Search Results (13,861)

The Portuguese agricultural sector has suffered a profound transformation over recent decades, evolving from traditional to increasingly technology-driven systems. Throughout this transition, climate and meteorological conditions have remained key drivers of agricultural productivity. Today, Portuguese agriculture faces growing challenges associated with climate change, including more frequent and intense heatwaves, droughts, and floods. Consequently, reliable climate information and decision-support tools are essential for strengthening resilience and promoting sustainable management. To address these needs, the Portuguese Institute for the Sea and Atmosphere (IPMA) developed two complementary climate service platforms for mainland Portugal: AgroClima and DataClima. The first provides observations from IPMA’s meteorological network, ECMWF forecasts, and agroclimatic indicators such as temperature, precipitation, soil water, and so-called agroclimatic warnings. The second offers historical climate information including WRFv4.2 simulations dynamically downscaled from ERA5 (1981–present), in situ observations (1941–present), and climate normals. Evaluation of the WRFv4.2 regionalization against IPMA observations shows a systematic underestimation of precipitation and air temperature, while mean wind speed is generally overestimated. Despite these biases, the downscaled WRFv4.2 dataset demonstrates sufficient accuracy to support operational climate services, providing valuable help for environmental monitoring, climate adaptation, and decision-making in agriculture and water resource management across Portugal. Full article

Dockless bike-sharing systems are increasingly important for sustainable urban mobility, yet they frequently suffer from spatial misallocation between supply and demand. Conventional demand forecasting relies primarily on historical trip records and therefore systematically underestimates true demand when users abandon bike-finding attempts. To address this limitation, we propose a user-behaviour-based dynamic clustering optimisation algorithm that integrates observed riding behaviour with latent unmet demand through the concept of Golden Distance—a district-adaptive service-radius threshold estimated heuristically from operational logs of successful rides and inferred unmet-demand events, serving as a behaviourally motivated proxy for aggregate walking tolerance. Building on a prior AIoT-enabled demand-prediction framework, the method first applies HDBSCAN density-based clustering to discover intrinsic demand topology, then selectively refines only those clusters that violate Golden Distance coverage constraints via an elongation-aware adaptive k-medoids formulation. District-level true demand is predicted using an XGBoost regression model and subsequently downscaled to the cluster level based on historical activity shares. Experiments on one full year of operational data from three Hong Kong districts (Tseung Kwan O, Sha Tin, and Tuen Mun) show that the proposed pipeline produces Golden-Distance-compliant service zones at a substantially finer operational resolution than the DBSCAN baseline: 86.5% to 91.6% of clustered demand points lie within the Golden Distance district of their assigned medoid, at cluster coverages of 68.0% to 78.5%, against 60.0% to 71.1% at coverages of 49.7% to 88.1% for the baseline. A resolution-fair evaluation shows that the cluster-level RMSE advantage reported previously largely reflects the finer reporting unit rather than better prediction: the same fixed district-level forecast spans a five- to seven-fold range of RMSE when scored on progressively coarser spatial units, and at a matched clustering resolution the DBSCAN baseline equals or exceeds the proposed pipeline on cluster-level metrics. Accuracy is therefore evaluated with scale-free metrics, on which the proposed method is not the more accurate of the two, and the contribution of this work is positioned on operationally deployable, behaviourally constrained spatial zoning rather than on per-cluster forecasting accuracy. Full article

The reliable monthly forecasting of fine particulate matter (PM2.5) requires artificial intelligence (AI) models that are accurate, temporally valid, and transparent. We develop a history-only rolling-decomposition framework with lightweight Mamba-inspired selective state-space backbones for 475 city-level administrative units in China. Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and three alternative decomposition strategies use only the PM2.5 data available before each decomposition cutoff, thereby avoiding future-information leakage and yielding inspectable multi-scale predictors. In the primary seven-model benchmark, CEEMDAN-DenseMamba achieved the lowest mean root mean squared error and mean absolute error (6.931 and 4.833 μg m⁻³, respectively). Equal-optimization reruns, paired moving-block bootstrap intervals, component-count sensitivity, city-wise diagnostics, and a parameter-matched gated recurrent unit baseline were then used to examine performance attribution. Under common optimization, the dense-connection contrasts showed paired error reductions with confidence intervals below zero, whereas the incremental CEEMDAN effect within a fixed backbone was smaller and its paired confidence intervals crossed zero. The recurrent baseline remained competitive. These findings support transparent, temporally valid multi-scale forecasting while limiting inference to future-month prediction for the known cities and the present experimental setting. Full article

