ABSTRACT Anthropogenic soundscape disturbance can constrain acoustic signalling, with potential consequences for reproductive behaviour in wild animals. We used passive acoustic monitoring to investigate how habitat, weather, and human-associated soundscape disturbance are associated with roaring activity of Iberian red deer (Cervus elaphus hispanicus) during the rut on Lousã Mountain, central Portugal. Twenty-nine AudioMoth recorders collected 150-s recordings every 7.5 min over 14 days, yielding 88,022 manually validated roars. We quantified anthrophony as relative acoustic energy in the 1–2 kHz band and modelled roar counts using a negative binomial generalised linear mixed model. Roaring activity was highest in shrublands and other open habitats, and decreased with increasing wind speed, rainfall and temperature. Importantly, roaring declined with increasing anthrophony and increased with distance from wind turbines, suggesting a possible avoidance of a broader disturbance gradient rather than only a purely acoustic effect. Daily roaring counts also tended to be lower towards the end of the week. Our results highlight that soundscape disturbance associated with human infrastructure and activity may influence both the location and timing of males’ vocalisations during the rut, demonstrating the value of soundscape metrics for identifying rutting areas potentially sensitive to disturbance and informing appropriate mitigation strategies.

Extreme weather poses serious challenges to wildlife, often forcing animals to alter their behaviour with potential population-level consequences. Following contemporary climate change, the number and intensity of Arctic storms are increasing, but the responses of Arctic species to wind speed and episodic storms are poorly documented. We aimed to quantify behavioural responses (i.e. changes in movement modes and habitat selection patterns) of muskoxen (Ovibos moschatus) to wind speed and to compare estimates from before, during and after storm to hurricane-level wind events (wind speeds ≥24.5 m/s on the Beaufort scale). Hourly positions of Global Positioning System (GPS)-collared adult muskoxen (N = 61) tracked in northeast Greenland between 2013 and 2024 were georeferenced with data on wind speed, precipitation class (with or without), elevation, terrain ruggedness and vegetation type. Statistical movement modes were estimated using unsupervised hidden Markov models, and habitat selection patterns were quantified using step selection functions for summer (June-August) and winter (September-May). Results showed that wind speed influenced the allocation of time between movement modes. Increasing wind speed reduced the amount of time spent in states characterized by intermediate step lengths and turning angles, suggesting a decrease in foraging activity. Changes in state-time budgets were most pronounced under wet conditions during summer, likely due to reduced fur insulation and increased thermoregulatory behaviour. Habitat selection patterns during calm conditions (e.g. selection for dense vegetation) were reinforced at increasing wind speeds. State-time budgets of muskoxen affected and unaffected by storms were comparable between days before and after episodic storm events. During storms, however, the allocation of time among movement modes shifted more towards states indicative of resting at the expense of states indicative of foraging and relocating, which differed from muskoxen unaffected by storms. Muskoxen adopt a simple energy conservation strategy of bedding down in dense vegetation habitat to buffer against negative impacts of high wind speed, without actively compensating for potential lost foraging opportunities during episodic storm events. However, as the Arctic climate continues to change, the energetic implications of increasing wind speed and frequency of storms on individual fitness and population performance represent a valuable area for future research.

Matched-field processing is highly sensitive to environmental mismatch, yet most robust formulations emphasize static uncertainties more than time-evolving environmental forcing. This study examines a representative low-frequency shallow-water scenario in which wind-driven mixed-layer deepening reshapes the upper-ocean sound-speed profile and perturbs modal horizontal wavenumbers, producing accumulated phase errors, ambiguity-surface distortion, and systematic range bias. To organize these effects beyond a single operating point, a conditional modal phase-spread analysis is introduced to show how wind-driven degradation depends jointly on wind state, propagation range, frequency, and source depth relative to the mixed layer. A physics-coupled particle filter (PC-PF) is then proposed, in which wind speed is treated as a dynamic hidden state and estimated jointly with source range through an embedded reduced-order environmental model. Broadband numerical experiments are used to assess mechanism and tracking performances. For the representative storm-evolution scenario considered here, a conventional static-model broadband Bartlett processor develops kilometer-scale range errors, whereas the proposed PC-PF substantially reduces the root mean square error and preserves track continuity. The formulation is intended as a reduced-order, acoustically informed framework for dynamic environmental adaptation in time-varying conditions.

