Search Results (245)

Against the background of intensifying global climate change, extreme precipitation events have become increasingly frequent. Improving the accuracy of short-term precipitation nowcasting is therefore essential for disaster prevention and mitigation. Traditional numerical weather prediction (NWP) approaches are constrained by computational latency and errors arising from physical parameterizations, making it difficult to satisfy real-time forecasting requirements at high spatiotemporal resolution. Using the SEVIR dataset, this study conducts a systematic comparison of two Transformer-based deep learning models—Earthformer and LLMDiff—for short-term extreme precipitation nowcasting. Model performance is evaluated using the Critical Success Index (CSI), Probability of Detection (POD), and Success Ratio (SUCR). Results indicate that, for 0–30 min lead times, Earthformer more efficiently captures both local and long-range spatiotemporal dependencies via its Cuboid Attention mechanism and shows a slight advantage for low-intensity precipitation. As the lead time extends to 60 min, LLMDiff demonstrates stronger longer-horizon skill due to its diffusion-based probabilistic modeling and a frozen large language model (LLM) module, which enhance the representation of uncertainty and longer-term evolution of precipitation systems. However, LLMDiff tends to produce a higher false-alarm rate. Overall, Earthformer is better suited for rapid early warning of light precipitation, whereas LLMDiff is more appropriate for high-accuracy nowcasting of heavy precipitation, offering useful insights for intelligent forecasting of extreme weather. Full article

Landslides, often triggered by intense or prolonged rainfall, pose significant risks to communities and infrastructure. Identifying accurate rainfall thresholds is crucial for predicting landslide events and developing effective early warning systems. This study, conducted on São Miguel Island (Azores), aimed to improve the predictive capability of rainfall thresholds by integrating both rainfall preparatory and rainfall trigger thresholds. Using data from 61 landslide events and rainfall measurements recorded at four stations between 1977 and 2020, the study applied the Generalised Extreme Value (GEV) distribution with Maximum Likelihood Estimation (MLE) to identify the cumulative rainfall–duration pair with the highest return period for each event, thereby establishing a preparatory threshold. The trigger threshold was determined by analysing the rainfall amount recorded on the day of the event while also accounting for the duration of the preparatory rainfall period. The final threshold combines both the preparatory and trigger thresholds, and an event is detected when both thresholds are exceeded. Preparatory thresholds showed similar patterns across the stations, with Sete Cidades and Furnas recording the highest cumulative rainfall values, while Santana and Ponta Delgada exhibited lower thresholds. The trigger thresholds at Furnas reflected the highest daily rainfall intensities. The analysis also indicated that the rainfall intensity required to trigger landslides decreases with increasing durations of the antecedent rainfall. Performance of the thresholds using ROC metrics revealed that the combined threshold outperformed the preparatory threshold alone by reducing false positives (FPs) and improving predictive accuracy. In all cases, the combined threshold demonstrated superior performance in detecting landslide events, highlighting its effectiveness in landslide prediction. This study provides a detailed analysis of rainfall thresholds for landslides on São Miguel Island and underscores the advantages of the combined threshold approach for improving landslide prediction and supporting the development of robust early warning systems. Full article

Two key challenges in financial distress prediction are pronounced class imbalance between majority and minority classes and the persistent misclassification of hard-to-learn samples. To tackle these issues, this study proposes an ensemble framework called Dynamic Weight-Adjusted Random Forest Boost (DWARFB). The proposed method incorporates a dynamic sample-weighting mechanism that leverages cumulative misclassification information, adaptive minority–majority class ratios to address class imbalance issue, and a real-time performance-driven strategy to integrate models’ prediction results. The effectiveness of DWARFB is evaluated using a financial dataset from the China Stock Market & Accounting Research (CSMAR) database. Comparative experiments against eight benchmark Random Forest (RF) approaches show that DWARFB delivers superior balanced performance, and a stable precision–recall trade-off, which effectively reduces both false negatives and false positives in the prediction. Moreover, a loss-based feature contribution metric provides economically meaningful insights into the key financial determinants of distress, enhancing model interpretability. Overall, DWARFB demonstrates strong reliability and adaptability and offers a practical solution for early financial distress warning in imbalanced and dynamic financial environments. Full article