Proactive lane-change risk assessment requires estimating whether an intended maneuver will lead to unsafe interactions before the maneuver is completed. This is challenging in naturalistic driving data because actual crashes are extremely rare, and binary collision labels provide little discriminative information for learning risk. We propose IntentDiff, an intent-conditioned diffusion framework for proactive lane-change risk assessment. The framework uses predicted future trajectories as the basis for risk estimation. A vectorized scene context learning module combines a VectorNet backbone with a Vector Quantized Variational Autoencoder (VQ-VAE) to map agent–map interactions into discrete intent codes. These codes organize complex traffic situations into interpretable intent prototypes and provide semantic guidance for trajectory generation. Conditioned on the learned intent code, a diffusion model generates kinematically consistent multimodal trajectories of the target vehicle. On the forecast trajectories, Monte Carlo rear-end risk is evaluated against the four bounding vehicles and fused into a Lane-Change Risk Index (LCRI). On the highD dataset, the framework attains an average displacement error of 0.42 m over a 5-s horizon. The forecast-based LCRI agrees closely with the index computed from realized future trajectories, indicating that most high-risk lane changes can be identified before the maneuver is completed. Grouping LCRI by intent code further reveals systematic variation in risk across lane-change maneuvers, suggesting that the learned codebook captures risk-relevant interaction patterns in addition to maneuver semantics. Full article

The extended Kalman filter (EKF) remains one of the most widely used tools for state estimation, tracking, forecasting, and data assimilation in nonlinear stochastic dynamical systems. This paper does not propose a replacement for the EKF or a modification of the filter itself. Instead, it investigates how the classical EKF may fail when used as a default estimation tool in a unified but practically interpretable nonlinear tracking problem. A stochastic moving-target observation model with angular and range measurements from two identical independent radar channels co-located at the origin of the coordinate system is used as the test environment. To the standard EKF scheme, we add only the technique of linear pseudomeasurements: the EKF is kept in its classical recursive form, and only the observation representation is changed, while the filtering algorithm itself remains unchanged. Within this framework, several systematic model modifications are considered: inaccurate state initialization, absence of prior information about the mean motion parameter, jump-like changes of motion parameters, and incorrect specification of observation-noise characteristics. The experiments show that EKF instability is not limited to explicit divergence. It may also appear as hidden degradation of estimation quality, coordinate-selective failure, physically counterintuitive accuracy behavior, and cases in which the filter remains formally bounded but performs worse than a simple direct estimate based on current measurements. In several experiments, unstable behavior becomes visible only when the Monte Carlo sample size is increased. The results provide a classification of qualitatively different EKF failure modes and support practical diagnostic criteria for testing EKF applicability in nonlinear observation models. Full article

Increasing variability in electricity production and demand contributes to greater electricity price volatility across different markets. The effective implementation of business processes related to broadly defined electricity trading therefore requires reliable electricity price forecasts. Consequently, electricity price forecasting has grown in both importance and research interest. This article presents an in-depth literature review of short-term point electricity price forecasting at the native temporal resolutions of the reviewed markets, primarily hourly and 30 min intervals, with selected studies using 15 min and 5 min intervals. The review concerns forecasts of individual market-interval prices and does not address forecasts of daily aggregated prices. The analysis covers the input data used in forecasting models, the main forecasting approaches, modelling techniques, and error measures. Forecast quality is examined with respect to market characteristics, forecasting methods, and explanatory variables. General trends in the reported results are identified, together with descriptive relationships among selected error measures. Particular attention is given to the RMSE-to-MAE ratio, referred to in this review as the Error Dispersion Factor (EDF), which is treated solely as a descriptive summary of the relative inequality of absolute forecast-error magnitudes in a given sample. The article concludes with findings and recommendations concerning best practices in electricity price forecasting. The review differs from broader conceptual and market-specific surveys by focusing narrowly on short-term point forecasts for individual market delivery intervals and by providing a structured quantitative synthesis of studies published between January 2021 and May 2026. Full article

