[Significance] The characteristics of the lunar surface, including its mineral compositions, geological formations, environmental factors, and temperature variations, are essential for advancing our understanding of the Moon. These features provide a wealth of scientific data for lunar research, such as resource distribution, environmental characteristics, and evolutionary history. Spectral imagers, which detect mineral compositions in a nondestructive way, play a crucial role in analyzing the mineral compositions of the lunar surface and have become key payloads in scientific exploration missions. With the increasing demand for high-precision lunar exploration data and advancements in spectral imaging technology, there is a growing trend toward acquiring lunar remote sensing data with higher spatial and spectral resolution across a broad spectral range. This trend is shaping the future of lunar orbit exploration, allowing for unprecedented detail in probing the Moon's surface. However, the higher resolution of spatial and spectral data also introduces significant challenges in data processing. [Progress] This paper begins by summarizing existing lunar spectral orbit data, including payload parameters and associated scientific findings. It then explores specific technical challenges in the data processing chain, such as pre-processing and the calculation of lunar surface parameters. Mapping surface compositions through spectral remote sensing is particularly complex due to the mixing of minerals within rocks, which can obscure clear spectral signatures. To address these challenges, various theoretical and empirical approaches have been developed. This paper proposes technical methods and potential solutions to overcome these obstacles.[Conclusions] In conclusion, detailed studies of lunar surface characteristics and the acquisition of high-resolution spectral data are vital for advancing lunar science. Lunar hyperspectral data are expected to support manned lunar exploration and scientific research by enabling the identification of various minerals on the Moon's surface and determining their abundance through hyperspectral observations. Advances in spectral imaging technology and the development of solutions for processing high-resolution data will significantly enhance lunar and planetary science capabilities. These efforts will pave the way for deeper insights into the Moon's geology and potential resource utilization.

[Significance] Lunar remote sensing is a critical method to ensure the safety and success of lunar exploration missions while advancing lunar scientific research. It plays a significant role in understanding the Moon's geological evolution and the formation of the Earth-Moon system. Accurate lunar topographic maps are essential for mission planning, including landing site selection, navigation, and resource identification. These maps also provide valuable data for studying planetary processes and the history of the solar system. [Progress] In recent years, with growing global interest and investment in lunar exploration, remarkable progress has been made in remote sensing technology. These advancements have significantly improved the precision, resolution, and coverage of lunar topographic mapping. Various lunar remote sensing missions, such as China's Chang'e program, NASA's Lunar Reconnaissance Orbiter, and missions by other space agencies, have acquired substantial amounts of multi-source, multi-modal, and multi-scale data. This wealth of data has laid a solid foundation for technological breakthroughs. For instance, high-resolution laser altimetry, optical photogrammetry, and synthetic aperture radar have provided detailed datasets, enabling refined mapping of the Moon's surface. However, the dramatic increase in data volume, complexity, and heterogeneity presents challenges for effective processing, integration, and application in topographic mapping. This paper provides a comprehensive overview of the current state of lunar topographic remote sensing and mapping, focusing on the implementation and data acquisition capabilities of major lunar remote sensing missions during the second wave of lunar exploration. It systematically summarizes the latest research progress in key surveying and mapping technologies, including laser altimetry, which enables precise elevation measurements; optical photogrammetry, which reconstructs surface features using high-resolution imagery; and synthetic aperture radar, which provides unique insights into topographic and subsurface structures. [Prospect] In addition to reviewing recent advancements, the paper discusses future trends and challenges in the field. Key recommendations include enhancing sensor functionality and performance metrics to improve data quality, optimizing the lunar absolute reference framework for consistency and accuracy, leveraging multi-source data fusion for fine-scale modeling, expanding scientific applications of lunar topography, and developing intelligent and efficient methods to process massive amounts of remote sensing data. These efforts will not only support upcoming lunar exploration missions, such as China's manned lunar landing program scheduled for 2030, but also contribute to a deeper understanding of the Moon and its relationship with Earth.

