The Guangdong-Hong Kong-Macao Greater Bay Area (GBA) is one of the most economically active regions in China. Its extensive fish pond aquaculture activities play an important role in the sustainable supply of regional aquatic products. However, the carbon balance contribution caused by the high-intensity aquaculture activities in these fish ponds is still lacking in quantitative research. This study used the Sentinel-3 Ocean and Land Colorimeter (OLCI) data, combined with a semi-analytical inversion algorithm based on key bands (Oa08, Oa09, Oa017), and performed atmospheric correction through the SeaDas software to improve the reflectance accuracy, providing reliable data for carbon concentration inversion, and then estimating the carbon concentration of phytoplankton in fish ponds in the GBA from 2016 to 2024.The results show that: ① Overall, the carbon concentration of phytoplankton in fish ponds in the GBA presents a spatial distribution of high in the northwest and low in the southeast, mainly showing a trend of gradually weakening from inland cities to coastal cities, and the concentration differences between different cities are obvious. ② The seasonal changes in the carbon concentration of phytoplankton in aquaculture fish ponds in the GBA are more significant than the interannual changes. The high concentration season is autumn, with an average concentration of 2.193 mg/L; followed by summer and winter, with 1.314 and 1.465 mg/L respectively; the lowest concentration is in spring, which is 0.632 mg/L. ③ The high concentration areas are mainly distributed in inland areas such as Zhaoqing City and Foshan City in the northwest, and the low concentration is mainly distributed in the coastal areas in the southeast. ④ The interannual variation of phytoplankton carbon concentration showed a statistically insignificant increasing trend, and showed a slight upward trend year by year in the Sen's slope estimation, with an annual increase rate of 0.01~0.05 mg/L, mainly concentrated in Huizhou, Foshan and Hong Kong. The study verified the application potential of the Sentinel-3 OLCI satellite in estimating carbon flux in aquaculture areas, clarified the spatiotemporal dynamic changes of phytoplankton carbon concentration in small fish pond ecosystems, and its important impact on the regional carbon cycle. The results provide a scientific basis for optimizing environmental management in the GBA and studying carbon dynamics in similar aquaculture systems.

As the core driving force of urban development, the spatial distribution and quantitative changes of population are of key significance to urban planning. Based on the GIS center of gravity model, grey GM(1, 1) model, BP neural network and GA-BP neural network, this study used the year-end permanent resident population data of Guangzhou City from 2010 to 2024 to explore the evolution path of population center of gravity and the future population number of Guangzhou City. By comparing the three forecasted population models, it was found that GA-BP neural network had the best prediction effect and predicted the population in 2035 accordingly; ① The results show that the population center of gravity in Guangzhou City shifted to the southeast from 2010 to 2024, with the largest shift from 2020 to 2024 and will continue to shift to the southeast in the future; ② It is estimated that by 2035, the population of Liwan, Yuexiu, Huangpu and Conghua will be basically stable, the population of Haizhu, Tianhe and Baiyun will decrease, and the population of Panyu, Huadu, Nansha and Zengcheng will increase; ③ The central urban area needs to continue to relieve the high-density population, and Nansha and Panyu should prevent the risk of “separation of industry and city” and simultaneously support public service facilities. The research results can provide a scientific basis for the territorial spatial planning and policy formulation of Guangzhou City, help to enhance the city's competitiveness and sustainable development capacity, and provide a reference for the high-quality development of urbanization in China.