We apply the free energy principle to free markets by treating the Market as a random dynamical system with an attracting set, i.e., some characteristic states. This licenses a normal form for stochastic dynamics that inherits from the Helmholtz–Hodge decomposition. Equipped with this functional form—and a suitable parameterization—one can create a generative model of fluctuations in the value of assets and accompanying indicator variables. This affords the opportunity for prospective (ex ante) prediction, scenario modelling and forecasting that could, in principle, be applied to any complex dynamical system exhibiting stochastic chaos. Here, we illustrate the application to portfolio management—in the context of financial services—and use the (posterior) predictive densities over future paths to evaluate the expected free energy that underwrites active inference. In this application, active inference reduces to risk-sensitive control, which can be used to model the optimal decision-making of an agent or investor. In this setting, an investor is characterized by their prior preferences for a high rate of return under drawdown constraints. Using numerical studies and historical financial data, we quantify the improvement in portfolio management, relative to baseline policies. Full article

An atmosphere–ocean–wave-coupled model, the UWIN-CM, running in real-time in the Hong Kong Observatory, has adopted radar data assimilation (DA) and been incorporated with a HAILCAST module in WRF for hail size simulation. This paper analyzes the distribution of maximum hail sizes forecast by the UWIN-CM in comparison with the Maximum Estimated Size of Hail (MESH) from radar for six hail cases reported by human or derived from radar in the Pearl River Estuary of Guangdong Province in 2026. The performance of hail forecast with and without radar DA is compared. The model with DA performed simulates hail at the locations that were consistent with the reported data and/or MESHs in five out of six cases. The timings of simulated hail aligned with reports/MESHs in three cases, and these were delayed by approximately 1 to 3 h in the remaining two. For the average hail diameters, those from simulations were smaller than MESH by around 30 to 50% in two of the five successful cases, while the other three cases had larger diameters than MESH by about 90 to 140%. Without radar DA, hails were simulated in only two out of six cases. The rate of detection was enhanced with radar DA, potentially benefiting from the improved moisture distribution and wind field of the initial conditions. The possible factors behind the differences in hail size between the simulations and MESH are discussed, including the choice of microphysics scheme and the differences in climatology between the PRE, and the contiguous US where the HAILCAST parameters were calibrated. For practical use in operation, parameter tuning of HAILCAST based on the hail climatology of the PRE would be needed. This work paves the way for providing forecasters with early alerts about the occurrence of hail in the region. Full article

Coastal humid-heat and photochemical-oxidant episodes are commonly studied through concentration estimation, leaving it unclear whether satellite observations improve warning decisions under explicit false-alarm constraints. This study introduces CoAST-EWS Japan, a six-station, validation-locked hindcast benchmark across Osaka Bay and Tokyo Bay. Models were developed using June–July 2023 data, calibrated and thresholded on August 2023 predictions, and retrospectively evaluated on June–August 2025 station-hour observations. The strong non-satellite route combines recent ground history, ERA5 meteorology, CAMS composition, and static station geometry. Adding previous-day MODIS thermal context to an otherwise identical XGBoost route increased average precision from 0.3153 to 0.3429, reduced the Brier score from 0.05032 to 0.04874, and improved recall/F1 under a validation-locked budget of 0.5 false alarms per station-day (FPDs) from 0.1864/0.2511 to 0.2402/0.3042. Japan-local calendar-day intervals supported the improvements in Brier score, recall, and F1. In a dimension-matched comparison using the same 18 MODIS variables, previous-day context increased average precision over the same-day route by 0.0378 (95% CI: 0.0144–0.0603), demonstrating that the timing advantage was not attributable to a larger satellite feature set. The MODIS increment was strongest during high-heat issue times and in Osaka Bay, and its ranking value was reproduced by a 36 h Temporal FLOW model. Matched spatial controls identified distance-based coastal context as the most stable 24 h graph component, while wind-aligned information operated as a complementary route. These results establish latency-aware MODIS thermal context as a measurable decision input for neighborhood-scale coastal compound warning. Strict station-level localization, cross-bay transfer, and forecast-consistent deployment define the next validation frontier. Full article