[Objectives] The identification and classification of lunar impact craters are critical for selecting spacecraft landing sites and estimating the Moon's geological age. However, the complex morphological features created by impact processes post significant challenges to studying micro-scale lunar surface features, which are often indivisible at the pixel level. Addressing these challenges requires a scale-adaptive approach that incorporates micro-scale characteristics to refine lunar impact crater classification maps. [Methods] This study introduces a scale-adaptive algorithm based on geomorphons for the automatic classification of micro-scale lunar surface features. First, terrain parameters are optimized to define local ternary patterns of lunar geomorphology. These patterns are then used to determine lunar geomorphons. Next, the geomorphons are aggregated according to rules based on relief amplitude and slope to identify lunar impact geomorphic units on a larger scale. Finally, a classification map of lunar impact craters in the Gagarin Crater region is constructed using the identified geomorphons. [Results] The proposed method successfully identifies the optimal parameters for adaptively scaling lunar geomorphons by incorporating the unique characteristics of lunar surface features. Using a four-parameter constraint window, lunar geomorphons are refined at locally optimal spatial scales through the computation of local ternary patterns integrated with the theory of lunar geomorphological evolution. The results reveal that the generated maps of lunar geomorphons exhibit significant spatial aggregation, well-defined classification boundaries, and high accuracy in representing lunar impact craters. The method effectively captures the internal structural details of impact craters, providing a pixel-level depiction of their morphological features. The multi-scale identification of impact craters achieves a precision of 88.24%, a recall of 84.96%, and an F1 score of 86.57%. A classification schema for impact craters was established, including simple pit, small-scale bowl, small-scale flat bottom, small-scale central peak, medium flat bottom, medium central peak, large ring plain, and giant complex. [Conclusions] This method demonstrates robustness and high efficiency in crater identification, offering multi-scale geomorphological units and serving as a foundational tool for scale-based lunar scientific research. It provides technical support for identifying and classifying multi-scale lunar impact craters, contributing to advancements in lunar morphological and geological analysis.

[Objectives] Korolev basin is located within the Feldspathic Highlands Terrane and in the highest topographic area of the moon. It serves as the only farside landing site for the Lunar Geophysical Network mission to conduct geophysical measurements. The study of microwave thermal radiation characteristics of surface materials in Korolev basin can provide new and significant scientific references for the mission. [Methods] The microwave radiometer on Chang'e-1/2 satellites have made the first on-orbit passive microwave brightness temperature measurements of lunar regolith. The microwave radiometer data are sensitive to the temperature and composition of regolith, enabling the characterization of shallow regolith's thermal properties. Therefore, based on the data from the microwave radiometer on Chang'e-2 satellite, this study employed the centroid interpolation method based on Delaunay tetrahedralization to generate brightness temperature (TB) maps and then generate brightness temperature difference (dTB) maps. These maps were then combined with (FeO+TiO₂) abundance map generated from Clementine UV/VIS data, rock abundance map generated from Diviner data, thorium distribution map, and Bouguer gravity map to evaluate the microwave thermal radiation properties of the surface regolith in Korolev basin. [Results] (1) The four-quadrant analysis and profile analysis methods are employed, revealing that the TB behaviors of the central and northwestern part of Korolev basin are similar to those observed in Compton-Belkovich (C-B) region, both exhibiting microwave warm anomalies; (2) Along the inner rim of Korolev basin, there is a distribution of regions with low TB both at day and night; (3) In the western part of the basin, there is a region where the Diviner data don’t detect the presence of rocks, while the TB maps here show characteristics affected by rocks. Considering the different sensitivities of microwave and thermal infrared data to rock size, rocks not identified by thermal infrared data are distributed in this region. [Conclusions] (1) Radioactive elements are likely to be enriched at the bottom of Korolev basin, while the surface Th abundance cannot represent the Th abundance in the deep; (2) The impact events have exposed the purest anorthosite from the subsurface, leading to the occurrence of low TB anomalies. These findings provide new and important scientific references for further investigating the thermal state of the lunar highlands and the thermal activity of the shallow crust..

