To address challenges facing vision-language models in low-resource scenarios, such as lack of large-scale annotated data and effective transfer methods, a code-switching Chinese Minority pre-trained language model visual question answering (CCMPLM-VQA) method is proposed in this work. With a cross-lingual masked modeling approach using code-switching, model dependence on annotated training data is reduced. A language adapter (LA) with novel structures is introduced to effectively improve multimodal alignment of CCMPLM-VQA. The effectiveness of the proposed method is verified. Compared with the best benchmark model, CCMPLM-VQA improves zero-shot performance on real-world general visual reasoning dataset by approximately 12%. Additionally, its zero-shot performance on cross-lingual real-world general visual reasoning datasets also outperforms existing methods by about 1%.

A Chinese scene text editing (CSTE) method, based on research but incorporating uncertainty modeling and identifying effective technical solution, is proposed in this work. This new method optimizes prediction noise through an uncertainty-guided adjustment mechanism, improving the accuracy of noise estimation, thereby enhancing the clarity and structural integrity of the generated text. Additionally, by filtering irrelevant information from both textual and visual features, the method improves cross-modal alignment capabilities, achieving a seamless fusion of text and background textures.

To address insufficient modeling of inter-variable dependencies and suboptimal adaptation to spatiotemporal dynamics in multivariate time series forecasting, this paper proposes a Transformer network, STARFormer, incorporating spatiotemporal dimension reconstruction. This approach utilizes a segmented encoding mechanism that transforms single-dimensional temporal information into 2D vector matrices through dimension inversion. A dual-phase attention architecture is developed to hierarchically capture cross-temporal and cross-dimensional dependencies, effectively strengthening temporal representation learning. Further, this work introduces a dynamic graph module to model evolving dependencies between temporal patterns and spatial structures. Experimental evaluations across five real-world datasets demonstrate that STARFormer outperforms state-of-the-art Transformer-based models in multivariate forecasting tasks.

To address privacy concerns in federated learning under heterogeneous data environments, we propose a heterogeneous federated unlearning algorithm that integrates knowledge distillation (KD) with a forgetting mechanism. The proposed method extracts generic knowledge from the global model and distills it to local clients, thereby preserving the learning capability of local models, whilst effectively removing information associated with sensitive data, to enable privacy-oriented unlearning. The proposed approach is found to maintain strong model performance across diverse data structures and device configurations, while significantly enhancing data privacy. This study provides an effective technical solution for privacy preservation in heterogeneous federated learning settings.

A multimodal recommendation model that integrates high-order semantic enhancement with adaptive modal feature fusion (MMSAF) is proposed. The model employs graph convolutional neural network (GCNN) to perform high-order semantic enhancement, enabling the capture of deeper associations between users and items, thereby more accurately reflecting users' complex preferences. The effectiveness and applicability of this model are validated. An adaptive modality fusion mechanism is introduced to dynamically adjust weights of modal features based on their relative importance in different contexts, enabling flexible adaptation to diverse user preferences. Experimental results demonstrate that MMSAF outperforms existing mainstream methods across multiple benchmark datasets in terms of recommendation accuracy and generalization capability.

Aiming at the collaborative problem of resource load balancer and energy consumption optimization in cloud data center virtualization deployment, an efficient method is proposed to achieve some balance between these two objectives. Specifically, a load balancing model that integrates the multi-dimensional resource utilization balance is constructed, along with an energy consumption model that reflects energy loss under different system operating states. A bi-objective optimization function with resource constraints is then formulated. An improved hybrid sparrow search algorithm (HSSA) is then designed to minimize this optimization function. Cloud simulations show that, in a heterogeneous cloud environment with 200 nodes, the proposed method improves performance by 45.28% compared to the ant colony system (ACS), by 58.06% and 8.38% respectively compared to the first fit decreasing (FFD) algorithm and genetic algorithm (GA).

