Search Results (643)

In the era of smart agriculture, accurate identification of plant species is critical for effective crop management, biodiversity monitoring, and the sustainable use of medicinal resources. However, existing deep learning approaches often underperform when applied to fine-grained plant classification tasks due to the lack of domain-specific, high-quality datasets and the limited representational capacity of traditional architectures. This study addresses these challenges by introducing a novel, well-curated leaf image dataset consisting of 39 classes of medicinal and aromatic plants collected from the Aromatic and Medicinal Plant Research Station in Odakkali, Kerala, India. To overcome performance bottlenecks observed with a baseline Convolutional Neural Network (CNN) that achieved only 44.94% accuracy, we progressively enhanced model performance through a series of architectural innovations. These included the use of a pre-trained VGG16 network, data augmentation techniques, and fine-tuning of deeper convolutional layers, followed by the integration of Squeeze-and-Excitation (SE) attention blocks. Ultimately, we propose a hybrid deep learning architecture that combines VGG16 with Batch Normalization, Gated Recurrent Units (GRUs), Transformer modules, and Dilated Convolutions. This final model achieved a peak validation accuracy of 95.24%, significantly outperforming several baseline models, such as custom CNN (44.94%), VGG-19 (59.49%), VGG-16 before augmentation (71.52%), Xception (85.44%), Inception v3 (87.97%), VGG-16 after data augumentation (89.24%), VGG-16 after fine-tuning (90.51%), MobileNetV2 (93.67), and VGG16 with SE block (94.94%). These results demonstrate superior capability in capturing both local textures and global morphological features. The proposed solution not only advances the state of the art in plant classification but also contributes a valuable dataset to the research community. Its real-world applicability spans field-based plant identification, biodiversity conservation, and precision agriculture, offering a scalable tool for automated plant recognition in complex ecological and agricultural environments. Full article

This study presents a comprehensive case of green remodeling applied to a local fire station in Seoul, South Korea. The project aimed to improve energy performance through an integrated upgrade of passive systems (exterior insulation, high-performance windows, and airtightness) and active systems (electric heat pumps, energy recovery ventilation, and rooftop photovoltaic systems), while maintaining uninterrupted emergency operations. A detailed analysis of annual energy use before and after the remodeling shows a 44% reduction in total energy consumption, significantly exceeding the initial reduction target of 20%. While electricity use increased modestly during winter due to the electrification of heating systems, gas consumption dropped sharply by 63%, indicating a shift in energy source and improved efficiency. The building’s airtightness also improved significantly, with a reduction in the air change rate. The project further addressed unique challenges associated with continuously operated public facilities, such as insulating the fire apparatus garage and executing phased construction to avoid operational disruption. This study contributes valuable insights into green remodeling strategies for mission-critical public buildings, emphasizing the importance of integrating technical upgrades with operational constraints to achieve verified energy performance improvements. Full article

Cowpea (Vigna unguiculata [L.] Walp) is a vital food security crop in sub-Saharan Africa, including Eswatini. The productivity of the crop is low (<600 kg/ha) in the country due to a lack of improved, locally adapted, and farmer-preferred varieties with biotic and abiotic stress tolerance. The objective of the study was to assess the agronomic performance of newly developed elite cowpea mutants to select best-yielding and adapted pure lines for production and genetic improvement in Eswatini. A total of 30 cowpea genotypes, including 24 newly developed advanced mutant lines, their 3 founder parents and 3 local checks, were profiled for major agronomic traits in two selected sites (Lowveld Experiment and Malkerns Research Stations) using a 6 × 5 alpha lattice design with three replications. A combined analysis of variance revealed that the genotype x location interaction effects were significant (p < 0.05) for germination percentage (DG %), days to flowering (DTF), days to maturity (DMT), number of pods per plant (NPP), pod length (PDL), number of seeds per pod (NSP), hundred seed weight (HSW), and grain yield (GYD). Elite mutant genotypes, including NKL9P7, BRR4P11, SHR9P5, and NKL9P7-2 exhibited higher grain yields at 3158.8 kg/ha, 2651.6 kg/ha, 2627.5 kg/ha, and 2255.8 kg/ha in that order. The highest-yielding mutant, NKL9P7, produced 70%, 61%, and 54% more grain yield than the check varieties Mtilane, Black Eye, and Accession 792, respectively. Furthermore, the selected genotypes displayed promising yield components such as better PDL (varying from 13.1 to 26.3 cm), NPP (15.9 to 26.8), and NSP (9.8 to 16.2). Grain yield had significant positive correlations (p < 0.05) with DG %, NSP, and NPP. The principal component analysis (PCA) revealed that 81.5% of the total genotypic variation was attributable to the assessed quantitative traits. Principal component (PC) 1 accounted for 48.6%, while PC 2 and PC 3 contributed 18.9% and 14% of the overall variation, respectively. Key traits correlated with PC1 were NPP with a loading score of 0.91, NSP (0.83), PDL (0.73), GYD (0.68), HSW (0.58), DMT (−0.60), and DTF (−0.43) in a desirable direction. In conclusion, genotypes NKL9P7, BRR4P11, SHR9P5, NKL9P7-2, Bira, SHR3P4, and SHR2P7 were identified as complementary parents with relatively best yields and local adaptation, making them ideal selections for direct production or breeding. The following traits, NPP, NSP, PDL, GYD, and HSW, offered unique opportunities for genotype selection in the cowpea breeding program in Eswatini. Full article