[Objectives] Typical elements on the surfaces of the Earth and other celestial bodies are essentially spatial entities with time-varying properties. These spatial entities or elements can be represented in either raster or vector data formats. The detection and extrapolation of their spatial patterns rely on a common theoretical foundation of spatial analysis. The spatial patterns and potential evolution of elements of the Earth and in deep-space scenarios are closely related to time and environmental factors. Their evolution, transitions, and correlations are a blend of necessity and contingency. This implies that detecting and extrapolating spatial patterns can be abstracted as a problem of occurrence probability for a phenomenon or scenario, consisting of transition probabilities and suitability probabilities governed by complex multifactorial mechanisms. [Methods] Therefore, we categorized the extensive factors influencing the spatial patterns and evolution of typical elements on Earth and in deep-space objects into five types: Topography (T), Constraints (C), Accessibility (A), Proximity (P) and Heterogeneity (H). We further proposed a universal theoretical paradigm, named TCAPH, for extrapolating occurrence-of-probability by integrating these five common types of factors. Based on the TCAPH theoretical framework, we developed a method for projecting elements' spatio-temporal evolution considering transition probability (TCAPH-Trans) and a method for spatial suitable site selection considering suitability probability (TCAPH-Suit). [Results] To validate the constructed theoretical paradigm and methods for spatial pattern extrapolation, we applied the TCAPH-Trans method to simulate and predict the spatial patterns of typical Earth's surface elements (such as land-use change). We also used the TCAPH-Suit method to extrapolate the possibility of establishing a lunar scientific research station in the de Gerlache area of the lunar south pole. [Conclusions] Case applications demonstrate that the proposed TCAPH theoretical paradigm is effective for analyzing and extrapolating both primary and derived spatial patterns. The TCAPH-Trans and TCAPH-Suit methods can address the challenges of calculating scenario transition probabilities and site selection suitability probabilities, enabling characterization and extrapolation of spatial patterns with minimized probabilistic mapping errors. The methods and models proposed in this study are suitable for analyzing and supporting decisions for various spatial entities across different scenarios, effectively extending applications for spatial elements from Earth's surface to deep-space celestial objects. Future work will focus on developing an integrated platform system that combines spatial scene selection, spatial element analysis, and spatial pattern derivation to provide real-time and dynamic spatial pattern detection and evolution services, thereby expanding the application areas of the TCAPH model framework.

[Objectives] Rovers play an essential role in lunar exploration, serving as vital tools for scientists aiming to unravel the Moon's geological history and exploit its potential water-ice reserves. However, navigating the lunar surface with rovers presents significant safety risks due to the complex and often hazardous terrain, compounded by the lack of a consistent and reliable light source. The absence of pre-existing, high-resolution data—such as LiDAR—prior to exploration missions poses a considerable challenge in evaluating the safety of potential rover paths. Given these constraints, developing a reliable pre-assessment method is crucial for enhancing the success rate of lunar rover missions. [Methods] This paper introduces a 3D simulation method for lunar rover exploration, leveraging the Visualization Toolkit (VTK) to address these challenges. Our method integrates three critical aspects. Firstly, it offers high-resolution visualization of the lunar surface terrain, capturing intricate details down to the meter scale. Secondly, it simulates the dynamic illumination environment on the lunar surface, accounting for the varying illumination conditions due to the Moon 's rotation and orbital position. Thirdly, it models the rover's position and attitude transformations as it navigates the terrain. [Results] The effectiveness of this simulation approach is demonstrated through a case study focusing on the Shackleton Connecting Ridge region at the lunar South Pole, an area of significant interest due to its challenging topography and potential for water-ice deposits. The 3D simulation accurately depicts the undulating terrain of impact craters and allows for a thorough assessment of the rover's route safety by visualizing the potential hazards along the path. Moreover, the simulation offers an intuitive representation of the rover's movement, including real-time adjustments in position and attitude, which are critical for ensuring the rover’s stability and operational safety over long distances. Additionally, our method includes a real-time update feature for the dynamic illumination scene, enabling direct observation of how changing light conditions affect the rover's path during the mission. This capability is particularly important for assessing the feasibility of navigating through areas that may experience prolonged periods of darkness or extreme shadowing, which could impede the rover's progress or jeopardize its safety. The goal of this research is to improve the reliability and safety of future lunar rover missions by providing a robust pre-assessment tool that can verify the feasibility of proposed exploration routes. [Conclusions] This method thus offers crucial a priori information, serving as an essential guarantee for the successful execution of future lunar exploration endeavors.