Mammalian Rpn13, a 26S proteasome ubiquitin receptor subunit, binds to ubiquitin substrate at the N terminal but deubiquitin enzyme UCH37 at the C terminal. Rpn13 is known to mediate degradation of most proteins via ubiquitin-proteasome, to regulate autophagy, neuron function, and germ cell development. Proteasome inhibitors are used for tumor prevention and treatment, targeting Rpn13 can overcome resistance against proteasome inhibitors. Transcriptional repression mechanism of its gene expression is unclear. Interactive analysis of gene expression profile revealed that Rpn13 gene was highly expressed in many cancer types, especially in colon cancer, lymphoma, pancreatic, rectal, gastric and thoracic cancers. The activity of the regulatory region −454~−58 bp of Rpn13 promoter was found to be high in double luciferase report assay. Binding sites of multiple transcription factors, including CTCFL (CCCTC binding factor-like), CTCF (CCCTC binding factor), NRF1 (Nuclear respiratory factor 1) and FOXO3 (Forkhead box O3), were predicted on the promoter sequence of Rpn13. We constructed expression plasmids of these transcription factors and found that they all inhibited Rpn13 promoter. By further mutating corresponding transcription factor binding sites of Rpn13 promoter, CTCFL was found to bind −63~−52 bp sequence of the Rpn13 promoter. These studies not only analyzed Rpn13 gene expression in cancer tissues, but also revealed possible transcription inhibition mechanism of Rpn13, providing useful clues for exploring mechanism of Rpn13 in tumor occurrence and development.

The aim of this work was to investigate the expression and prognostic significance of microtubule-associated serine/threonine kinase 1 (MAST1) in glioma tissues. MAST1 mRNA expression in tumor tissues and prognosis of patients with gliomas were analyzed with the GEPIA 2 database. mRNA seq~~325 and mRNA seq~~693 transcriptome data and clinical data were obtained from the Chinese Glioma Genome Atlas (CGGA) database, differential expression of MAST1 mRNA in glioma tissues with different clinical features was analyzed. Impact of MAST1 expression on prognosis of patients with glioma was evaluated using Kaplan-Meier survival curves and multivariate Cox regression analysis. Potential signaling pathways involved in MAST1 were explored by Gene Set Enrichment Analysis (GSEA) software. MAST1 mRNA expression was found to be low in glioma tissues. MAST1 mRNA expression was correlated with patient age, WHO grade, IDH mutation, and 1p/19q co-deletion status. Low MAST1 expression was found associated with poor prognosis in glioma patients. Further enrichment analysis revealed involvement of MAST1 expression in multiple biological processes. We conclude that MAST1 is a potential indicator for prognostic evaluation in glioma patients.

Miyun Reservoir is located in the East Asian-Australasian flyway. The reservoir is a habitat and feeding ground for birds, ecologically important in biodiversity conservation. At the present, little is known about the diversity of birds in Miyun Reservoir. Bird observation data platforms are used to examine bird diversity in Miyun Reservoir, to provide information for conservation and monitoring of bird resources. The composition and diversity of birds in Miyun Reservoir were studied from January 2020 to December 2022 using the point-count sampling method. A total of 20 orders, 65 families and 324 species of birds were recorded, accounting for 62.4% of the recorded bird species in Beijing. Out of these birds, 14 species are Nationally First-Grade Protected Animals and 53 species are Nationally Second-Grade Protected Animals. Passage migrants, residents, summer visitors and winter visitors accounted for 38.9%, 13.6%, 28.7% and 16.4% of the total respectively. From 2020 to 2022, no significant differences were found in bird species richness (P = 0.591), Shannon-Weiner index (P = 0.995), Pielou index (P = 0.806) and Simpson index (P = 0.144). Species composition and diversity of bird species continued to remain stable during those three years. The highest values of the Shannon-Weiner index all occurred in summer, indicating that the Miyun Reservoir is rich in food sources in summer, providing suitable habitat for birds. The Pielou index for the summers of 2021 and 2022 were significantly higher than those for the other three seasons, indicating that the bird community structure tended to be stable in summer after peak spring migration, proportion of each bird species is relatively balanced. The highest value of Simpson index appears in autumn, when birds are migrating. The number and relative abundance of winter visitors such as Bean goose, Daurian Jackdaw and Ruddy Shelduck are increased, thereby raising the Simpson index. Miyun Reservoir therefore is rich in bird resources and has important biodiversity protection value. It is necessary to strengthen ecological and environmental protection, to carry out bird habitat restoration work, to establish long-term and stable ecological beaches, and to continue to carry out ecological restoration research and conservation effectiveness assessment, for effective protection and management of bird diversity.