Wireless Body Area Networks (WBANs) enable real-time health monitoring essential for timely clinical intervention, yet their performance is frequently hindered by sensor degradation, noise interference, and strict low-latency constraints in resource-limited environments. Conventional preprocessing approaches indiscriminately reprocess all incoming data, including uncorrupted samples, thereby increasing computational overhead, introducing latency, and potentially distorting valid physiological trends. This study introduces a unified real-time monitoring framework tailored for WBAN systems. The key contributions include: (1) an adaptively gated multi-stage preprocessing pipeline that selectively restores corrupted samples while preserving clean data, (2) an overlap-aware sliding-window mechanism enabling low-latency operation, and (3) a clinically informed risk assessment strategy for early-warning support. By avoiding unnecessary modification of intact signals, the framework maintains physiological integrity while substantially improving reconstruction and predictive reliability. Across multiple vital signs, the proposed approach achieves substantial reconstruction gains, with Mean Squared Error (MSE) reductions ranging from 53% to 67% under strong degradation conditions. An adaptive ARIMA-based forecasting layer captures short-term physiological dynamics with directional accuracies of approximately 65–70% for one-step (10 s) ahead prediction. Early-warning behavior is intentionally conservative, prioritizing false alarm suppression over aggressive alerting. Per-signal evaluation reveals high sensitivity for blood pressure signals, whereas glucose and certain high-variability modalities exhibit conservative sensitivity under modality-specific thresholds. Importantly, the aggregated multi-modal risk decision achieves strong overall system-level performance, with sensitivity and specificity of 0.89 and 0.92, respectively. Overall, the proposed framework establishes a robust, low-latency, and computationally efficient foundation for dependable physiological monitoring in WBAN environments, leveraging selective processing to optimize both resource utilization and clinical reliability. Full article

Rip currents pose a serious maritime safety hazard, as they can quickly carry swimmers away from the shore, often leading to drownings caused by panic. Traditional beach flags and signs often fall short due to the complexities involved in issuing real-time warnings. In this study, a framework for rip current warning based on deep learning was introduced and evaluated. The framework consists of automated object detection, adaptive time-averaged image generation, and expert validation protocols. The YOLOv4 deep learning model was trained and evaluated using three distinct datasets derived from two primary sources: a publicly available dataset sourced from peer-reviewed literature and a custom-built dataset compiled for this study. The results indicate that the models performed effectively, even under challenging environmental conditions, such as fluctuating lighting, camera motion, and varying wave dynamics. A significant novelty of this framework is the adaptable time-averaging feature, which filters out potential false positives generated by the deep learning model. This feature also allows for rapid detection in emergency situations while identifying persistent rip channel patterns for long-term risk assessments. Furthermore, the rip current alerts are not solely activated by automated results. Rather, they are contingent on the verification of dangerous conditions by trained personnel, such as lifeguards or beach management officers. The results of implementing a pilot version of this framework demonstrate its practical viability for real-world deployment, marking a significant advancement in transitioning deep learning-based rip current detection from controlled environments to practical, real-time warning systems. Full article