[Objectives] Existing models for calculating spatial similarity relationships of point groups on maps often rely on questionnaires, expert evaluations, or predefined weights to select and determine feature indicator weights. However, this reliance introduces limitations in objectivity and generalizability. [Methods] To address these issues, this study undertakes the following research: (1) A systematic review, enumeration, and analysis of characteristic indices commonly used in spatial similarity models for point groups. The primary research focuses on attribute similarity, distribution range similarity, distribution density similarity, distance similarity, directional similarity, and topological similarity. (2) Several methods were employed to calculate the weight values and ranking of each feature index: the Analytic Hierarchy Process (AHP), the Fuzzy Comprehensive Evaluation Method(FCEM), Rough Set Theory (RS), and a composite weighting method integrating AHP and RS through the Lagrangian optimization decision model, referred to as the AHP-RS combined weighting method. (3) The weights derived from the AHP-RS combined weighting method were incorporated into the point group spatial similarity calculation model. [Results] The results show that： (1) Comparative analysis revealed that the AHP-RS combined weighting method not only maintains the same weight ordering as AHP and FCEM but also significantly reduces the unreasonable weight ordering issues caused by RS's high sensitivity to data. This method effectively decreases weight disparities, making the evaluation more reflective of actual conditions. (2) The results of point group spatial similarity calculations at various scales are highly consistent with human perception of point group similarity, further validating the rationality and effectiveness of the AHP-RS combined weighting method. [Conclusions] This study ultimately provides a set of characteristic index weights, ω_topology, ω_distance, ω_attribute, ω_direction, ω_scope, ω_density=(0.36, 0.15, 0.10, 0.15, 0.12, 0.12), offering objective and standardized weight values for point target spatial similarity calculation model. These weights establish a reliable theoretical foundation for spatial similarity calculations and comprehensive mapping.

[Objectives] Accurate rainstorm prediction plays an important role in disaster prevention and mitigation, industrial and agricultural production, and transportation, making it crucial for safeguarding social and economic development as well as people's property. However, existing intelligent rainstorm prediction methods fail to fully account for the uncertainties inherent in the rainstorm process itself, as well as in observation and modelling, which limits the improvement of prediction accuracy and stability. [Methods] To address this issue, an "estimation-correction" recurrent network based on filtering theory is proposed. This network estimates the meteorological state with the constraints of substantial derivative and corrects the state according to estimation errors, enabling accurate and reliable rainstorm prediction. The "estimation-correction" network consists of two main units, the state estimation unit and the state correction unit. Constrained by substantial derivative, the state estimation unit estimates the meteorological state and error for the next time step based on historical meteorological states. Guided by estimation error, the state correction unit corrects meteorological state by fusing estimation and observation errors. The two units work together to enhance prediction accuracy and stability. [Results] Experiments conducted on the ERA5 and NCEP reanalysis datasets demonstrate that the proposed method improves the Critical Success Index (CSI) of rainstorm prediction by 5% compared to other methods. Furthermore, it achieves good stability, as indicated by a stability metric (SPREAD≈0.5). [Conclusions] These results validate the feasibility of integrating filtering theory with deep learning to address uncertainty in rainstorm prediction.