With rapid development of the Chinese economy in recent years, intensity of water resources development and utilization has increased significantly, leading to degradation of some water ecosystems. Assessment of the health status of river and lake ecosystems has therefore become a top priority in current water environment governance and protection efforts. This study focuses on the Yellow River and Nansi Lake basins in Shandong province. In-depth research is done on physical morphology, hydrology and water resources, water environmental quality, and aquatic biological conditions. Principal component analysis, entropy method, and AHP hierarchical analysis are used, to establish a comprehensive evaluation system for water ecological health in Shandong province, and to conduct comprehensive evaluation of water ecological health of 12 typical river and lake water systems. The Shandong section of the Yellow River basin is found to maintain an overall healthy state, but some tributaries are sub-healthy due to insufficient ecological flow, dam construction, and damage to benthic organisms and macrophyte diversity. The overall health of the Nansi Lake basin is good, but Zhuzhaoxin River, Xinwanfu River, Guangfu River, and Baima River are sub-healthy due to natural shoreline damages, sediment pollutions, decreased diversity of macrobenthos and attached algae. Water ecological protection strategies are proposed hereinafter: to strengthen riverbank management, improve ecological flow protection mechanisms, control local sources of pollution. Construction of an efficient and scientific water ecological monitoring, evaluation, and management system is of great significance for promoting sustainable water resources utilization and ecological environment protection.

Sea cucumber (Apostichopus japonicus) is an important economic species in the Bohai and Yellow Seas of China. In recent years, due to anthropogenic factors, the marine environment has undergone dramatic changes, and the wild sea cucumber resources are facing a crisis. Restoration of sea cucumber populations and development of artificial aquaculture facilities would require assessment of the suitability of their habitats. Principles of ecological dynamics, physiological characteristics and behavioral habits of sea cucumbers are used in this study to select key habitat factors related to the growth of sea cucumbers. Two-dimensional hydrodynamic model of the Bohai and Yellow Seas and ecological growth model for sea cucumbers are used to develop a habitat suitability index (HSI) model for sea cucumbers. A new habitat factor, water flow velocity, was incorporated. A combination of subjective and objective weighting methods were used to make the sea cucumber habitat suitability assessment more comprehensive and objective. This new model was used to assess the habitat suitability of sea cucumbers in the Bohai Sea and the northern Yellow Sea, providing theoretical support for habitat restoration and fishery production.

With the accelerated pace of urbanization, the urban hydrological cycle has undergone significant changes, leading to complication of rainfall runoff characteristics, frequent occurrences of waterlogging events, and increasingly severe water pollution issues. As an effective tool to address these challenges, storm water management model (SWMM), with its powerful simulation functions and flexibility, has gained widespread application in urban rainfall runoff simulation, urban drainage system planning and design, urban non-point source pollution simulation, and LID (low impact development) facility effect analysis. This paper reviews the current application status of the SWMM model, analyzes the main existing problems in SWMM research, and proposes prospects for the functional improvement and practical application of the model.

Land use change data in Zaozhuang, Shandong Province from 2000-2010 and from 2010-2020 are analyzed. Land use transfer matrices, land use dynamic degree, geometric features, and NMDS ordination are used to examine land use changes in both temporal and spatial dimensions from the perspectives of area, rate, and morphology. The land use types in Zaozhuang are primarily cultivated land, grassland, urban, industrial, and residential areas, with distinct spatial boundaries. Land use changes in 2000-2010 were found to be minimal, classified as extremely slow, with geometric characteristics of changing water areas differing markedly from other land types, heavily influenced by natural factors. Land use changes in 2010-2020 became pronounced, with mutual transformations occurring among all land use types, particularly in cultivated land, urban, industrial, residential areas, and grassland. The geometric features of these three types were similar but significantly different from other categories. During this later period, with the development of agriculture and industry, land structure in Zaozhuang became more rational, and the level of land intensification improved. This study provides new insights and recommendations for researching land use changes from a spatial morphology perspective.

Induction and detection of tunneling nanotubes: comprehensive experimental design and practice. One important way to improve quality of undergraduate practical class is to transform cutting-edge theories and research techniques of cell biology into practical experiments. Knowledge of tunneling nanotubes (TNT) is used to design a comprehensive and exploratory experiment, including establishment of serum-starvation model and tunneling nanotubes induction, observation of tunneling nanotubes structure by staining of cell microfilaments and nuclei, and detection of cell cycle by flow cytometry. Students would first establish serum-starvation model by decreasing serum concentration for different cell types, then use cell crawler for fluorescence staining, and observe the bridge-like TNT structure under a fluorescence microscope. Flow cytometry is used to detect changes of the proportion of G1 cells in the cell cycle in starvation. These experiments facilitate student hands-on experience and stimulate their interest in exploration.