Background: Proximal gastrectomy (PG) with double tract reconstruction (DTR) offers organ preservation for early gastric cancers, leading to reduced vitamin B12 deficiency, less weight loss, and improved quality of life. The JCOG1401 study confirmed excellent long-term outcomes for PG in stage I gastric cancer. However, in locally advanced proximal gastric cancer (LAPGC), preserving the gastric body and lymph node station 4d may compromise margin clearance and adequate lymphadenectomy. Methods: We propose a modified PG that removes the distal esophagus, gastroesophageal junction (GEJ), cardia, fundus, and gastric body, preserving only the antrum and performing DTR. Lymphadenectomy is also adapted, removing stations 1, 2, 3a, 4sa, 4sb, 4d, 7, 8, 9, 10 (spleen preserving), 11, and lower mediastinal nodes (stations 19, 20, and 110), while preserving stations 3b, 5, and 6. Indications for this procedure include GEJ (Siewert type II and III) and proximal gastric cancers with ≤2 cm distal esophageal involvement and ≤5 cm gastric involvement. Results: In our initial experience with 14 patients, we achieved R0 resection in all patients, adequate lymph node harvest (median 24 nodes, IQR 18–38), and no locoregional recurrences at a median follow-up of 18 months. We also found favorable postoperative weight loss, reflux, and anemia in the PG cohort. Conclusion: While larger studies and long-term data are still needed, our early results suggest that modified PG—despite sparing only the antrum—retains the key benefits of PG over total gastrectomy, including better weight maintenance and improved hemoglobin levels, while maintaining oncologic outcomes for LAPGC. Full article

Focusing the over simplification of existing models in simulating river flow evolution process and lack of coordination of hydropower generation and export, this paper proposes a hydropower generation and export coordinated optimal operation model that, at the same time, incorporates dynamic water flow delay by finely modeling the water flow evolution process among cascade hydropower stations within a river basin. Specifically, firstly, a dynamic water flow evolution model is built based on the segmented Muskingum method. By dividing the river into sub-segments and establishing flow evolution equation for individual sub-segments, the model accurately captures the dynamic time delay of water flow. On this basis, integrating cascade hydropower systems and the transmission system, a hydropower generation and export coordinated optimal operation model is proposed. By flexibly adjusting the power export, the model balances local consumption and external transmission of hydropower, enhancing the utilization efficiency of hydropower resources and achieving high economic performance. A case study verified the accuracy of the dynamic water flow evolution model and the effectiveness of the proposed hydropower generation and export coordinated optimal operation model. Full article

Air quality (AQ) modeling is at the forefront of estimating pollution levels in areas where the spatial representativity is low. Large metropolitan areas in Asia such as Beijing face significant pollution issues due to rapid industrialization and urbanization. AQ nowcasting, especially in dense urban centers like Beijing, is crucial for public health and safety. One of the most popular and accurate modeling methodologies relies on black-box models that fail to explain the phenomena in an interpretable way. This study investigates the performance and interpretability of Explainable AI (XAI) applied with the eXtreme Gradient Boosting (XGBoost) algorithm employing the SHapley Additive exPlanations (SHAP) and the Local Interpretable Model-Agnostic Explanations (LIME) for PM_2.5 nowcasting. Using a SHAP-based technique for dimensionality reduction, we identified the features responsible for 95% of the target variance, allowing us to perform an effective feature selection with minimal impact on accuracy. In addition, the findings show that SHAP and LIME supported orthogonal insights: SHAP provided a view of the model performance at a high level, identifying interaction effects that are often overlooked using gain-based metrics such as feature importance; while LIME presented an enhanced overlook by justifying its local explanation, providing low-bias estimates of the environmental data values that affect predictions. Our evaluation set included 12 monitoring stations using temporal split methods with or without lagged-feature engineering approaches. Moreover, the evaluation showed that models retained a substantial degree of predictive power (R² > 0.93) even in a reduced complexity size. The findings provide evidence for deploying interpretable and performant AQ modeling tools where policy interventions cannot solely depend on predictive analytics tools. Overall, the findings demonstrate the large potential of directly incorporating explainability methods during model development for equal and more transparent modeling processes. Full article