With the widespread adoption of cloud computing technology, modern cloud platforms have become increasingly complex and dynamic, posing significant challenges for efficient resource management. Accurate forecasting of cloud resource loads has therefore become essential for improving service quality, optimizing resource utilization, and reducing operational costs. To address the intrinsic characteristics of cloud load time series, including nonlinear fluctuations, multi-scale temporal dependencies, and redundant high-dimensional features, this paper proposes the MST-iTransformer model, which integrates multi-scale temporal encoding, sparse attention, and adaptive feature selection mechanisms. Specifically, a multi-scale temporal encoding module is developed to capture and fuse temporal dependencies across multiple periodic scales. Furthermore, an adaptive feature selection module is introduced to dynamically assign importance weights to resource features, enhancing informative variables while suppressing redundant ones. Meanwhile, a sparse attention mechanism is incorporated to reduce computational overhead while maintaining forecasting accuracy. The proposed model is evaluated on the Alibaba Cluster Trace dataset. Experimental results demonstrate that MST-iTransformer achieves MSE, RMSE, and MAE values of 0.5559, 0.7456, and 0.4968, respectively. Compared with the original iTransformer, the proposed model achieves simultaneous reductions in prediction errors and inference latency, validating the effectiveness of the multi-scale temporal encoding, sparse attention mechanism, and adaptive feature selection modules in improving forecasting accuracy and computational efficiency. These improvements provide reliable prediction support for resource scheduling and elastic scaling in cloud data centers. Full article

To address the reduced accuracy of photovoltaic power forecasting caused by the fluctuating characteristics of ultra-short-term photovoltaic output, this paper proposes an ultra-short-term photovoltaic power forecasting method that integrates static and dynamic features extracted from ground-based cloud images with deep learning. The proposed method employs distortion correction, histogram equalization, and other techniques for image quality enhancement. Static cloud-image features are extracted using a threshold segmentation algorithm based on the maximum inter-class variance method. To characterize the dynamic evolution of cloud clusters, an optical flow method is introduced to accurately capture their motion speed and direction. The extracted static and dynamic cloud features are then combined to form a fused dataset. Finally, an ultra-short-term forecasting model combining a convolutional neural network and Autoformer is developed. Comparative results under different weather conditions show that the incorporation of multiple cloud-image features significantly improves forecasting accuracy, thereby validating the effectiveness of multimodal data fusion and the proposed deep learning architecture. Full article

The growing adoption of renewable energy and small-scale power producers has increased the need for reliable and transparent peer-to-peer (P2P) energy trading. Traditional centralized markets often struggle with high transaction fees, limited transparency, and a greater risk of manipulation, which restrict efficient energy distribution. To overcome these issues, this study presents a decentralized P2P trading framework that implements a fully functional blockchain-based trading system with smart grid simulation and demonstrates a prototype machine learning forecasting module (Random Forest, 84% accuracy) designed for future integration. The trading mechanism is developed using Ethereum smart contracts and a custom ERC-20 token, the TUM Energy Coin (TEC), enabling secure and traceable energy exchange. System security is strengthened through dual confirmation steps, role-based access control, and consensus-driven market clearing. A double-sided auction model is used to match buyers and sellers fairly. Real-time grid behavior such as fluctuating loads, prosumer generation, and consumer demand is modeled using MATLAB Simulink to reflect realistic operating conditions. To enhance decision-making, a Random Forest model is integrated for load forecasting and dynamic pricing, achieving an accuracy of 84%. The simulation results show improved transaction throughput, more stable pricing, and strong resilience against false-data injection attacks. The primary novelty of this work lies in (1) an entirely operational and validated blockchain-trading system simulation with synchronized time using Simulink, (2) a working Random Forest forecasting tool demonstrating feasibility for incorporation in the future, and (3) an analysis of the system’s robustness in the case of FDIA attacks. The authors point out that the ML component used is a prototype and not yet integrated into the functioning block chain. Full article

