability across tasks, regions, or trajectory data. Recent advances in generative AI have demonstrated the potential of foundation models in NLP and computer vision, motivating the need for a trajectory foundation model capable of learning universal patterns from large-scale mobility data to support diverse downstream applications. [Methods] This paper first reviews the research progress of various specialized trajectory models. It then categorizes trajectory modeling tasks into conventional tasks (e.g., trajectory similarity computation, interpolation, prediction, and classification) and generation task (i.e., trajectory generation), and elaborates on recent advances in trajectory foundation models for these two types of tasks. [Conclusions] The paper argues that trajectory foundation models for conventional tasks should enhance not only task generalization but also spatial and data generalization. Trajectory foundation models for generation task must address the challenge of spatial generalization, enabling the generation of large-scale trajectory data "from scratch" based on easily obtainable macro-level urban data or features. Furthermore, integrating trajectory data with other data types (e.g., text, maps, and other geospatial data) to construct multimodal geographic foundation models, as well as developing application-oriented trajectory foundation models for fields such as transportation, public health, and public safety, are promising research directions worthy of future exploration.

[Significance] As a fundamental unit for representing complex processes in the real world, events have garnered increasing attention across disciplines in recent years. With the advancement of major initiatives such as Digital Twin and China's 3D realistic geospatial scene, there is an urgent need to reexamine event definitions, classification schemes, and data modeling systems from an entity-based perspective to support multidimensional information representation and deep semantic interpretation. [Progress] This paper first provides a systematic review of the conceptual connotations and definitional characteristics of events from a cross-disciplinary perspective. It then summarizes the theoretical foundations and structural features of existing event classification approaches. Furthermore, based on modeling principles and application scopes, a tripartite framework for event data models is proposed, comprising: (1) semantic element-based models, (2) ontology-based event models, and (3) spatiotemporal entity-oriented event models. These models are systematically analyzed and comparatively evaluated in terms of their expressiveness across key dimensions such as time, space, participating objects, event relationships, and hierarchical structures, revealing their respective advantages and limitations in supporting dynamic process modeling, generality, and extensibility. [Methods] Building on the above, a spatiotemporal entity-based definition of events is proposed, followed by two novel classification approaches for entity-oriented event modeling and analysis: one based on the completeness of event records, and the other on the types of involved change processes. [Prospect] Future research should advance event data models from structural organization toward semantic integration and intelligent construction. This entails integrating the strengths of diverse models to build a unified framework; fostering synergy between qualitative and quantitative modeling mechanisms for integrated analysis; and incorporating artificial intelligence to enhance model intelligence and adaptability.

[Objectives] Visual balance is one of the most important factors affecting the efficiency of map information communication and visual aesthetics. Although existing balance calculation methods are effective for traditional maps, their applicability is significantly reduced when applied to WeMaps with personalized element configurations. The dynamic nature of elements on WeMaps leads to complex distribution patterns and challenging quantification, posing a major obstacle to visual balance calculation for WeMaps. [Methods] To address this issue, a multifactor-driven method for calculating the visual balance of WeMaps is proposed. First, adaptive co-occurrence filtering and affiliation matrix partitioning are applied to each WeMap to optimize image distribution features and determine the degree of attribution for each element. This enables the extraction of map layout elements based on robust fuzzy C-means clustering. Next, three visual perception factors, brightness, contrast, and saliency, are introduced to quantify the visual weight, detail, and focus of graphical elements, respectively. Finally, using the Euclidean distance and angle between the visual center of gravity and the visual center of the graphical elements, a model for calculating the visual balance of WeMaps is constructed. An experimental dataset of 78 WeMaps was built using data from project designs, news sources, and official tourism websites. The effectiveness of the proposed method was validated through six parameter adjustment experiments, evaluation surveys, comparative experiments, and visual feature variation tests. [Results] The experimental results show that when the distance weight α is set to 0.8 and the angled weight β to 0.2, the proposed method aligns best with human aesthetic perception. [Conclusions] The proposed method is highly consistent with expert cartographic experience and effectively simulates human visual cognition, demonstrating strong practical applicability.