[Objectives] The spatial distribution of the urban population in mountainous cities varies significantly due to the topography of the area. Information on the spatial distribution of populations in mountainous cities is crucial for scientific research, regional policymaking, resource allocation, and disaster assessment and protection. [Methods] Therefore, to accurately address the population distribution in these cities, this paper proposes a spatialization method based on residential identification and an improved random forest model. This method accounts for the spatial heterogeneity of population distribution in mountainous cities. To avoid assigning populations to non-residential areas, the grid of residential areas is first identified, and this grid is used during the feature variable selection stage. Next, a dataset of feature variables that reflect the characteristics of mountain cities is constructed. The spatial heterogeneity of population distribution is considered, and the feature variable set is clustered using the Gaussian Mixture Model algorithm. The Bootstrap sampling method is then used to randomly select an equal number of feature variables from each category, merging them into a new feature variable set, which is used as the training data to construct the Random Forest Model, thereby improving the traditional random forest approach. To verify the validity of the method, this paper uses Chongqing Municipality, a mountainous city, as the experimental area. The population spatialization results for Chongqing Municipality, based on a 150m grid, are obtained. These results are compared with those from the traditional random forest model, the WorldPop dataset, and the LandScan dataset. Additionally, the importance of each characteristic variable is measured using the random forest model. [Results] The experimental results show that the overall accuracy of the proposed method is 82.9%, which is 2.7% higher than that of the traditional random forest model and 2.94% and 10.91% higher than those of the WorldPop and LandScan datasets, respectively. Across the entire experimental area, compared to the WorldPop and LandScan datasets, the MAE (Mean Absolute Error) is reduced by 212.63 and 35.11, and the RMSE (Root Mean Square Error) is reduced by 1 354.34 and 524.54, respectively. In high-density hilly areas and mountainous regions, the proposed method yields better accuracy, demonstrating the effectiveness of the method in spatializing population data in mountainous cities. [Conclusions] In addition, the simulation results exhibit more distinct population distribution heterogeneity across different population density zones and topographic areas compared to the two open population datasets, offering richer insights into population density.

[Objectives] The impressive generalization capabilities of foundation models have made them a popular topic in the field of artificial intelligence. However, due to the unique characteristics of remote sensing imagery, these models cannot be directly applied to remote sensing visual tasks. Additionally, existing change detection methods often rely heavily on a substantial number of manually labeled samples and fail to effectively utilize available vector data, making them inefficient for automated detection. To address these challenges, this paper proposes a top-down change detection method that extracts valuable information from foundation models and vector data to minimize human involvement. [Methods] First, an improved Simple Linear Iterative Clustering (SLIC) algorithm is applied to segment and label bi-temporal images of the same area, constrained by vector boundaries, thereby generating training and testing datasets. Next, the generalization reasoning ability of the Contrastive Language-Image Pre-training (CLIP) model is leveraged to refine the training data. Then, a Channel Attention Bilinear Convolutional Neural Network (CAB-CNN) is employed for fine-grained scene classification. Finally, change rules and post-processing techniques are integrated to identify changing vector patches, with the Segment Anything Model (SAM) utilized to refine these patches. To validate the effectiveness and generalizability of the proposed method, two study areas are selected for quantitative experiments. The first is Guangling District, Yangzhou City, Jiangsu Province, using remote sensing images from 2022 and 2023, along with land use vector data from 2022. The second is Qianzhou Street, Huishan District, Wuxi City, Jiangsu Province, utilizing remote sensing images from 2018 and 2020, as well as land use vector data from 2018. [Results] The proposed method achieves 86.47% accuracy and 90.46% recall in Guangling District, representing improvements by 6.28% and 7.90%, respectively, compared to the pixel-based change detection method. In Qianzhou Street study area, the accuracy and recall reached 89.75% and 91.37%, respectively, exceeding the pixel-based method by 8.65% and 6.79%. [Conclusions] This method effectively utilizes existing vector data and foundation models to detect and refine changed vector patches with significantly reduced labor cost. It is particularly beneficial for applications such as detecting cropland conversion to non-agricultural use, identifying illegal land use, and updating forested land cover maps.