Sugarcane provides over 70% of the world's sugar raw materials, making an accurate prediction of its maturity period is crucial for ensuring sugar production and optimizing harvest resource allocation. Satellite remote sensing offers a powerful tool for large-scale crop growth monitoring, yet challenges persist in understanding the underlying mechanisms of maturity prediction. This study addresses this gap by focusing on two key meteorological factors affecting sugarcane development—temperature and solar radiation—using effective accumulated degree days (AGDD) and accumulated solar radiation (ASR) as driving variables. We developed a hybrid remote sensing-meteorological maturity prediction model through three key steps. First, normalized difference vegetation index (NDVI) time series were reconstructed using Double-Logistic models with spatial-temporal fused Landsat-MODIS time series datasets. Second, three critical growth stages including germination, tillering and elongation, were extracted from the reconstructed NDVI time series. Third, the AGDD-based and ASR-based maturity prediction models initiated from each identified growth stage were developed. The predicted maturity dates were validated using the multi-year phenological data from Guangxi and Yunnan provinces. The validation results demonstrated that the ASR-based models consistently outperformed AGDD-based models across all tested growth stages, and models initialized from the germination stage achieved the highest accuracy (RMSE=17.7 days, RRMSE=5.3%). This research establishes a methodological framework for large-scale sugarcane maturity prediction, offering technological support for precise agriculture management in sugarcane production systems.

Driven by the digital economy, the digital transformation of the construction waste recycling industry has emerged as an inevitable pathway to advance the sector's in-depth integration and high-quality upgrading. Developing scientifically sound and effective transformation strategies is therefore critical to addressing the current bottlenecks constraining industrial development. To tackle the challenge of stakeholder collaboration amid this industrial transformation, this study first constructs a four-party evolutionary game model involving the government, developers, construction firms, and recycling enterprises. It then systematically analyzes the strategy selection logic of each stakeholder and the stability conditions of the system's equilibrium points. Further, the study integrates system dynamics (SD) to establish an SD-evolutionary game hybrid model. Scenario simulations are conducted focusing on core variables, including the scale of government subsidies, enterprises' digital transformation costs, transformation benefits for individual stakeholders, and collaborative transformation gains for two or three parties. These simulations help unveil the dynamic evolutionary patterns and feedback mechanisms underlying multi-stakeholder strategies, ultimately leading to the proposal of collaborative strategies for the industry's digital transformation. The results indicate three key findings: ① Government subsidies positively incentivize enterprises' willingness to transform, but such subsidies require dynamic adjustment—excessively high or low subsidy levels will diminish stakeholders' transformation motivation. ② The additional benefits gained by developers, construction firms, and resource utilization enterprises through collaborative transformation exhibit a significant positive correlation with each stakeholder's willingness to pursue digital transformation. ③ The improvement of regional digital infrastructure and the development level of the digital economy directly influence the effectiveness of transformation cost control, while enhancing collaborative innovation capabilities within the industrial chain constitutes the core pathway to increasing stakeholders' collaborative benefits. This research not only provides theoretical support for clarifying the multi-stakeholder behavioral logic of the construction waste recycling industry in the context of digital transformation but also offers practical references for governments to formulate industrial collaborative governance policies and for enterprises to optimize their transformation decisions.

"Zero-waste City" is a new urban development model and governance concept. Based on the evaluation index system of "Zero-waste City" construction pilot work plan in China, and 233 questionnaire data analysis, the research used the applied management entropy theory and busselator model, to investigate and evaluate the system dissipation structure of "Zero-waste City" management system. It concluded that, the dissipation structure of "Zero-waste City" management system in China has not yet been formed, and the main factors affecting the formation of "Zero-waste City" in China are the number of units for the construction of zero-waste cities, the amount of safe disposal of industrial hazardous materials, the coverage rate of medical waste collection and disposal systems, the amount of waste discharged from construction, the penetration rate of sanitary toilets in rural areas, the political completion rate of informal landfill sites, the formulation of policy regulations or policy documents in cities without waste, and the quantification of household waste. The formation of a "Zero-waste City" needs to reduce its internal positive entropy environment, and requires the government, enterprises, and the public to establish a sufficiently large negative entropy external environment to enable "Zero-waste City" management system in China to evolve into a dissipative structure. Finally, the suggestion was put forward to construct a "Zero-waste City".