In contemporary wireless communication systems, achieving precise localization of communicating devices and accurate channel estimation is crucial for enhancing operational efficiency and reliability. This study introduces a novel approach that integrates the localization task and channel estimation into a single framework. We present a multi-task neural network architecture capable of simultaneously estimating channels from multiple base stations in a blind manner while estimating user terminal coordinates in given indoor environments. This approach exploits the relationship between channel characteristics and spatial information, using the same channel state information (CSI) data to perform both tasks with a single model. We evaluate the proposed solution, assessing its effectiveness across differing antenna spacing configurations and indoor test environments using both WiFi and 5G orthogonal frequency-division multiplexing (OFDM) systems. The results show performance benefits, achieving comparable channel estimation results to other studies while simultaneously providing a localization estimate, resulting in reduced model overhead while leveraging spatial context. The presented system demonstrates potential to improve the efficiency of communication systems in real-world applications, aligning with the goals of emerging integrated sensing and communication (ISAC) systems. Results based on experimental data using the proposed solution show a 50th percentile localization error of 1.62 m for 3-tap channels and 0.89 m for 10-tap channels. Full article

In a multi-cell network, interference management between adjacent cells is a key factor that determines the performance of the entire cellular network. In particular, in order to control inter-cell interference while providing a high data rate to users, it is very important for the base station (BS) of each cell to appropriately control the transmit power in the downlink. However, as the number of cells increases, controlling the downlink transmit power at the BS becomes increasingly difficult. In this paper, we propose a multi-agent deep reinforcement learning (MADRL)-based transmit power control scheme to maximize the sum rate in multi-cell networks. In particular, the proposed scheme incorporates a long short-term memory (LSTM) architecture into the MADRL scheme to retain state information across time slots and to use that information for subsequent action decisions, thereby improving the sum rate performance. In the proposed scheme, the agent of each BS uses only its local channel state information; consequently, it does not need to receive signal messages from adjacent agents. The simulation results show that the proposed scheme outperforms the existing MADRL scheme by reducing the amount of signal messages exchanged between links and improving the sum rate. Full article

This study presents a flood risk assessment of five rural bridges along the monsoon-prone Khar–Mohmand Gat corridor in Northwestern Pakistan using an integrated hydrologic and hydraulic modeling framework. Hydrologic simulations for 50- and 100-year design storms were performed using the Hydrologic Engineering Center’s Hydrologic Modeling System (HEC-HMS), with watershed delineation conducted via Geographic Information Systems (GIS). Calibration was based on regional rainfall data from the Peshawar station using a Soil Conservation Service Curve Number (SCS-CN) of 86 and time of concentration calculated using Kirpich’s method. The resulting hydrographs were used in two-dimensional hydraulic simulations using the Hydrologic Engineering Center’s River Analysis System (HEC-RAS) to evaluate water surface elevations, flow velocities, and Froude numbers at each bridge site. The findings reveal that all bridges can convey peak flows without overtopping under current climatic conditions. However, Bridges 3 to 5 experience near-critical to supercritical flow conditions, with velocities ranging from 3.43 to 4.75 m/s and Froude numbers between 0.92 and 1.04, indicating high vulnerability to local scour. Bridge 2 shows moderate risk, while Bridge 1 faces the least hydraulic stress. The applied modeling framework effectively identifies structures requiring priority intervention and demonstrates a practical methodology for assessing flood risk in ungauged, data-scarce, and semi-arid regions. Full article