Constructing a prediction model for archaeological sites can accurately identify the potential spatial extent of archaeological sites, which can help to prospectively protect undiscovered archaeological sites. A collection of 1 916 Neolithic archaeological sites (excluding burial sites) in Shandong Province was studied in this paper, with 1916 non-site locations used as negative controls (at a 1:1 ratio). Eight natural geographic environmental influencing factors (elevation, slope, slope orientation, profile curvature, planar curvature, micro-geomorphology, slope position, and distance to water boundaries) were selected to construct an index system for archaeological site prediction model. Using GIS spatial analysis method and the XGBoost algorithm, we constructed a predictive model for archaeological sites, which is used to predict the spatial extent of potential sites and analyze the importance of influencing factors. The archaeological site prediction model constructed with XGBoost algorithm achieves high accuracy, with an AUC value of 0.85. The model trained with optimal samples categorizes the results into low, medium, and high grades, mapped the spatial distribution of site probabilities, and revealed that sites primarily cluster in plains. Based on the XGBoost algorithm for ranking importance of influencing factors, slope position, microgeomorphology and elevation are the top three factors influencing the siting of Neolithic archaeologicalsites. The XGBoost algorithm has good stability and prediction ability. The constructed model provides a new research method for prediction of archaeological sites, and provides important technical support for field archaeological excavation and elucidates the relationship between prehistoric settlement patterns and geographical environments.

Forest fires are major natural disasters and key to the loss of forest biodiversity and degradation of ecosystem functions. Fuel moisture content (FMC) is closely related to occurrence of forest fires and is a critical parameter in forest fire risk assessment and in calculating fire spread rates. In this study, we used Google Earth Engine (GEE) platform to generate spatiotemporal distribution maps of global FMC and analyzed spatiotemporal differences in FMC across various regions and vegetation types. First, ground-measured data and spectral indices are used to construct empirical estimation models for equivalent water thickness (EWT) and dry matter content (DMC) for forest and grassland vegetation types, respectively. Then, EWT and DMC are used as intermediate variables, to develop a stepwise estimation model for FMC, to ultimately produce global monthly FMC spatiotemporal distribution maps for the period from 2010 to 2022. To assess estimation accuracy, both direct and indirect validations are conducted by integrating site data (Global-LFMC) and standardized precipitation evapotranspiration index (SPEI) data. For forests, normalized difference infrared index (NDII) is used to estimate EWT (r = 0.89) and normalized difference water index (NDWI) is used to estimate DMC (r = 0.26). For grasslands, simple ratio water index (SRWI) is used to estimate EWT (r = 0.85) and NDWI used to estimate DMC (r = 0.27), both yielding high accuracy. Thus, these methods are used to construct stepwise estimation model for FMC. Direct validations indicate that the correlation coefficient between estimated FMC and measured values is 0.79 (with a significance level of P < 0.000 1), demonstrating significant correlation. Indirect validations reveal that FMC exhibits a trend consistent with SPEI, albeit with a temporal lag. The spatiotemporal dynamics of global vegetation FMC is then further explored. FMC shows significant seasonal variations and marked differences between the Northern and Southern Hemispheres as well as among different vegetation types. In tropical regions, FMC values remain relatively high throughout the year with minimal fluctuations. In temperate regions, FMC values increase significantly during the spring and summer. In polar and arid regions, FMC values are relatively low year-round. In mid-to-high latitude areas, FMC is significantly influenced by extreme seasonal climate events. Areas with low FMC coincide with regions of high fire occurrence. Evergreen forests maintain high moisture content throughout the year, whereas FMC in deciduous forests and grasslands fluctuates markedly with seasons. Spatial distribution characteristics of FMC outliers are confirmed. Analysis of these outliers reflects the significant impact of factors such as climate and sunlight on vegetation water status, further elucidating influence of extreme climate on FMC. MODIS-derived global monthly FMC spatiotemporal distribution maps generated for 2010–2022 reveal FMC distribution characteristics across different regions and vegetation types, providing ground for formulating scientific prevention and control strategies in high fire-risk areas, and they are of significant values for forest fire prevention and ecosystem management.

Relationship between distribution of field songs in Hubei and geographic environment in which the field songs are nurtured is explored in this paper, to provide new ideas and methods for empirical research on regional music. A total of 1 248 sample data sets of field songs are selected. Geographic information system (GIS) is used to build a database for distribution of preliminarily selected field songs and influencing factors of the music elements. An analytical model is constructed of the system of influencing factors of the field songs based on random forest and shapley additive explanations (SHAP) interpretable algorithms. Validity of the model is evaluated through the receiver operating characteristic (ROC) curve, to analyze field songs distribution, relationship between music elements and geographic environment, and to analyze relationship between distribution of field songs, music elements and geographic environment. The model of the influence factor system of field songs constructed based on the random forest has a good prediction effect, and its area under the curve (AUC) value is 0.82. Importance ranking of influence factors for generation of field songs and music elements shows that multi-year average rainfall and multi-year average temperature are the main factors for the cultivation of field songs in Hubei. Random Forest and SHAP algorithms can predict distribution patterns of Hubei field to a certain extent, which is of great significance to the study of regional music culture and geographic correlation.