The study of extreme fire phenomena is limited by the experimental capabilities, especially in terms of geometric scale. Scaling tools provides a solution to extrapolate limited laboratory-scale results to large real-world-scale scenarios. In this paper, a new scaling law is proposed for merging bushfires (which propagates quasi-steadily) through a relationship between the normalised rate of spread of junction fire and Byram’s convective number. The proposed law accounts for wind and slope effects, highlighting the role played by the two forces governing the flame-front dynamics and the plume trajectory: buoyancy force and wind inertia. A large set of numerical simulations of the junction fire at a wide range of scales, slopes, wind speeds, junction angles and two types of fuel (grass and shrub) was carried out using fully physical modelling. Results show that the normalised rate of spread of a junction fire depends only on a modified expression of Byram’s convective number and on fuel type. Moreover, the proposed expression of Byram’s number yields a unified scaling law for both junction fires and single straight fire lines for quasi-steady fires. The research helps assess the effects of some topographical parameters in extreme fires, improving situational awareness, operational predictions and firefighter safety. • A modified Byram’s convective number is proposed for junction fires. • Normalised junction fire rate of spread is scaled using Byram’s convective number. • Unified scaling law is found for both junction fire and single straight fire line. • The scaling law depends on fuel type. • Fire scale, when considered in isolation, does not affect junction fire behaviour.

Dust storms originating from the desiccated bed of Lake Urmia pose a serious threat to public health, agriculture, and ecosystem sustainability. This study assessed dust storm susceptibility using four hybrid models integrating deep learning and machine learning approaches, namely CNN–RF, CNN–LR, SVM–RF, and SVM–LR, developed from 16 environmental predictors and verified dust occurrence and non-occurrence samples. The model performance was evaluated using standard validation metrics, which revealed that the CNN-based hybrid models consistently outperformed the traditional machine-learning combinations. The resulting susceptibility maps indicate extensive high-risk zones across the basin, with substantial exposure of the population, agricultural lands, and orchards. Feature analyses highlight pH soil, salinity, Normalized Difference Vegetation Index (NDVI), wind speed and sun hours as dominant drivers. The high-resolution maps delineate priority areas for mitigation and land-use planning and offer a transferable framework for dust storm risk assessment in arid and semi-arid lake basins experiencing severe desiccation.

Physics-based simulations, while central to transmission tower fragility analysis, are often computationally prohibitive for system-level or regional-scale assessments, where the objective is to evaluate the performance of multiple transmission towers with diverse geometries over large geographic regions. The study addresses these challenges by developing a generalized surrogate wind fragility modeling framework. This framework facilitates rapid and comprehensive vulnerability assessment of transmission towers by integrating structural design, fragility analysis, and deep-learning-based surrogate modeling. This integration allows the framework to explicitly account for site-specific environmental conditions and tower-specific design characteristics across diverse geographic regions in the United States, thereby enhancing its generalizability and applicability to a wide range of tower designs and locations. Moreover, contrary to traditional fragility models, the framework considers a more realistic representation of extreme wind events, where multiple hazard variables (i.e., wind speed, direction, and rainfall intensity) influence tower fragility concurrently. The resulting surrogate models provided high prediction accuracies. For unseen towers, the models achieved a mean square error of 0.0202 and an R 2 of 0.899 and demonstrated consistency with conventional physics-based fragility models. Moreover, the inclusion of location-specific design parameters enabled the models to adapt to different regional performance objectives, while considering tower-specific parameters in the design phase facilitated rapid vulnerability assessments for various tower designs within the transmission network. Overall, the proposed framework can potentially reinforce decision-making for grid resilience planning by offering a practical and computationally efficient solution for system-level vulnerability analysis of transmission tower networks. • The framework combines design and fragility processes for model generalizability. • The interaction of three environmental parameters is considered in tower fragility. • The developed surrogate models provide a high prediction accuracy. • Surrogate models can be generalized for various tower configurations and locations. • The framework provides a computationally-efficient solution for network-level analysis.

Pine wilt disease (PWD) poses a severe threat to the stability of China's ecosystems and their carbon sequestration capacity. However, the complex interactions between natural and anthropogenic spread pathways remain poorly quantified. By analyzing epidemic subcompartment records from 2014 to 2023, this study established quantitative thresholds to distinguish natural from anthropogenic spread, simulated the historical anthropogenic spread process, and identified the spatiotemporal drivers underlying both spread patterns. The research findings indicate a maximum natural spread distance of 2.725 km per year. Distances beyond this threshold were attributed to anthropogenic spread. Across the decade, anthropogenic spread averaged 36.916 km year, peaking at 91.959 km in outbreak intensive years. A geodetector analysis revealed that natural spread was primarily influenced by normalized difference vegetation index (NDVI), wind speed, and temperature, whereas anthropogenic spread was influenced by road density. By integrating road hierarchy data, wood processing plant locations and wood transportation logistics, the study simulated the anthropogenic spread pattern and mapped cross-regional long-distance spread corridors. This study provides the distance-based quantitative criterion distinguishing natural and anthropogenic PWD spread, and demonstrates that long-distance dispersal is overwhelmingly driven by human activities. The integrated probability modeling and road network framework offers an operational tool for identifying high-risk pathways and supports targeted surveillance, quarantine design, and regional management strategies for PWD and other human-vectored forest pests. © 2026 Society of Chemical Industry.