[Objectives] To address issues in existing remote sensing building extraction models, including poor feature representation ability due to redundancy, unclear building boundaries, and the loss of small buildings, [Methods] we propose a detail enhancement and cross-scale geometric feature sharing network (DCS-Net). This network consists of an Information Decoupling and Aggregation Module (IRDM), a Local Mutual Similarity Detail Enhancement Module (LMSE), and a Cross-scale Geometric Feature Fusing Module (CGFF), designed to guide small target inference. The IRDM module separates and reconstructs redundant features by assigning weights, thereby suppressing redundancy in both spatial and channel dimensions and promoting effective feature learning. The LMSE module enhances the accuracy and completeness of building edge information by dynamically selecting windows and specifying pixel clustering based on local mutual similarity between encoder-decoder features. The CGFF module computes the feature block relationships between the original image and various semantic-level feature maps to compensate for information loss, thereby improving the extraction performance of small buildings. [Results] The experiments in this paper are based on two public datasets: the WHU aerial dataset and the Massachusetts building detection dataset. The experimental results demonstrate the following: (1) Compared with existing building extraction algorithms such as UNet, PSPNet, Deeplab V3+, MANet, MAPNet, DRNet, Build-Former, MBR-HRNet, SDSNet, HDNet, DFFNet, and UANet, DCS-Net has achieved significant improvements across various evaluation metrics, demonstrating the effectiveness of the proposed method. (2) On the WHU dataset, the Intersection over Union (IoU), F1 score, and 95% Hausdorff Distance (95%HD) reached 92.94%, 96.35%, and 75.79%, respectively, outperforming the current best algorithm by 0.79%, 0.44%, and 1.90%. (3) On the Massachusetts dataset, the metrics were 77.13%, 87.06%, and 205.26, with improvements of 0.72%, 0.43%, and 13.84%, respectively. [Conclusions] These results indicate that DCS-Net can more accurately and comprehensively extract buildings from remote sensing images, significantly alleviating the issue of small building loss.

[Objectives] Unmanned Aerial Vehicle (UAV) and satellite remote sensing technologies have been successfully applied to estimate soil organic carbon and other attributes. However, their application to soil texture estimation remains relatively limited, highlighting the need for further research in this area. This study focuses on three farmland plots located in Zhuozhatan Village (Huzhu County), Nilongkou Village (Lalongkou Town, Huangzhong District), and Baitu Village (Lushar Town, Huangzhong District) within the Huangshui River Basin of Qinghai Province. It explores the potential of UAV and satellite remote sensing technologies for estimating soil texture content at the field scale. [Methods] Using UAV platforms equipped with two hyperspectral cameras, field-scale imaging of farmland soils was conducted. Additionally, a field spectrometer was used to collect in-situ soil spectra, and a total of 838 soil samples were collected from 2022 to 2024. Satellite imagery was also obtained for the same time periods, including GF1/2/7 (Gaofen 1/2/7), Sentinel-2A, and ZY1-02D (Ziyuan 1-02D). Laboratory analyses determined soil particle size distribution and acquired indoor soil spectral data. Based on these datasets, statistical modeling and soil texture content estimation were performed using the XGBoost (Extreme Gradient Boosting) method for laboratory, field in-situ, UAV, GF, ZY1-02D, and Sentinel-2 spectral data. Spatial distribution maps of soil texture content were then generated. [Results] ① Among the XGBoost model results, the highest model accuracy for UAV image spectra achieved an RPD (Ratio of Performance to Deviation) of 2.441, while the optimal RPD values for GF1/2/7, ZY1-02D, and Sentinel-2 satellite imagery were 1.815, 1.601, and 1.561, respectively. ② The estimation accuracy based on UAV and satellite imagery was lower than that derived from field spectrometer measurements. The accuracy ranking was as follows: laboratory spectra > field in-situ spectra > UAV image spectra > GF1/2/7 satellite image spectra > ZY1-02D satellite image spectra > Sentinel-2 satellite image spectra. Among soil texture components, clay content estimation showed the highest accuracy (RPD = 2.70), followed by silt (RPD = 2.24) and sand (RPD = 1.91). ③ Sand and clay content exhibited a negative correlation with soil spectral reflectance, whereas silt content displayed a positive correlation. The sensitive bands for sand, silt, and clay content were primarily concentrated in the near-infrared region (780~2 400 nm). ④ The content of sand, silt, and clay exhibited minor variations over three years, demonstrating relative stability. The mapping results for the three plots showed soil texture contents predominantly in the following ranges: 67% < sand ≤ 83%, 10.6% < silt ≤ 19.1%, and 3.2% < clay ≤ 6.6%. [Conclusions] At the field scale, UAV imagery was identified as the most effective data source for soil texture content mapping, providing strong support for precision agricultural management. While GF1/2/7 and ZY1-02D satellite imagery were found to be sufficient for texture mapping, Sentinel-2 satellite imagery was too coarse for field-scale mapping.