Spaceborne infrared small-target detection plays a critical role in space-sky early warning, disaster rescue, and reconnaissance tracking, benefiting from all-time, all-weather, and wide-area monitoring capabilities. The deployment of high-spatial-resolution infrared payloads (ground sampling distance, GSD < 10 m) has introduced pronounced scale diversity among targets, leading to size-sensitive performance degradation in existing detectors and heightened risks of missed detections or false alarms in mixed-size scenarios. Furthermore, multi-frame infrared small-target detection methods often face challenges in maintaining consistent temporal coherence during feature propagation across sequences. To overcome these limitations in high-resolution spaceborne infrared videos, we propose DTRFR, an end-to-end unified detection framework built on an enhanced recurrent feature refinement architecture. This approach incorporates a realistic SITP-QLSD dataset derived from QLSAT-2 infrared backgrounds, featuring diverse scenes, multi-size small targets, and a dedicated generalization sub-test set with extremely small targets partially unseen in training; a multi-scale IRFeatureExtractor leveraging parallel convolutions and dilated receptive fields for improved cross-scale discrimination and clutter suppression; and an adaptive gating pyramid deformable alignment module to optimize sequence alignment and enhance temporal consistency, enabling robust performance across various clutter levels and dynamic backgrounds. Extensive evaluations on SITP-QLSD demonstrate that DTRFR attains competitive performance, achieving mIoU of 74.32% and Pd of 94.51% on the main set, with strong robustness on the generalization sub-test set (Pd = 92.37%). Compared to single-frame and multi-frame baselines, the proposed method achieves higher detection accuracy with significantly reduced false alarms, benefiting from multi-scale feature extraction that enables robust detection of small targets of different sizes in infrared videos. Full article

In protected horticultural production, tomato late blight shows strong environmental inducibility, with a short latent period, rapid risk accumulation, and a limited control window, which challenges conventional post-event disease monitoring. To address this, a tomato late blight risk perception and predictive control approach for protected production is proposed, integrating deep temporal modeling of environmental factors, visual symptom perception, and risk-driven greenhouse control to enable prospective assessment and proactive intervention. Based on disease mechanisms and real greenhouse conditions, an artificial intelligence (AI) framework covering perception, prediction, and regulation is constructed, moving beyond reliance on visible symptoms alone. Long-term evolution of key variables, including temperature, air humidity, leaf wetness, and light intensity, is modeled using deep temporal networks, while early weak lesions and subtle texture changes are captured by visual models. Cross-modal fusion in a unified risk space generates continuous risk scores to drive greenhouse regulation. Experiments on a multimodal dataset from a real greenhouse in Bayannur, Inner Mongolia, show that the proposed method outperforms vision-based and environment-based baselines in recognition and risk prediction. It achieves about 0.95 accuracy, 0.94 F1-score, and over 0.97 area under the receiver operating characteristic curve (AUC), while providing more than 20 h of early warning before disease onset. In environmental modeling, the deep temporal model consistently surpasses threshold-based methods, logistic regression, and long short-term memory/gated recurrent unit (LSTM/GRU) baselines in risk lead time, false alert rate, and prediction stability. Full article

Weather radar provides continuous, large-scale observations of aerial biological activity. However, biological echoes typically exhibit weak signals, sparse distributions, and non-stationary abrupt variations, causing existing extrapolation models to suffer from over-smoothing and loss of detail and making it difficult to capture their short-term evolution effectively. To address this issue, we propose an Integrated Self-Attention Long Short-Term Memory (ISA-LSTM) model that integrates a self-attention mechanism within the Predictive Recurrent Neural Network (PredRNN) framework. Coupled convolutional modules are introduced to enhance feature interactions between inputs and hidden states, while a spatiotemporal self-attention mechanism improves long-term dependency modeling and local detail preservation. Experiments conducted on 6000 biological echo samples from three weather radars in the Poyang Lake region demonstrate that the proposed model achieves superior extrapolation accuracy and stability compared with existing methods, maintaining a low false-alarm rate for lead times of up to 50 min. The results suggest that ISA-LSTM offers an effective deep learning approach for biological echo extrapolation, with applications in aviation safety and agricultural pest and disease early warning. Full article