This paper presents a computational framework for predicting the heat transfer from a burning battery energy storage system (BESS) to adjacent enclosures using computational fluid dynamics (CFD) modeling representative of a large-scale fire test (LSFT) configuration. Simulations were conducted using representative input parameters informed by literature and experimental data. The predicted heat fluxes were in agreement under quiescent conditions with three empirical models. A full factorial sensitivity analysis examined the influence of heat release rate, wind speed and direction, ventilation configuration, radiative fraction, and atmospheric absorption parameters. The results demonstrate that predicted peak heat flux and total heat transfer to neighboring enclosures are sensitive to the flame geometry resulting from the interaction of the heat release rate, enclosure ventilation, and ambient winds. The radiative fraction also impacted the predicted exposure, although atmospheric absorption had less impact in the near field. This work highlights the need to evaluate a range of scenarios when conducting a hazard assessment in BESS installations. CFD models provide a critical role in filling the gap from variability in real-world installations and statistical power of individual full-scale tests. Recommendations are provided for additional measurements in LSFTs to improve model validation and reduce uncertainty in separation distance evaluations. • Explores the impact of different potential test configurations on the heat transfer. • Presents systematic analysis of physical inputs affecting heat transfer from BESS. • Discusses the model sensitivity to combustion reaction and atmospheric attenuation. • Provides a detailed sensitivity analysis of the computational domain. • Recommends measurements to collect in LSFT to maximize value for future modeling.

ABSTRACT Maritime shipping is a cornerstone of South Korea’s economy, yet emissions from port operations remain a major contributor to urban air pollution. However, few studies have quantitatively isolated the specific contribution of ship traffic to local PM2.5 levels within port-city environments, especially at fine spatial scales. The current study examines the dispersion of PM2.5 around Busan Port, with a focus on buffer zones of 3 and 10 km from the emission control area (ECA) boundary. We integrate high-resolution automatic identification system (AIS) vessel movement data with ground-based air quality measurements and meteorological observations to capture both emission sources and atmospheric transport dynamics. Two advanced gradient boosting models, LightGBM and XGBoost, were developed and evaluated using K-fold cross-validation; LightGBM achieved strong predictive performance (R² = 0.882), demonstrating the value of data-driven modeling for port air-quality assessment. The results indicate that ship-derived PM2.5 emissions are the main predictor in the 3-km zone, while wind speed is the most critical factor driving dispersion in both zones. SHAP and PDP analyses provide a transparent interpretation, confirming the relative influence of ship emissions versus meteorological controls. The novelty of this work lies in the spatially explicit integration of AIS-based emissions with meteorological variables within multiple buffer zones.

Firebrands are recognized as a major source of wildland fuel ignition and a critical driver of fire spread in wildland–urban interface (WUI). This study experimentally examines the ignition behaviour of two widely present vegetative fuel beds in the WUI, pine needle and eucalyptus, when exposed to glowing firebrands under both no wind and wind conditions. Key ignition parameters, including fuel consumption rate, rate of spread, and flame development, were evaluated. Pine needle beds consistently exhibited more intense burning behavior than eucalyptus, with higher fuel consumption rates, faster fire spread, and greater flame heights. For the same fuel load, the average peak values of mass loss rate, rate of spread, and flame height in pine needle fuel beds were approximately 4.5, 1.5, and 1.85 times greater, respectively, than in eucalyptus. Increasing fuel load resulted in increased mass loss and flame height by factors of approximately 1.7 and 1.3 times, respectively, while reducing the rate of spread to about 0.9 times. A notable flame separation phenomenon was also observed during spot fire, where the flame front detached and subsequently created two flame zones. These findings highlight the importance of fuel structure in determining ignition intensity and fire spread under firebrand exposure. • Fuel type, load, and wind govern ignition and fire propagation dynamics • Pine needle beds burn faster with higher consumption rates, and greater flames • Once spot ignition was initiated in the fuel, it leads to sustained fire propagation • Fires in fuel beds with lower loads spread faster but produce smaller flames • Spot fires cause flame separation, forming two flame zones