[Objectives] Traditional settlements are rich in local cultural values and represent the culmination of long-standing agricultural civilizations and deep-seated wisdom. Revealing their traditional cultural connotations through the lens of information geography holds significant value. However, there is currently a lack of theoretical research on the informational attributes of the Cultural Landscape Genes of Traditional Settlements (CLGTS) from a social-cultural semiotic perspective. [Methods] To address this gap, this paper introduces the concept of Symbolic Information Entropy of CLGTS (SIE-CLGTS), grounded in information entropy theory. Based on the varying modes of expression of CLGTS symbols, this study defines corresponding entropy calculation methods, including gray-scale distance, spectral analysis, Bayesian probability statistics, adjacency relationships, and structural elements. Using the landscape gene symbols of Zhongtian Village in Hunan Province as a case study, we conducted experiments to measure SIE-CLGTS and analyze the results. Additionally, a simulated tourist guide route was designed to explore potential applications of SIE-GLGTS. [Results] The results indicate that: (1) The information entropy-based method can effectively quantify CLGTS; (2) SIE-CLGTS reflects the cultural attributes embodied in these symbols; (3) SIE-CLGTS shows great potential for applications such as the preservation and sustainable development of traditional villages. [Conclusions] This paper represents a scientific exploration of the informational characteristics of landscape genes from an information science perspective, contributing to the deeper application of landscape gene theory.

[Objectives] Buildings are fundamental elements on maps. In the fields of similarity relations and geospatial cognition, detecting building shapes presents both a significant challenge and a key research focus. Moreover, it has the potential to support cartographic generalization technologically. However, existing supervised learning methods for grid-based building shape recognition require a large number of labeled samples. To address this limitation, this paper proposes a Shape Recognition Method for Rasterized Buildings Based on a Contrastive Transfer Model. This approach combines transfer learning techniques with self-supervised feature extraction strategies. The goal is to extract shape features during a self-supervised learning phase, followed by supervised learning for shape classification, thereby minimizing labeling effort and reducing training costs. [Methods] In this study, the Contrastive Learning-based Transfer Model (CLTM) is applied for shape recognition focused on raster data. The process is as follows: First, the shapes of rasterized buildings are extracted through pre-processing. These shapes are binarized and standardized in size to eliminate the effects of pixel noise and size differences. Second, a high-dimensional feature vector is generated by encoding the building shapes using a contrastive learning model. This step extracts the unique characteristics of each shape and optimizes the model via a contrastive loss function. Finally, the model parameters are updated, and transfer learning is used to evaluate model performance through a downstream shape prediction task. [Results] Experimental results indicate that the classification accuracy of the proposed method reaches 93.79%, which is slightly higher than the AlexNet method but lower than ResNet50. In the shape recognition application, t-SNE visualization is used to clearly display the clustering trends of different shape categories in two-dimensional space. The clustering results indicate that shapes of the same category are closely grouped, while dissimilar shapes are well separated. This confirms the model's effectiveness, as it can accurately distinguish between similar and dissimilar shapes. [Conclusions] The proposed method performs well in building shape classification by leveraging data augmentation and a contrastive loss function to train the model and extract useful features from unlabeled data. This significantly reduces the need for manual data annotation and mitigates the influence of human visual bias. Compared to fully supervised methods, it offers distinct advantages, exhibiting strong shape recognition capabilities and providing an efficient and reliable approach for analyzing geospatial elements.

[Objectives] Small object detection is of great significance in both military and civil applications. However, due to challenges such as low resolution, high noise environments, target occlusion, and complex backgrounds, traditional detection methods often struggle to achieve the necessary accuracy and robustness. The problem of detecting small objects in complex scenes remains highly challenging. Therefore, this paper proposes a hybrid feature and multi-scale fusion algorithm for small object detection. [Methods] First, a Hybrid Conv and Transformer Block (HCTB) is designed to fully utilize local and global context information, enhancing the network's perception of small objects while optimizing computational efficiency and feature extraction capability. Second, a Multi-Dilated Shared Kernel Conv (MDSKC) module is introduced to extend the receptive field of the backbone network using dilated convolutions with varying expansion rates, thereby enabling efficient multi-scale feature extraction. Finally, the Omni-Kernel Cross Stage Model (OKCSM), constructed based on the concepts of Omni-Kernel and Cross Stage Partial, is integrated to optimize the small target feature pyramid network. This approach helps preserve small object information and significantly improves detection performance. [Results] Ablation and comparison experiments were conducted on the VisDrone2019 and TinyPerson datasets. Compared to the baseline model YOLOv8n, the proposed method improves precision, recall, mAP@₅₀, and mAP@_50:95 by 1.3%, 3.1%, 3%, and 1.9%, respectively on VisDrone2019, and by 3.6%, 1.3%, 2.1%, and 0.7%, respectively on TinyPerson. Additionally, the model size and GFLOPs are only 6.3 MB and 11.3 G, demonstrating its efficiency. Furthermore, compared with classical algorithms, such as HIC-YOLOv5, TPH- YOLOv5, and Drone-YOLO, the proposed algorithm demonstrates significant advantages and superior performance. [Conclusions] The algorithm effectively improves detection accuracy, confirming its strong performance in addressing small object detection in complex scenes.