Background: Clinicians rely on clinical judgement and vital sign monitoring to identify patient deterioration, commonly supported by systems such as the National Early Warning Score 2 (NEWS2). However, NEWS2 is associated with a high false-positive burden, contributing to alert fatigue in increasingly pressured clinical environments. Consequently, there is a growing need for early warning systems (EWS) that not only detect deterioration but do so with higher precision to prioritise clinically meaningful alerts. We aimed to develop and validate a prognostic EWS capable of predicting real-time clinical deterioration in hospitalised adult patients. Methods: We conducted a retrospective observational cohort study using routinely collected Electronic Patient Record (EPR) data. A Cox proportional hazards model with time-varying covariates was developed to estimate dynamic risk of deterioration. Deterioration was defined as unplanned transfer to intensive care, unplanned surgery, or in-hospital death. Data for model development comprised 37,989 adult inpatient episodes admitted between January 2022 and October 2024, and were initially split into training, temporal validation and test datasets. An extended evaluation period included 11,048 patients admitted through September 2025. Model performance was compared with NEWS2 at the emergency-response threshold (≥7). Results: The final model produced a tiered “traffic-light” risk profile and demonstrated substantially higher precision than NEWS2 while maintaining comparable recall in our test data. At the red alert threshold, precision was 60% compared with 16% for NEWS2 ≥7, with 82% versus 43% of alerts occurring within 24 h of deterioration. Performance remained consistent across the extended evaluation period. Conclusions: A survival-based EWS incorporating time-varying covariates achieved higher precision and improved temporal alignment with deterioration events compared with NEWS2. A tiered amber–red alert framework may support more targeted escalation, reduce alert fatigue, and enhance early identification of clinical deterioration. Full article

Rural transportation corridors remain weakly instrumented for continuous low-altitude airspace monitoring. At the same time, Vehicle-to-Everything (V2X) roadside units (RSUs) are increasingly deployed for transportation safety services. This work investigates whether existing RSUs can be extended with passive, cooperative RF sensing to detect small UAVs without modifying standards-compliant ITS communications in the protected 5.9 GHz band. A calibrated simulation study evaluates corridor-scale operation under realistic propagation conditions, including terrain masking and narrowband interference. All results reported in this paper are derived from simulation and do not include field measurements or hardware prototyping. False alarm performance under diverse ISM emitters is not quantified. The results show that cooperative processing across neighboring RSUs improves epoch-level verified detection coverage compared with single-RSU sensing. Bearing variability is reduced for weak or partially masked signals. These gains result from feature-level validation across spatially separated receivers rather than deterministic signal combining. RF calibration constrains detections to physically plausible kilometer-scale ranges. The resulting angular accuracy is sufficient for early warning and track initiation, but not for precise localization. Overall, the findings indicate that existing V2X infrastructure can support supplementary early warning capability for corridor-scale airspace monitoring while preserving primary V2X safety functions. Full article

To address critical limitations in conventional forward collision warning (FCW) systems including inadequate road condition detection accuracy, significant warning area prediction errors, and poor environmental adaptability on wet/snow-covered roads, this study develops an adaptive collision warning framework based on real-time road slippery states recognition. An enhanced ED-ResNet50 model is proposed, incorporating grouped convolutions within the backbone network and embedding ECA attention mechanisms after the second/third residual blocks alongside DDS-DA modules after the fourth block, significantly improving discriminative capability for pavement texture analysis under adverse conditions. This vision-based recognition system synchronizes with YOLOv8 for preceding vehicle detection, enabling the construction of a friction-sensitive safety distance and the time-to-collision model that dynamically calibrates warning thresholds according to instantaneous vehicle velocity and road adhesion coefficients. Real-vehicle validation demonstrates an 8.76% improvement in overall warning accuracy and 7.29% reduction in lateral and early false alarm rates compared to static-threshold systems, confirming practical efficacy for safety assurance in inclement weather. Full article

Steep slope failures adjacent to residential areas are becoming an increasingly serious hazard. However, satellite-based monitoring is often limited by revisit time and spatial resolution, which can impede the timely identification of small, precursory deformations. To support dense in situ surveillance, embedded glass fiber-reinforced polymer (GFRP) sensor rods were installed in a susceptible slope, and ground-displacement data were recorded at 5 min intervals for five months. Based on these multivariate time series, we propose PRISM-TAD, a masked Transformer-based anomaly detection approach that integrates kinematic priors computed from displacement and velocity to model normal slope dynamics and detect departures from typical behavior. The proposed method was benchmarked against six baselines: robust velocity threshold screening, PCA-based reconstruction, Isolation Forest, one-class SVM, a 1D convolutional autoencoder, and a standard Transformer reconstructor. In a field test using a documented slope failure case in Seocheon, PRISM-TAD generated an alert approximately 22 h before collapse while yielding the lowest false alarm rate. Although some baseline methods showed longer nominal lead times, they produced substantially more false positives. Overall, the results suggest that coupling high-frequency IoT displacement sensing with domain-informed deep learning can enhance the operational reliability of early warning for slope failures. Full article