Accurate and robust forecast of meteorological variables such as wind speed, solar irradiance and ambient temperature is challenging because of their nonlinear, non-stationary behavior as well as season-dependent dynamics. This study proposes an adaptive least squares regression fusion-based ensemble method for multi-horizon multi-season weather forecasting by combining three individual deep learning models: a proposed multi-scale three-branch convolutional neural network with bidirectional long short-term memory, a classical long short-term memory, and convolutional neural network with bidirectional long short-term memory. The least squares regression fusion adaptively assigns the appropriate weights to each individual model according to the weather variable, season, and forecast horizon. Evaluation tests conducted on a one-year dataset for different short-term horizons reveal the superior performance of least squares regression fusion compared to all individual deep learning models in terms of accuracy and performance stability. For instance, at one-hour forecasting horizon, the outcomes show, a root mean square error reduction ranged from 3–12% for irradiance, 3–10% for wind speed and 4–10% for temperature. Additionally, a high coefficient of determination was observed, approximately equal to 0.99, implying a strong temporal correlation between the predicted and observed weather variables throughout the four seasons. Statistical analyses, including paired t-tests with false discovery rate correction, confirm that least squares regression fusion consistently outperforms individual models, achieving the highest win rates for wind speed (65.8%) and irradiance (55.8%), while remaining competitive for temperature (46.4%). Overall, the adaptive least squares regression fusion framework effectively integrates heterogeneous deep learning models, dynamically adjusting their corresponding contribution, and achieves an effective forecast of weather variables for short-term multi-horizons and across all four seasons. • Multi-season, multi-horizon short-term forecasting of three weather variables. • Adaptive LSR-based fusion of MS-3B-CNN-BiLSTM, LSTM, and CNN-BiLSTM with chronological train–test splits. • LSR fusion outperforms individual DL models across all seasons and forecast horizons. • Statistical validation using paired t-tests with false discovery rate correction confirms the superiority of LSR fusion. • Robustness and high accuracy of the model are achieved specially for temperature and irradiation.

Moon-based synthetic aperture radar (SAR) offers a promising platform for long-term monitoring with global coverage of Earth's vegetation tipping elements. However, wind-induced vegetation motion during aperture synthesis can degrade image stability and spatial resolution. To address this, we integrated multi-year wind speed records, a parameterized internal clutter motion model, and Moon-based SAR geometry to systematically analyze the dynamic scattering behavior across representative vegetation types under prevailing wind conditions. The results reveal two distinct degradation mechanisms: high-latitude sensitivity, driven by orbital geometry, and high-biomass severity, driven by vegetation structure. Crucially, we demonstrate that the unique ultra-high orbit enables Moon-based SAR to overcome these challenges, achieving sub-100 m resolution across all representative ecosystems by surpassing the traditional half-antenna length limit. Furthermore, the analysis indicates that C-band may be preferentially considered in system design, as it achieves a comparatively favorable trade-off between temporal coherence required for imaging stability and structural sensitivity across diverse vegetation conditions. These findings support the design and capability assessment of Moon-based SAR systems optimized for vegetation observation in the context of ecosystem transition monitoring.

The XGBoost wind speed prediction model based on VMD-LSTM error correction

Climate-adaptive stochastic framework for annual maximum wind speed prediction

Multi-step prediction method for short-term wind speed based on SSA-STVFEMD-Kransformer(EPSR-D-25-05656)

Abstract Ragweed ( Ambrosia artemisiifolia L.) pollen represents a major aeroallergen in urban environments across Central and Southeastern Europe, yet quantitative assessments of its seasonal dynamics and meteorological drivers remain limited for many regions. This study examines the characteristics of the ragweed pollen season and daily meteorological drivers in Zagreb, Croatia, utilizing five consecutive years of monitoring data (2020–2024). Daily airborne pollen concentrations were analysed alongside local meteorological variables using a main pollen season (MPS) definition using the 5–95% cumulative sum method. To account for the highly skewed and zero-inflated nature of pollen data, negative binomial (NB) and zero-inflated negative binomial (ZINB) regression models were applied, complemented by nonparametric correlation analysis and multivariate ordination. Ragweed pollen seasons in Zagreb were consistently confined to a narrow late-summer window, with minimal interannual variability in timing and durations of approximately 30–36 days. In the context of a consistent phenological structure, atmospheric temperature was identified as the primary meteorological factor influencing daily fluctuations in pollen levels, showing a clear positive correlation with concentrations. In contrast, relative humidity and precipitation showed negative associations during the flowering period, whereas wind speed displayed weaker and context-dependent relationships. ZINB models enabled separate modelling of structural zeros and daily pollen concentration intensity. Multivariate analyses independently confirmed a dominant thermal–humidity axis supporting a dominant thermal–humidity association underlying pollen variability. These findings demonstrate that ragweed pollen dynamics in Zagreb are governed by a stable and highly concentrated seasonal structure, within which short-term meteorological variability, particularly temperature, controls daily exposure levels. The integration of phenologically informed MPS definitions with zero-inflated modelling provides an interpretable analytical framework for understanding urban ragweed pollen dynamics and supports evidence-based interpretation of aerobiological monitoring data.