[Objectives] Revealing the structural characteristics of tourist flow networks is a prerequisite for achieving complementary advantages and coordinated development among attractions.[Methods] In this study, we employs methods such as travel chain extraction, social network analysis, and community detection to construct a research framework to analyze multi-scale tourist flow networks based on large-scale mobile phone data. The structural characteristics of the tourist flow network in Changsha are explored at microscopic, mesoscopic, and macroscopic scales.[Results] (1) Microscopic scale: The tourist flow network of Changsha shows a significant centralization trend, where a few core attractions such as the Yuelu Mountain and Orange Island have great influences on the whole network. Only 33% of attractions show structural hole efficiency and effectiveness above average, while their constraint is below average, indicating prominent structural holes and limited overall connectivity and efficiency. (2) Mesoscopic scale: The tourist flows of Changsha are highly concentrated, showing obvious spatial clustering characteristics and forming six tourism communities. There are usually two core attractions in each community to drive tourists to visit the surrounding attractions. In addition, the development of tourism communities is unbalanced, with a highly large community centered on Yuelu Mountain and Orange Island. (3) Macroscopic scale: The spatial distribution of the tourist flow network presents the characteristics of single-core strong concentration and overall dispersion, showing a multi-layer structure with the city center as the core and spreading outwards. The global efficiency of the network is only 0.367, with some marginal attractions having poor accessibility. The core attraction plays limited "trickle-down" effects on marginal attractions.

[Objectives] The influence of negative sample selection and machine learning models on landslide susceptibility evaluation cannot be overlooked. [Methods] To investigate the impact of these two factors on landslide susceptibility assessment, this study examines the Nujiang Valley section of the Nujiang River Basin. A weighted information quantity model was proposed to optimize negative sample selection. Thirteen influencing factors, including topography, land use, and average annual rainfall, were selected. Three machine learning models were employed: Support Vector Machine (SVM), Convolutional Neural Network (CNN), and Gradient Boosting Decision Tree (GBDT). A comparative analysis of landslide susceptibility was conducted against traditional random sample selection methods. Additionally, the effect of rainfall factors on susceptibility classification was analyzed. [Results] The results indicate that: (1) The optimized negative sample selection improved landslide density by 0.0103, 0.0639, and 0.004 0, respectively, for the three models. The AUC values increased by 0.033, 0.018, and 0.008, respectively. (2) Among the susceptibility evaluation models, the GBDT model performed best, improving accuracy by 3.8% and 1.7% compared to the SVM and CNN models, respectively. (3) Incorporating average monthly rainfall data for summer and winter (2019—2020) into the GBDT model revealed an increase in high and relatively high susceptibility zones during summer, particularly in the southern regions of Liuku Town and Shangjiang Town. [Conclusions] The optimization of negative samples based on the weighted information quantity model is reasonable and effective. As a landslide susceptibility evaluation model, the GBDT model is the most suitable for the disaster-prone environment of the Nujiang Valley, where precipitation significantly impacts landslide susceptibility.