[Objectives] Cross-view image geolocation refers to a technology that determines the geographical location of an image by matching it with reference images taken from different perspectives and possessing precise location information. This technology plays a crucial role in real-world applications such as Unmanned Aerial Vehicle (UAV) navigation, environmental monitoring, and target positioning. Currently, most deep learning-based cross-view image retrieval and geolocation methods for drone-satellite tasks rely heavily on supervised learning. However, the scarcity of high-quality labeled data presents a significant limitation, hindering the generalization capability of these models. Moreover, existing methods often fail to effectively model the spatial layout of images, making it difficult to bridge the substantial domain gap between cross-view images, thereby limiting the accuracy and robustness of geolocation tasks. [Methods] To address these challenges, this paper proposes a novel cross-view image retrieval and localization architecture called DINO-MSRA. The architecture first employs the DINOv2 large model framework, fine-tuned by Conv-LoRA, as the feature encoder. This enhances the model's feature extraction capabilities with fewer parameters, improving both efficiency and accuracy. Second, we design a spatial relation-aware feature aggregator based on the Mamba module (MSRA) to more effectively aggregate image features. By embedding spatial configuration features into the global descriptor, this module significantly improves the model's performance in cross-view matching tasks, especially in complex scenarios where spatial relationships between objects are crucial. Finally, the InfoNCE loss function is adopted to train the model, optimizing contrastive learning and ensuring more accurate retrieval and localization results. [Results] Extensive comparative and ablation experiments were conducted on the University-1652 and SUES-200 datasets. The experimental results show that for drone-view target localization (drone→satellite) and drone navigation (satellite→drone) tasks, the proposed method achieves R@1 accuracies of 95.14% and 97.29%, respectively, on the University-1652 dataset, representing improvements of 0.68% and 1.14% over the current best algorithm, CAMP. On the SUES-200 dataset at an altitude of 150 meters, R@1 accuracies reach 97.2% and 98.75%, which are 1.8% and 2.5% higher than CAMP, respectively. Moreover, the proposed method requires significantly fewer parameters than existing algorithms, only 19.2% of those used by Sample4Geo. [Conclusions] In summary, the proposed DINO-MSRA architecture outperforms current state-of-the-art methods in cross-view image matching, achieving higher accuracy and faster inference speed. These results demonstrate its robustness and practical application potential in challenging real-world scenarios.