With the rapid advancement of remote sensing technology, remote sensing images are increasingly being used in applications such as geographical monitoring, disaster warning, and urban planning. However, detecting small objects—such as vehicles and small buildings—in such imagery remains challenging due to complex backgrounds, weak features, and interference from factors like terrain, clouds, and lighting, leading to high rates of missed detections and false alarms. To tackle these issues, this paper proposes an improved YOLOv8-based framework for small object detection in remote sensing images. The enhancements include a multi-scale feature fusion mechanism, optimized data augmentation strategies incorporating super-resolution techniques, and a redesigned loss function that emphasizes small objects. These refinements significantly improve the model’s ability to extract discriminative features and detect small targets against cluttered backgrounds. Experimental results demonstrate superior performance across multiple metrics, including precision, recall, mAP50, and mAP50-95, particularly for challenging categories like small vehicles and buildings. This research not only provides an effective solution to the key technical bottleneck in small object detection, advancing the progress of related algorithms, but also offers important theoretical and practical experience for subsequent work. Full article

Climate change is expected to intensify droughts, thereby increasing the need for reliable predictive tools. In this study, one-month-ahead forecasts of the Palmer Z-Index were generated using long-term monthly data from two meteorological stations (17112 Çanakkale and 18084 Biga) located in the Troy region. The input features included current and lagged meteorological variables, multi-month rolling statistics, and seasonal encodings. Eight machine learning models, including linear and ensemble tree-based approaches, were evaluated using time series cross-validation. Drought events were defined based on Palmer Z-Index and standardized drought indicators, and model performance was assessed using commonly adopted accuracy and detection measures. Shapley Additive Explanations (SHAP) analysis was used to quantify the feature contributions. Gradient Boosting achieved the highest predictive accuracy at the main station, while XGBoost and CatBoost also performed strongly. High accuracy was maintained at the second station, demonstrating the spatial robustness of the model. The machine learning-predicted Palmer Z-Index values showed strong agreement with observed hydrological drought conditions; severe drought events were detected with high confidence and low false alarm rates. SHAP results identified precipitation inputs as the most dominant driver of Z-Index variability. Overall, the findings suggest that ML-based models can provide timely and interpretable forecasts for operational drought early warning systems. Nonetheless, further research is needed to test the generalizability of these findings under different climate regimes and data conditions. Full article

Glide-snow avalanches pose a major challenge for operational forecasting and local avalanche authorities. Although their key prerequisite, a moist interface between the snowpack and smooth ground, is well known, predicting the timing of glide-snow avalanches remains difficult. We analyzed five seasons of avalanche monitoring data in the Planneralm area of Styria, Austria. Glide-snow avalanche activity in the study area follows typical temporal patterns, with the highest release probability in the early afternoon and peak activity from mid-March to mid-April. Using meteorological data and avalanche observations as input, we trained machine-learning models to predict hours with glide-snow avalanche release. The most significant predictors were the mean air temperature of the preceding 48h, the day of the winter season, the hour of the day, and the decrease in snow height. The combination of those variables suggests a longer-term predisposition toward glide-snow avalanche release, as well as short-term driving factors. Our decision tree model correctly identified the vast majority of avalanche hours (recall 90%) at the cost of a moderate false alarm rate (15%). Our model could support operational glide-snow avalanche forecasting by identifying hours with elevated glide-snow potential that warrant increased attention and may require warnings or temporary closures by local authorities. Full article