[Objectives] Although the use of street view data to calculate the Green View Index (GVI) has emerged as a method for evaluating urban greening levels, systematic research on the spatiotemporal dynamics of GVI remains limited. [Methods] This study explores the spatiotemporal characteristics and influencing factors of urban GVI using street view big data, providing a new method for assessing urban street greening levels. This study proposes the GSENet semantic segmentation model for calculating and analyzing the GVI in Lanzhou's main urban area. The GSENet model incorporates a GSE-Block feature calibration module within its encoder, combining spatial and channel attention mechanisms. The decoder adopts an efficient self-attention module (Mix-transformer), which introduces a scaling factor and replaces the fully connected layer with a 1×1 convolution, combining the global modeling capability of Transformers with the local processing ability of convolution. Using the GSENet model, this study calculates the GVI of Lanzhou's main urban area based on Baidu Street View data and explores its spatiotemporal variation patterns through hotspot analysis, statistical analysis, and correlation analysis. [Results] The results reveal several key findings: (1) Utilizing ResNet50 as the backbone, the GSENet model achieves a Mean Intersection over Union (MIOU) of 74.7%, outperforming mainstream models such as PSPNet and DeepLabV3. The model demonstrates superior performance in identifying large-area categories such as vegetation and buildings, achieving an F1 score of 0.95. (2) Between 2019 and 2023, the average GVI increased by 2.3% compared to the period from 2014 to 2018. Notably, 70.9% of the sampled points showed a positive GVI trend, although only 8.4% experienced an increase greater than 10%. Anning District recorded the most substantial improvement, with a GVI rise of 3.5%, while Chengguan District saw the smallest growth, at only 1.9%. Spatial analysis identified that the central-western and northeastern parts of the study area experienced significant GVI increases, particularly in regions surrounding universities. In contrast, GVI declined notably in commercial centers and transportation hubs. (3) The influence of street view features and social factors on GVI changes exhibits spatiotemporal heterogeneity. Building density shows a negative correlation with GVI changes. The correlation between road width and GVI changes is relatively weak while the correlation between population density and GVI changes varies across different scales, with a stronger positive correlation at the street scale. [Conclusions] The experimental results highlight the effectiveness of this research in enhancing the perceived greening of urban streets. Furthermore, the findings provide valuable insights for urban planners aiming to optimize green space distribution and improve urban environments.

[Objectives] The increasing frequency and intensity of drought hazards, driven by climate change and human activities, have profoundly impacted the socio-economic systems and ecosystems surrounding Poyang Lake, the largest freshwater lake in China. Effective adaptive drought management plans require a thorough understanding of drought dynamics and the propagation processes from meteorological to hydrological drought. [Methods] This study introduces a Standardized Water Extent Index (SWI) to identify hydrological drought events in Poyang Lake from 2000 to 2023. The SWI is derived from reconstructed monthly water extent data based on Landsat and MODIS satellite imagery. Key drought event characteristics, including duration, interval, and severity, are quantified across various time scales (1-month, 3-month, and 6-month). Concurrently, the Standardized Precipitation Index (SPI), calculated using ERA5-Land monthly climate data, is employed to identify meteorological drought events in the river basins feeding Poyang Lake. The propagation and recovery lags between meteorological and hydrological droughts are analyzed using an event-based approach. [Results] The results reveal several key findings: (1) The reconstructed monthly water extent time series reveals significant interannual fluctuations in Poyang Lake, with coefficients of variation of approximately 0.13 and 0.12 for the annual maximum and minimum water extents, respectively. These findings demonstrate that Poyang Lake is vulnerable to drought during both the flood season (April to October) and the non-flood season (November to March). (2) At the 1-month time scale, the median duration, interval, and severity of severe droughts (-1.5<SWI≤-1.0) are 1 month, 6.5 months, and -0.25, respectively. For extreme droughts (SWI≤-1.5), the corresponding values are 1 month, 8 months, and -0.29. As the time scale increases, drought durations shorten, intervals between droughts lengthen, and severity intensifies. (3) The propagation time from moderate meteorological droughts to moderate hydrological droughts is approximately 2 months in the Xiushui, Xinjiang, and Raohe river basins, shorter than in other basins. This suggests that meteorological droughts in these basins have a more immediate influence on hydrological droughts in Poyang Lake. Recovery lags from meteorological droughts to hydrological droughts show strong consistency with the meteorological-hydrological propagation times across different time scales and drought severities. [Conclusions] The proposed standardized water extent index has proven to be effective and efficient in identifying hydrological drought events. Additionally, the methods for quantifying the propagation dynamics from meteorological droughts to hydrological droughts provide valuable contributions to predicting hydrological droughts. The findings offer critical insights into the characteristics of water extent variations and the interplay between meteorological and hydrological droughts, which can guide adaptive drought management and mitigation strategies in the Poyang Lake region.