[Objectives] High-resolution remote sensing image segmentation provides essential data support for urban planning, land use, and land cover analysis by accurately extracting terrain information. However, traditional methods face challenges in predicting object categories at the pixel level due to the high computational cost of processing high-resolution images. Current segmentation approaches often divide remote sensing images into a series of standard blocks and perform multi-scale local segmentation, which captures semantic information at different granularities. However, these methods exhibit weak feature interaction between blocks, as they do not consider contextual prior knowledge, ultimately reducing local segmentation performance. [Methods] To address this issue, this paper proposes a high-resolution remote sensing image segmentation framework named CATrans (Cross-scale Attention Transformer), which combines cross-scale attention with a semantic-based visual Transformer. CATrans first predicts the segmentation results of local blocks and then merges them to produce the final global image segmentation. It introduces contextual prior knowledge to enhance local feature representation. Specifically, we propose a cross-scale attention mechanism to integrate contextual semantic information with multi-level features. The multi-branch parallel structure of the cross-scale attention module enhances focus on objects of varying granularities by analyzing shallow-deep and local-global dependencies. This mechanism aggregates cross-spatial information across various dimensions and weights multi-scale kernels to strengthen multi-level feature representations, enabling the model to avoid deep stacking and multiple sequential processes. Additionally, a semantic-based visual Transformer is adopted to couple multi-level contextual semantic information. Spatial attention is used to reinforce these semantic representations. The multi-level contextual information is grouped to form abstract semantic concepts, which are then fed into the Transformer for sequence modeling. The self-attention mechanism within the Transformer captures dependencies between different positions in the input sequence, thereby enhancing the correlation between contextual semantics and spatial positions. Finally, enhanced contextual semantics are generated through feature mapping. [Results] This paper conducts comparative experiments on the DeepGlobe, Inria Aerial, and LoveDA datasets. The results show that CATrans outperforms existing segmentation methods, including Discrete Wavelet Smooth Network (WSDNet) and Integrating Shallow and Deep Network (ISDNet). CATrans achieves a Mean Intersection over Union (mIoU) of 76.2%, 79.2%, and 54.2%, and a Mean F1 Score (mF1) of 86.5, 87.8%, and 66.8%, with inference speeds of 38.1 FPS, 13.2 FPS, and 95.22 FPS on the respective datasets. Compared to the best-performing method, WSDNet, CATrans improves segmentation performance across all classes, with mIoU gains of 2.1%, 4.0%, and 5.3%, and mF1 gains of 1.3%, 1.8%, and 5.6%. [Conclusions] These findings highlight that the proposed CATrans framework significantly enhances high-resolution remote sensing image segmentation by incorporating contextual prior knowledge to improve local feature representation. It achieves an effective balance between segmentation performance and computational efficiency.

[Objectives] Accurate extraction of global surface water bodies is crucial for water resource management and climate monitoring. Synthetic Aperture Radar (SAR), with its active microwave imaging capabilities, enables all-weather, all-day monitoring of water body changes in areas with persistent cloud cover and heavy rainfall. However, the dense river networks and complex terrain of the Qinghai-Tibet Plateau, along with the imaging characteristics of SAR, pose significant challenges that often hinder the accurate differentiation of mountain shadows, bare land, and water bodies. To address these challenges, this paper proposes a Multi-Scale Feature Fusion Water Body Extraction model, MSFSwin (Multi-Scale Feature Fusion Swin), which leverages the multi-channel information of SAR. [Methods] By integrating dual-polarization features from ascending and descending Sentinel-1 images with Digital Elevation Model (DEM) data, we construct a multi-channel remote sensing water dataset. To enhance the perception of multi-scale water bodies and compensate for the Swin Transformer's limitations in cross-window feature integration, we introduce an Atrous Spatial Pyramid Pooling (ASPP) module that aggregates features with different receptive fields. Additionally, a multi-level decoding structure is designed to strengthen cross-scale information interaction through hierarchical feature fusion, enabling refined water extraction. To validate the effectiveness and robustness of the MSFSwin model, experiments were conducted in typical water-covered regions of the Qinghai-Tibet Plateau. We performed both qualitative and quantitative comparisons against several deep learning models (e.g., Swin Transformer, Segformer, ViT) and the KNN algorithm. [Results] The experimental results show that the MSFSwin model outperforms Swin Transformer, Segformer, and ViT in water body extraction. In the river-lake confluence area, the BF-score of the proposed method improved by 4.15% compared to the baseline model, while the IoU in areas with small water bodies increased by 3.52%, enabling high-precision automatic water extraction in complex terrains. [Conclusions] By leveraging the MSFSwin model and a multi-channel fusion strategy, this study achieves clear distinction between water bodies and mountain shadows, substantially improving the robustness and adaptability of water body extraction in complex terrains. The proposed approach offers a reliable solution for high-precision, all-weather, all-day water body monitoring. Code Link: https://github.com/infinitas732/code.