Modal shift in urban transportation is forecast to increase the frequency of tram interactions with Vulnerable Road Users (VRUs), who are disproportionately susceptible to injury. This report describes the outcomes of a research programme to identify the limitations of existing tram–VRU crash-compatibility guidelines. A targeted search of Scopus, Google Scholar and the grey literature resources yielded 27 results for inclusion in this study, covering three areas affecting tram–VRU collisions and interactions. These areas were: body regions susceptible to injury, speed and location of reported collisions, and VRU rider preferences around roads and tram infrastructures. This informed recommendations on future tram–VRU crash-compatibility guidelines. Injury criteria correlating to head, thorax and lower extremities should be considered, to enable up to ~78–84% of serious (AIS3+) injuries from real-world tram–pedestrian collisions to be captured in crash-compatibility studies. Although most tram–VRU collisions occur around stops or crossings, increasing vehicle impact speed in testing from 20 km/h to 30 km/h would enable most high-severity injuries from collision data to be captured. Based on VRU-stated or observed preferences, cyclists and e-scooter users should be considered travelling perpendicular to road infrastructures in tram–VRU collisions. Implementation of such recommendations into tram–VRU crash-compatibility guidance would account for a greater proportion of real-world interactions. Full article

Accurate short-term wind-power forecasting supports renewable-energy integration. This study presents a reproducibility and implementation audit of MoRAM, an integrated architecture that combines feature-token attention, a residual Mamba layer, dense two-expert blending, and a one-dimensional convolution (Conv1D)–linear residual output path to forecast 12 h of normalized turbine power from 48 h multivariate histories in Dataset A (200 turbines; 8760 hourly records). Source reconstruction found that the historical PyTorch attention layer used its default sequence-major interface on batch-major data, thereby mixing samples; the dense-window record is therefore retained only as an implementation audit. We corrected the layer to batch-first feature-token semantics and ran an attention-only ablation using all turbines, a 12 h window-start stride, and five paired seeds. Corrected MoRAM achieved mean absolute error (MAE) 0.2428 ± 0.0029 per unit (p.u.) and root-mean-square error (RMSE) 0.3035 ± 0.0059 p.u.; bypassing only attention achieved MAE 0.2298 ± 0.0027 p.u. and RMSE 0.2839 ± 0.0028 p.u. (mean ± sample standard deviation). The attention-free condition had lower overall MAE and RMSE in every seed and, after averaging across seeds, at every forecast horizon. The principal validated contribution is therefore an auditable reconstruction—complete data protocol, source/log mapping, corrected tensor semantics, and a reproducible five-seed negative ablation—rather than evidence that every constituent module or the integrated architecture is superior. Full article

The histopathology area is being redefined with the use of artificial intelligence (AI), providing robust tools to help bridge cellular morphology, functional assays, and multi-omics data in stem cell research. The ability of stem cells to undergo self-renewal and differentiation is key in regenerative medicine, but their research requires the careful characterization of morphological and molecular phenotypic traits. Traditional histopathology is invaluable, but its application can be limited by inter-observer variability and restricted scalability. These limitations are circumvented by AI-based techniques, such as machine learning and deep learning, which are capable of classifying cells, performing quantitative morphometry, and forecasting stem cell behavior. Adding AI to genomics, proteomics, and metabolomics will contribute to the further identification of biomarkers and pathways that regulate stem cell fate. This convergence provides new possibilities for precision medicine, personalized therapies, and translational uses like drug discovery and disease modeling. However, its potential has not yet been realized because of the existing difficulties in data quality, variability, regulatory control, and ethical issues, especially in terms of the transparency and justice of AI systems. Emphasized areas for the future include explainable AI, federated learning, and a multimodal framework that integrates imaging, sequencing, and clinical data. Interdisciplinary partnerships and adequate regulatory frameworks will help AI-enabled histopathology reshape stem cell studies and speed up the process of translating regenerative medicine into clinical applications. Full article