MicroRNAs (miRNAs) play key regulatory roles on cellular functions such as apoptosis and angiogenesis. The pathogenesis of complex diseases such as cancer is fundamentally linked to dysregulation of these cellular functions. Quantifying impact on abnormal cellular functions of miRNAs enables an understanding of the mechanism of disease progression and identifying novel therapeutic targets. Although traditional molecular biological experiments can verify the relationship between miRNAs and abnormal cellular functions in diseases, they have significant limitations such as long experimental cycle and high cost. To address such problems, in this paper, the impact of miRNAs on abnormal cellular functions in disease by the shortest path between differentially expressed miRNAs and functional genes was quantified. The proposed method is applied to lung cancer and colorectal cancer datasets to identify miRNAs associated with abnormal disease functions. The results of literature and KEGG pathway enrichment analysis show that SP-MFA can effectively identify miRNAs associated with abnormal cellular functions in diseases.

The involvement of microRNAs (miRNAs) in various diseases has been established. Thus, discovering miRNA-disease associations (MDAs) aids in the understanding of disease pathogenesis and makes it easier to develop effective therapeutics that target miRNAs. Compared with traditional biological experimental methods, identifying MDAs through computational methods has been proven to be low-cost and efficient. This study introduces the MDAES method for MDA prediction by integrating miRNA expression data and similarity networks. In MDAES, the optimal kernels of miRNAs and diseases are first constructed through multiple kernel learning (MKL) from their similarity kernels. Next, MDAES applies a regression model to obtain feature representations of miRNAs as well as diseases. Finally, by integration of the similarity network features of miRNAs and diseases, as well as the features of miRNA expression data as input to the deep neural network (DNN), MDAs are predicted. As a result, MDAES exhibits good performance on three benchmark datasets (AUCs >0.95) and outperforms the currently employed methods. Case study assessments of both breast and lung cancers demonstrate that 96% and 100%, of the top 30 predictions are linked to the two cancers, respectively. Overall, MDAES has great potential as a reliable and practical model to identify MDAs.

Influenza viruses are major respiratory pathogens that can cause large-scale human outbreaks and pose a serious threat to global health. Many historical influenza pandemics have originated from swine or avian influenza viruses or reassortant hybrids. The development of interpretable and generalizable predictive models for human-to-human transmission is therefore critically important for early warning systems. This study presents a deep learning model incorporating attention mechanisms to predict the transmission risk of swine influenza virus to humans. The model combines convolutional neural networks and bidirectional gated recurrent units to capture local patterns and long-range dependencies in viral genomic sequences. Its key innovation lies in an interpretable attention mechanism that not only improves performance but also identifies and visualizes critical genomic regions related to host adaptation, such as PB1, PA, and HA segments. Using five-fold cross-validation, the model achieved an average AUROC exceeding 0.99, outperforming baseline models by approximately 0.150 in AUROC. The attention weights showed strong agreement with established biological evidence, confirming the model's biological relevance. In summary, the proposed model demonstrates high predictive accuracy and provides valuable support for public health surveillance and early warning systems against influenza transmission.

Vehicle-to-Infrastructure (V2I) technology, as a key component of the Vehicle Infrastructure Cooperative System (VICS), holds potential for alleviating traffic congestion, enhancing safety, and reducing emissions. This study aims to optimize signalized intersection control through V2I to minimize vehicle emissions. The spatiotemporal influence range of intersections was determined as 300 m based on DSRC communication. A dynamic speed control model and an emission model based on specific power method were established. Simulations were conducted via MATLAB to compare vehicle operational efficiency and emissions with and without V2I control. Simulation parameters included a control area of 600 m and traffic volume of 900 pcu/h. Results show that,V2I control significantly reduced emissions of CO, CO₂, NO_x, and HC by 13.5%, 9.6%, 19.9%, and 11.1%, respectively, while improving traffic efficiency, demonstrating its effectiveness in energy conservation and emission reduction.