[Objectives] Rainfall monitoring is essential for preventing meteorological disasters, protecting the ecological environment, and managing water resources scientifically. Current studies on rainfall estimation using machine learning algorithms often involve imbalanced datasets and employ random oversampling or undersampling techniques. These approaches may introduce highly correlated features, thereby reducing model accuracy and generalization capability. To address these issues, this study proposes a novel rainfall monitoring model that integrates machine learning techniques with satellite remote sensing data. [Methods] This method uses field-measured rainfall data and brightness temperature data from the Microwave Humidity Sounder (MWHS) of the FY-3D satellite's Microwave Temperature Sounder/ Humidity Sounder (TSHS) product, collected from 14 major rainfall events in Fujian Province from 2020 to 2022. This paper proposes a SMOTE-PCA-RF monitoring model for hourly rainfall to classify rainfall regions and intensity levels in Fujian Province. The model combines the Synthetic Minority Over-sampling Technique (SMOTE), Principal Component Analysis (PCA), and the Random Forest (RF) classifier. The performance of SMOTE-PCA-RF is compared with RF, PCA-RF, and SMOTE-RF models to determine the optimal approach. [Results] The results show that the SMOTE-PCA-RF model achieves the highest testing Threat Score (TS) for rainfall distribution estimation and the highest F1 score for rainfall grade classification, both reaching 0.60. Compared with other models, the TS increases by 3.45%~9.09%, and the F1 score increases by 9.09%~33.33%. Additionally, the study finds that while SMOTE improves classification performance, it may also increase overfitting and the False Alarm Rate (FAR). PCA, through dimensionality reduction, not only improves the model's generalization ability but also improves training efficiency by 9.75%~31.70%. A case study using the SMOTE-PCA-RF model shows that, although estimation accuracy decreases with increasing rainfall, the F1 score remains at 0.50, and the estimated results closely align with the spatial distribution of observed rainfall. [Conclusions] The study's findings provide technical support for rainfall monitoring and help relevant departments quickly and intuitively understand the spatial distribution and intensity of rainfall over large areas. This, in turn, enhances the ability to prevent and mitigate meteorological disasters.

[Significance] As the basic unit of a city, the carbon emission levels and accuracy of community-scale accounting directly impact the overall effectiveness of emission reduction in the construction industry. This paper reviews the main methods of carbon accounting, evaluates their advantages and disadvantages, and proposes a new approach to enhance the accuracy and comprehensiveness of community carbon accounting using digital twin technology. [Progress] This paper first introduces three traditional carbon accounting methods, namely the carbon emission factor method, the mass balance method, and the direct measurement method, and discusses their applications. It then identifies digital twin technologies suitable for community-scale carbon accounting, including Building Information Modeling (BIM), Geographic Information System (GIS), and the Internet of Things (IoT). The paper analyzes current development trends, including: (i) expanding the scope of carbon accounting to the community level using digital twin technology, (ii) strengthening the integration and interoperability of digital twin systems, and (iii) establishing a community carbon accounting framework grounded in digital twin technology. It further proposes integrating BIM, GIS, and IoT into a unified system based on the city information model to build a comprehensive community carbon emission platform. [Prospect] Looking ahead, the application of digital twin technology holds promise for enabling accurate carbon accounting, emission forecasting, reduction pathway planning, and performance evaluation for communities of varying scales and geographical contexts. Furthermore, with advances in AI technology, it is anticipated that city information models for community carbon accounting will increasingly integrate AI agents, leveraging the power of big data, large models, and high-performance computing, to create intelligent carbon accounting systems for the smart city era.