Statistics from hourly wind speed collected between 1960 and 2020 from 12 meteorological stations in the Hexi Corridor indicate that the western region of the corridor is located in an intermediate-to-high wind energy environment, whilst the eastern region is located in a low wind energy environment. Variation rules of annual average wind speed, probability of sand-driving wind, and drift potential differ between eastern and western regions of the corridor. WeibullCum distribution function is universally applicable to express annual cumulative probability distribution of hourly wind speeds at various meteorological stations. Interannual variation of constant term B in WeibullCum function exhibits periodic fluctuation, while C and D show a decreasing trend. The prediction equations of cumulative probability distribution function of wind speed are then established for eastern and western regions of Hexi Corridor, respectively. It is observed that prediction equations have high accuracy for annual average wind speed and sand-driving wind probability, whereas simulated drift potential is in overall less than the actual results. Actual statistical results from 2021 to 2024 further confirm the above conclusions. Wind variations exhibit distinct cycles in eastern and western regions of the Hexi Corridor. In the next 10 years, it is expected that peak values for annual average wind speed and sand-driving wind probability will occur around 2025 in the eastern region, but later than 2030 in the western region.

Increased frequency of intense typhoons is seen in recent years, particularly Super Typhoon “Yagi” (Typhoon No. 2411), which has caused substantial economic losses and social impacts in Hainan province in China. In this study the wind disaster assessment method based on accumulated wind energy is compared, thereby mitigating the limitations of traditional assessments based solely on maximum wind speed. The accumulated wind energy model is found to more precisely reflect actual disaster losses, especially in high-vulnerability areas such as Wenchang and Haikou. The coverage area of winds greater than level 12 is 1 473 km2, level 13 is 933 km2, level 14 is 101 km2. The overall wind energy of “Yagi” is higher than that “Rammasun”, and the total wind energy of Typhoon “Yagi” in Wenchang city is about 13 times that of “Rammasun”, at 500 000 kJ. This study reveals influence of factors such as typhoon wind speed, duration, and terrain on accumulated wind energy, thus elucidating regional disparities in wind disaster risk. Under the background of climate change, it is recommended to improve wind resistance design standards and infrastructure resilience based on typhoon wind energy estimation results, especially for rural housing, agricultural facilities, electricity, communication and other infrastructure.

Quantifying the dynamics of grasslands in Xinjiang's arid regions and their responses to key drivers is crucial for effective vegetation management in this ecologically fragile area. In the present study leaf area index (LAI) data collected during growing seasons from 2001 to 2020 were analyzed. Multiple analytical methods, including Sen's trend analysis, Mann-Kendall significance test, coefficient of variation, Hurst exponent, and structural equation modeling, were employed, to examine the spatiotemporal patterns of Xinjiang grasslands and their responses to topography, climate, and human activities. Spatial analysis revealed significant heterogeneity. The multi-year average LAI value from 2001 to 2020 was found to be 0.37 m&#x2212;2·m&#x2212;2, characterized by higher values in the north and west, lower values in the south and east. LAI exhibited a fluctuating upward trend with an average annual growth rate of 0.004 a&#x2212;1. Trend analysis indicated that approximately 67.3% of the areas showed improvement, while 19.2% experienced degradation, with degraded areas primarily concentrated in the vicinity of the Ili River valley and Tacheng prefecture. Further, structural equation model revealed that topography, climate, and livestock factors impacted substantially spatial distribution of LAI in Xinjiang grasslands. Precipitation and wind speed were identified as dominant factors affecting LAI, with total path coefficients respectively of 0.34 and &#x2212;0.32. Topography, through interactions with climate and livestock numbers, exerted a significantly positive impact on LAI. Overall, precipitation and altitude demonstrated positive influences, whereas wind speed, temperature, and solar radiation exhibited negative effects. Between 2001 and 2020, LAI of Xinjiang grasslands displayed marked fluctuations, primarily attributed to combined effects of natural environmental and human activity factors.