In this study, landslide risk assessment was conducted in the Renhe District, Panzhihua City, China. Firstly, based on 190 landslide points and 10 influencing factors, the landslide hazard was assessed using three models: random forest (RF), eXtreme Gradient Boosting (XGBoost), and Tabular Prior-data Fitted Network (TabPFN). The results indicate that the RF and XGBoost models exhibit comparable performance, both demonstrating strong generalization and accuracy, with the RF model achieving superior generalization, as evidenced by an area-under-the-curve (AUC) value of 0.9471. While the AUC value of TabPFN is 0.9243, indicating higher accuracy, it also poses a risk of overfitting and is therefore more suitable for applications involving small sample sizes and the need for rapid responses. The vulnerability assessment utilized the Analytic Hierarchy Process (AHP) to determine the weights of four disaster-bearing bodies, with sensitivity analysis revealing that road type was the most sensitive vulnerability factor. Finally, the landslide risk-assessment map of the Renhe District was produced by integrating the landslide hazard assessment map with the vulnerability assessment map. The findings indicate that the high-risk zones comprised 2.08% of the research region, which includes three principal train stations and necessitates enhanced protective measures. The medium-risk zones comprise 34.23% of the total area and are scattered throughout the region. It is important to enhance local capabilities for landslide monitoring and early warning systems. Relevant conclusions can provide a significant reference for landslide disaster prevention and mitigation work in the Renhe District and help ensure the safe operation of public transport infrastructure, such as railway stations and airports in the district. Full article

The problem of inadequate object detection accuracy in complex remote sensing scenarios has been identified as a primary concern. Traditional YOLO-series algorithms encounter challenges such as poor robustness in small object detection and significant interference from complex backgrounds. In this paper, a multi-scale feature fusion framework based on an improved version of YOLOv8_L is proposed. The combination of a graph attention network (GAT) and Dilated Encoder network significantly improves the algorithm detection and recognition performance for space remote sensing objects. It mainly includes abandoning the original Feature Pyramid Network (FPN) structure, proposing an adaptive fusion strategy based on multi-level features of backbone network, enhancing the expression ability of multi-scale objects through upsampling and feature stacking, and reconstructing the FPN. The local features extracted by convolutional neural networks are mapped to graph-structured data, and the nodal attention mechanism of GAT is used to capture the global topological association of space objects, which makes up for the deficiency of the convolutional operation in weight allocation and realizes GAT integration. The Dilated Encoder network is introduced to cover different-scale targets by differentiating receptive fields, and the feature weight allocation is optimized by combining it with a Convolutional Block Attention Module (CBAM). According to the characteristics of space missions, an annotated dataset containing 8000 satellite and space station images is constructed, covering a variety of lighting, attitude and scale scenes, and providing benchmark support for model training and verification. Experimental results on the space object dataset reveal that the enhanced algorithm achieves a mean average precision (mAP) of 97.2%, representing a 2.1% improvement over the original YOLOv8_L. Comparative experiments with six other models demonstrate that the proposed algorithm outperforms its counterparts. Ablation studies further validate the synergistic effect between the graph attention network (GAT) and the Dilated Encoder. The results indicate that the model maintains a high detection accuracy under challenging conditions, including strong light interference, multi-scale variations, and low-light environments. Full article

Over the past four decades, Wadi Fatimah in western Saudi Arabia has undergone significant environmental changes that have contributed to desertification. High-resolution spatial and temporal analyses are essential for monitoring the extent of desertification and understanding its driving factors. This study aimed to assess the spatial distribution of desertification in Wadi Fatimah using satellite and climate data. Landsat imagery from 1984 to 2022 was employed to derive land surface temperature (LST) and assess vegetation trends using the Normalized Difference Vegetation Index (NDVI). Climate variables, including precipitation and evapotranspiration (ET), were sourced from the gridded TerraClimate dataset (1980–2022). LST estimates were validated using MOD11A2 products (2001–2022), while TerraClimate precipitation data were evaluated against observations from four local rain gauge stations: Wadi Muharam, Al-Seal Al-Kabeer, Makkah, and Baharah Al-Jadeedah. A Desertification Index (DI) was developed based on four variables: NDVI, LST, precipitation, and ET. Five regression models—ridge, lasso, elastic net, polynomial regression (degree 2), and random forest regression—were applied to evaluate the predictive capacity of these variables in explaining desertification dynamics. Among these, Random Forest and Polynomial Regression demonstrated superior predictive performance. The classification accuracy of the desertification map showed high overall accuracy and a strong Kappa coefficient. Results revealed extensive land degradation in the central and lower sub-basins of Wadi Fatimah, driven by both climatic stressors and anthropogenic pressures. LST exhibited a clear upward trend between 1984 and 2022, especially in the lower sub-basin. Precipitation and ET analysis confirmed the region’s arid climate, characterized by limited rainfall and high ET, which exacerbate vegetation stress and soil moisture deficits. Validation of LST with MOD11A2 data showed reasonable agreement, with RMSE values ranging from 2 °C to 6 °C and strong correlation coefficients across most years. Precipitation validation revealed low correlation at Al-Seal Al-Kabeer, moderate at Baharah Al-Jadeedah, and high correlations at Wadi Muharam and Makkah stations. These results highlight the importance of developing robust validation methods for gridded climate datasets, especially in data-sparse regions. Promoting sustainable land management and implementing targeted interventions are vital to mitigating desertification and preserving the environmental integrity of Wadi Fatimah. Full article