[Objectives] China's transport sector is one of the fastest-growing sources of carbon emissions, with Road Traffic Carbon Emissions (RTCE) accounting for a large share. The way an urban road network is laid out may strongly influence RTCE, yet existing studies often ignore spatial non-stationarity and nonlinear effects. [Methods] This article takes 302 urban functional areas in China as the research object. Experiment data include 2019 urban road network data, road traffic carbon emission grid data, and population and GDP grid data. Firstly, ArcGIS and osmnx packages were used to visualize the road traffic carbon emissions, road grade distribution, traffic network density, and traffic network structure indicators of the 302 urban functional areas. The distribution characteristics of urban RTCE and urban road network were also analyzed. Then, the fitting effects of OLS, GWR, and MGWR models were compared and analyzed to identify the best model for relating road-network form to RTCE. Finally, based on the Multi-scale Geographically Weighted Regression model (MGWR) and SHAP analysis, the impact mechanism of road network morphology on RTCE was explored. [Results] ① The spatial distribution of Road Traffic Carbon Emissions (RTCE) exhibits a multi-center pattern, with core areas such as the Beijing-Tianjin-Hebei region (1 003.604 t/km²), the Yangtze River Delta (849.074 t/km²), the Pearl River Delta (1 615.291 t/km²), and provincial capital cities (1 168.886 t/km²), gradually decreasing toward the surrounding areas. The RTCE levels in the eastern region are generally higher than those in the central and western regions. In terms of the spatial distribution characteristics of road network morphology, the density of the traffic network and road hierarchy distribution resemble the RTCE distribution pattern. The southern regions exhibit higher Road Direction Richness (RDR), while the northern regions have higher road Grid Coefficients (GC). ② The impact of road network morphology on road traffic carbon emissions shows significant spatial heterogeneity. For example, Road Network Density (RND) has a more pronounced impact in the Pearl River Delta (0.636), while Road Direction Richness (RDR) has a greater influence in the Yangtze River Delta (0.259). Additionally, different road network morphological indicators vary considerably in their impact on RTCE across regions. ③ Road network morphology exhibits spatial non-stationarity and nonlinear effects on RTCE. For instance, the bandwidth of RND is only 45, whereas that of RCR is 215, indicating that different morphological characteristics affect RTCE at different spatial scales. In the SHAP analysis based on machine learning, which accounts for nonlinear impacts, RND is identified as the most important feature influencing road traffic carbon emissions. [Conclusions] This study employs the MGWR model and SHAP method to reveal the spatial non-stationarity and nonlinear influence mechanisms of road network morphology on road traffic carbon emissions. The results indicate that the impact of road network characteristics on traffic carbon emissions varies significantly across different regions. These differences are reflected not only in spatial distribution but also in the underlying mechanisms of influence. Therefore, when formulating low-carbon road network planning strategies, it is essential to fully consider the spatial heterogeneity, non-stationarity, and nonlinear characteristics of the road network. A comprehensive analysis from the multidimensional perspective of "density-hierarchy-structure" is recommended to promote low-carbon urban transportation. These findings provide a scientific basis for urban transportation planning and low-carbon development, contributing to sustainable urban development, improved traffic efficiency, and enhanced quality of life for residents.

[Objectives] The spatial conflict between three-dimensional (3D) urban growth and cultural heritage preservation poses a significant challenge to sustainable development. Simulating 3D urban expansion scenarios under heritage protection constraints can provide scientific evidence and decision-making support for optimizing spatial layouts, enhancing site management, and achieving long-term sustainability goals. However, few studies have incorporated heritage preservation factors into the modeling frameworks of three-dimensional urban growth scenarios, leaving a critical research gap in the simulation of 3D expansion in historical cities. [Methods] This study proposes an integrated 3D Zoning GA-CA-Markov model, specifically designed to account for the constraints imposed by historical and cultural heritage sites during urban expansion. The model introduces a spatially explicit zoning strategy that incorporates protection factors related to heritage areas. Using the Xi’an metropolitan area—rich in cultural relics and historical landmarks, as a case study, the methodology includes three main steps. First, three horizontal urban expansion scenarios under heritage protection constraints are simulated using the Zoning GA-CA-Markov model. Second, the newly developed land patches are classified into urban functional types using a random forest classifier. Third, the building heights across different functional zones are predicted using a LightGBM regression model, which generates a 3D urban form that reflects both spatial growth and vertical development dynamics under heritage limitations. [Results] The results indicate that: (1) The proposed 3D Zoning GA-CA-Markov model effectively reproduces the spatial patterns of urban expansion under cultural heritage constraints. The horizontal urban growth simulation achieves an overall accuracy of 89.65%, a Kappa coefficient of 0.758 2, and a Figure of Merit (FoM) of 0.274 0. The Root Mean Square Error (RMSE) of the building height prediction ranges from 1.7 to 2.8 meters. (2) Reasonable height limits and spatial planning can optimize urban growth patterns, achieving a balance between development and heritage conservation. Under the heritage protection scenario, newly developed land exhibits a predominantly low- to mid-rise 3D growth structure. Urban growth follows leapfrog and edge-expansion patterns, dominated by residential and public service functions. Height restrictions result in low-density, controlled development, with most new buildings limited to heights between 12 and 16 meters, and some areas below 12 meters. (3) In areas adjacent to heritage sites, building heights of residential and public service zones are significantly limited to maintain visual harmony and spatial coordination. In contrast, industrial land is less influenced by heritage constraints, with building heights primarily driven by economic and transportation factors. [Conclusions] The 3D Zoning GA-CA-Markov model can be broadly applied to simulate and predict 3D spatial expansion in historically and culturally significant cities. It offers practical support for the coordinated regulation of urban growth and heritage preservation strategies.