Categorical evaluation of methods for estimating aerodynamic parameters for vertical wind speed profiles over built-up areas: A systematic review

The transition to sustainable energy positions wind power as a key renewable solution. As demand grows, wind turbines are deployed across diverse terrains. However, wind’s stochastic nature and environmental variability complicate power forecasting, affecting grid stability. The study leverages data-driven techniques to enhance wind power forecasting using high-resolution SCADA system time-series data. Key operational parameters include wind speed, rotor speed, generator speed, nacelle orientation, ambient temperature and power output. A comparative analysis evaluates traditional machine learning models—Linear Regression, Decision Trees, Random Forests, Gradient Boosting and Support Vector Machines—against deep learning models like Long Short-Term Memory (LSTM) networks and a novel Recurrent Neural Network (RNN) architecture. The core contribution is an optimized Bidirectional LSTM-RNN model with permutation layers and attention. These layers capture long-range dependencies and nonlinear interactions in wind data. The structure improves long-range dependency capture and nonlinear interaction modeling. Bidirectionality enables learning from both past and future time steps, while attention mechanisms highlight critical temporal features. Experimental results demonstrate the proposed model’s superior performance, achieving a Mean Absolute Error (MAE) of 0.0994 and Root Mean Square Error (RMSE) of 0.1390, significantly outperforming traditional models (e.g., Random Forest: MAE 86.44, RMSE 220.30) and basic LSTM models (MAE 14.48, RMSE 15.27). Robust cross-validation confirms its ability to generalize across different temporal segments. Feature importance analysis improves interpretability, supporting informed decision-making in wind farm operations. The framework is scalable, modular and well-suited for real-time forecasting applications. The work presents a reliable deep learning model for wind power forecasting, enabling intelligent, data-driven energy management in modern power systems.

Ambient PM2.5 exposure and gestational diabetes mellitus: Evidence from an instrumental variable analysis in a prospective birth cohort study in China.

Air pollution remains a primary global health concern, with one of its most hazardous components, fine particulate matter PM2.5, being recognized as a significant contributor to premature mortality worldwide. Thus, forecasting PM2.5 concentrations is essential for policymaking and mitigating their harmful effects. Low-cost sensor (LCS) technology has advanced air quality monitoring by providing real-time PM2.5 measurements at finer spatial and temporal granularity due to their higher deployment density, enabling more comprehensive data for predictive modelling. However, these sensors can experience significant data gaps, necessitating effective imputation methods to improve the accuracy of PM2.5 predictions. In this study, we focus on reconstructing missing PM2.5 readings by evaluating the impact of different imputation strategies on PM2.5 forecasting accuracy using data from eight LCSs deployed across Edmonton, Alberta, from January 2023 to December 2024. The dataset includes 7 features: temperature, relative humidity, wind speed, wind direction, latitude, longitude, and altitude, integrated from both LCS sensors and adjacent air quality monitoring stations. We evaluate the efficacy of Kriging as a spatial imputation method for PM2.5 forecasting, comparing its performance along with mean imputation and k-nearest neighbours, against a no-imputation benchmark to assess whether imputation techniques provide meaningful improvements in forecasting accuracy. Additionally, we compare its performance across data with varying levels of missingness. Our comparative analysis examines how these imputation techniques affect various prediction models, including Random Forest, Extreme Gradient Boosting, Convolutional Neural Networks, Long Short-Term Memory (LSTM), and LSTM with an Attention mechanism. Our findings indicate that although differences among imputation methods were modest, Kriging consistently yielded the best results, particularly at 90% data availability for the deep learning models. The LSTM-AM architecture produced the best overall results in a short-horizon setting, [Formula: see text] = 0.9602, MAE = 1.7815μg/m[Formula: see text], and RMSE = 4.5208μg/m[Formula: see text], suggesting an optimal combination of Kriging imputation with LSTM-AM modeling for near-term PM2.5 forecasting applications.