Aerosols play a crucial role in air quality, climate regulation, and public health; their timely monitoring is hence fundamental. The aerosol optical depth (AOD) is the parameter used to investigate the spatial–temporal distribution of aerosols from space. Specifically, the AOD retrieved from the Multi-Angle Implementation of Atmospheric Correction (MAIAC) algorithm applied to a Moderate Resolution Imaging Spectroradiometer (MODIS) is suitable for aerosol investigation at a local scale by exploiting its high spatial resolution (1 km × 1 km). In this study, the MAIAC AOD retrieval over Rome (Italy) was validated for the first time, using ground-based data provided by an AERONET station operating in a semi-rural environment close to the city, over a time series from January 2001 to December 2022. Moreover, AOD trends were evaluated in a study area encompassing Rome and its surroundings, characterized by a transition zone between urban and rural environments. The results show a general underestimation of the MAIAC AOD; specifically, the validation process highlighted the less accurate performance of the algorithm under higher aerosol loading and with predominantly coarse mode aerosol. Interesting results were obtained concerning the influence of the geometrical configuration of satellite acquisition on the accuracy of the MAIAC product. In particular, the solar zenith angle, the relative azimuth and the scattering angle between the principal plane of the sun and satellite synergistically influence retrievals. Finally, the spatial distribution of the AOD shows a decreasing trend over the 2001–2022 period and a strong influence of the city of Rome over the whole study area. Full article

Today, Moon exploration is driven by the desire to expand the human presence beyond Earth and to use its resources. This requires the development of reliable navigation systems that can provide positioning information accurately and continuously on the lunar surface and orbits. Initiatives such as Moonlight (by ESA) and the Cislunar Autonomous Positioning System project (by NASA) are underway to address this challenge. The aim is to use ranging signals transmitted by satellites, similar to Earth’s GNSS, for lunar user positioning. This paper proposes a solution that involves local sensors deployed on the Moon surface to enhance the performance of the satellite system. These sensors can serve as differential reference stations, correcting satellite pseudorange measurements obtained by lunar surface receivers. The local sensor can also be used as a pseudolite, transmitting satellite-like signals to improve system availability and accuracy in obstructed areas. Additionally, the local sensor can act as an independent beacon that provides range and angle measurements. Higher navigation performance can be achieved by increasing the complexity of the system, depending on the implemented solution. This paper proposes and shows the concept, the intial design, and a preliminary definition of the protocol for the third solution. The three different solutions are compared in terms of position accuracy by exploiting the Cramér–Rao Lower-Bound formulation and Monte Carlo simulations. Finally, possible implementations for future use on the Moon are discussed. Full article

With an increase in the intensity and frequency of extreme rainfall events, there is a pressing need for accurate rainfall nowcasting applications. In recent years, precipitable water vapor (PWV) data obtained from GNSS observations have been widely used in rainfall prediction. Unlike previous studies mainly focusing on rainfall occurrences, this study proposes a transformer-based model for hourly rainfall prediction, integrating the GNSS PWV and ERA5 meteorological data. The proposed model employs the ProbSparse self-attention to efficiently capture long-range dependencies in time series data, crucial for correlating historical PWV variations with rainfall events. Additionally, the adoption of the DILATE loss function better captures the structural and timing aspects of rainfall prediction. Furthermore, traditional rainfall prediction models are typically trained on datasets specific to one region, which limits their generalization ability due to regional meteorological differences and the scarcity of data in certain areas. Therefore, we adopt a pre-training and fine-tuning strategy using global datasets to mitigate data scarcity in newly deployed GNSS stations, enhancing model adaptability to local conditions. The evaluation results demonstrate satisfactory performance over other methods, with the fine-tuned model achieving an MSE = 3.954, DTW = 0.232, and TDI = 0.101. This approach shows great potential for real-time rainfall nowcasting in a local area, especially with limited data. Full article