[Objectives] Under the background of National Cultural Park construction, tourism and recreational activities related to lineal cultural heritage sites, such as the Great Wall and the Grand Canal, are rapidly expanding along their routes. The landscape planning and design of these parks must address the question: "where can the landscape be seen?", a query rooted in the evaluation of resource value. The challenge involves two core issues in large-scale visual perception research of lineal cultural heritage: insufficient multi-dimensional semantic representation of landscapes, and computationally intensive visible location analysis. [Methods] Grounded in Gestalt theory, this study considers the multi-dimensional semantics embedded in the landscapes of large-scale lineal cultural heritage, including hierarchy systems, spatial structure, historical function, and morphological aesthetics. A method is proposed for the automatic extraction of landscape semantic feature points and the computation of visible locations. First, adhering to the principle of emphasizing the significance and visual value of heritage landscapes, these landscapes are abstracted and summarized into a set of feature points containing semantic information. Second, visible locations are computed based on these points, with the NetCDF multi-dimensional data format employed to integrate, organize, and store both the visibility computation results and semantic feature points. By using these feature points as a bridge, an integrated representation of multi-dimensional landscape semantics and their corresponding visible locations is achieved. Furthermore, through visible location mining, the process of "summarization—representation—restoration" of the landscape semantics is realized. [Results] The empirical study focuses on the Ming Great Wall in the Beijing-Tianjin-Hebei region. Based on datasets such as DEM and cultural relics surveys, a total of 53,454 feature points were extracted, followed by visibility computation and data organization. Field verification shows that the average coincidence rates for the number and content of feature points in visible locations are 76.37% and 70.69%, respectively. These results demonstrate that the proposed method can efficiently extract feature points that represent the semantics of large-scale heritage landscapes, and that the computed visible locations exhibit high reliability. Visible location mining enables the identification of landscape semantics at specific viewpoints, as well as the discovery of high-quality viewpoints for perceiving particular semantics, thus enabling two-way querying between semantic content and visible locations. [Conclusions] The proposed method for multi-dimensional semantic feature point extraction and landscape visible location computation offers new approaches and perspectives for the visual analysis of large-scale cultural heritage and the exploration of landscape visual value.

[Objectives] Points of Interest (POIs) have dual characteristics as geospatial entities and carriers of cultural information, serving as the data foundation for analyzing and identifying regional cultural expressions and functional traits. Identifying and analyzing the types and characteristics of tourism cultural scenes along the Grand Canal is of great significance for achieving differentiated and sustainable cultural tourism development. [Methods] By integrating POI data with scene theory, spatial entities are associated with cultural values, and quantitative statistics are combined with qualitative configuration analysis. A tourism-cultural amenity database was established using 476,968 POI records, categorized into 6 major categories and 24 sub-categories. The Delphi method was employed to determine scores for each subcategory related to tourism amenity scenes, which were then used to calculate the performance scores of tourism cultural scenes. Descriptive statistical analysis, K-means clustering, and hierarchical clustering were applied to identify types of tourism-cultural scenes. The clustering results were visualized on maps. Meanwhile, the characteristics, formation mechanisms, and corresponding countermeasures of these scene types were further analyzed. [Results] (1) The Jiangsu section of the Grand Canal exhibits distinctive local tourism-cultural characteristics, with strong regional identity and attractiveness. However, significant disparities exist in tourism-cultural value orientations, particularly in subcategories such as locality, glamour, exhibitionism, utilitarianism, and charisma, highlighting the heterogeneous features of tourism-cultural scenes in this area. (2) Cluster analysis classified 34 counties (cities or districts) along the Jiangsu section into four types: local scenes (10 regions), utilitarian scenes (8 regions), comfortable scenes (13 regions), and charming scenes (3 regions). Discriminant analysis validated the reliability of these clustering results. Each of the four scene types exhibits distinct characteristics. (3) The types of tourism-cultural scenes are influenced by the combined effects of multiple factors (economic development, urbanization, population, fiscal policy, transportation, and tourism resources), which can be summarized into three configuration-based influence paths. [Conclusions] This study introduces scene theory into cultural tourism research based on POI big data, offering a novel approach to promoting regionally differentiated and sustainable development of cultural tourism.