Atmospheric and Oceanic Sciences

Pedestrian detection from unmanned aerial vehicles (UAVs) holds significant value in security surveillance and emergency response applications. While visible-infrared (RGB-IR) fusion technology demonstrates potential in handling complex lighting conditions through cross-modal information complementarity, current mainstream fusion mechanisms still suffer from two evident shortcomings: (1) Existing approaches insufficiently account for the significant differences in noise distribution between infrared and visible images under varying imaging conditions, leading to unstable feature representations and posing fundamental challenges to subsequent effective fusion; and (2) Existing fusion strategies lack dynamic adaptability to features from different modalities, making it difficult to fully exploit complementary key information across modalities. To address these issues, this paper proposes a novel Laplacian Frequency Enhanced DETR (LF-DETR). The core innovations are threefold: (1) A Laplacian of Gaussian feature enhancement module is designed to independently enhance features in the visible and infrared branches at the early stage of feature extraction, effectively improving the representation quality of each modality. (2) A learnable frequency-domain fusion module is constructed to achieve adaptive complementary fusion of cross-modal features. (3) A dual-domain collaborative framework is proposed to integrate the above modules within a unified DETR architecture for RGB-IR pedestrian detection. Experimental results on the public RGBTDronePerson, VTUAV-det and DVTOD datasets demonstrate that LF-DETR achieves state-of-the-art performance, with particularly significant detection gains in challenging scenarios such as nighttime and low-light conditions, validating the effectiveness and superiority of the proposed method.

Unmanned aerial vehicles (UAVs) are increasingly used for crop monitoring, but their widespread adoption is limited since they often rely on non-standard specialized cameras equipped with near-infrared (NIR) sensors. More affordable and scalable crop monitoring solutions would be enabled, however, if data could be collected using standard RGB sensors. We compared visible-band indices that incorporate blue spectral range (NDGBI and NDRBI) with traditional NIR-based indices (NDVI and GNDVI) for their effectiveness in monitoring maize growth and nitrogen status. UAV multispectral data capture at different maize growth stages was complemented by ground-based spectroradiometer measurements for calibration and validation. Various agronomic and yield variables (including cornstalk NO3–N content, grain yield, grain moisture, number of corncobs, and grain test weight) were recorded to link spectral responses with plant performance and nutritional status. The results show that the overall performance of the RGB-based approach was comparable to that of the NIR-based approach, with the visible-band indices proving to be highly sensitive to physiological stress, chlorophyll degradation, and nitrogen variability in maize. Our findings highlight the potential of the RGB-based indices to complement or even replace specialized NIR-based indices, providing a cost-effective, high-resolution tool for precision agriculture.

Current synthetic aperture radar automatic target recognition (SAR ATR) tasks face challenges including limited training samples and poor generalization capability to novel classes. To address these issues, few-shot class-incremental learning (FSCIL) has emerged as a promising research direction. Few-shot learning facilitates the expedited adaptation to novel tasks utilizing a limited number of labeled samples, whereas incremental learning concentrates on the continuous refinement of the model as new categories are incorporated without eradicating previously learned knowledge. Although both methodologies present potential resolutions to the challenges of sample scarcity and class evolution in SAR target recognition, they are not without their own set of difficulties. Fine-tuning with emerging classes can perturb the feature distribution of established classes, culminating in catastrophic forgetting, while training exclusively on a handful of new samples can induce bias towards older classes, leading to distribution collapse and overfitting. To surmount these limitations and satisfy practical application requirements, we propose a Few-Shot Class-Incremental SAR Target Recognition method based on a Dynamic Task-Adaptive Classifier (DTAC). This approach underscores task adaptability through a feature extraction module, a task information encoding module, and a classifier generation module. The feature extraction module discerns both target-specific and task-specific characteristics, while the task information encoding module modulates the network parameters of the classifier generation module based on pertinent task information, thereby improving adaptability. Our innovative classifier generation module, honed with task-specific insights, dynamically assembles classifiers tailored to the current task, effectively accommodating a variety of scenarios and novel class samples. Our extensive experiments on SAR datasets demonstrate that our proposed method generally outperforms the baselines in few-shot class incremental SAR target recognition.

This study develops a practical framework for forecasting long-term drought conditions in Karaman Province, a semi-arid region of Turkey, where accurate climate information is vital for water planning and agriculture. Since the area has limited rainfall records and strong year-to-year fluctuations, traditional modeling approaches often fall short. To better capture local conditions, drought intensity was defined using a simple monthly wetness anomaly measure based directly on precipitation; here, positive values indicate wetter months and negative values indicate drier ones. This makes the method suitable for regions where detailed hydrological data are scarce. Rainfall observations from 1965 to 2011 were expanded using a combination of kernel density estimation and Cholesky-based correlation reconstruction. These steps preserved the main statistical and temporal patterns of the original data while increasing sample diversity. The enriched dataset was then used to train artificial neural networks to predict both precipitation and drought intensity. The models reached R2 values of 0.76 and 0.72, with mean absolute errors of 12.8 mm and 28.4%, which represents an improvement of roughly 10–15% over traditional statistical methods. They were also able to capture the seasonal and year-to-year variability that strongly affects drought conditions in the region. To understand what drives the predictions, the model was examined with LIME, which consistently highlighted lagged rainfall and seasonal indicators as the most influential inputs. A walk-forward validation approach was also used to mimic real forecasting conditions and demonstrated that the model remains stable when projecting into the future. Overall, the proposed framework offers a reliable and practical basis for early-warning efforts and drought-management strategies in semi-arid regions like Karaman.

This study examines two wildfire events in the southern Amazon in August 2021, addressing the challenges in investigating the development of pyro-convective clouds in tropical regions. The analysis combines the Normalized Difference Vegetation Index, Fire Radiative Power derived from the Suomi-NPP and NOAA-20 satellites, and meteorological conditions from thermodynamic profiles and atmospheric modeling. The Meso-NH model was applied exploratorily with two simulations that allow convection, at a 2.5 km resolution. In the first case, a pyro-convective cloud (PyroCu) formed directly from active fires. In the second, a deep convective cloud developed over dispersed fire hotspots, exhibiting characteristics compatible with pyro-convective activity, although uncertainties remain regarding its classification as a true PyroCb. The results indicate that background thermodynamic instability primarily controls vertical plume development, modulating the influence of fire intensity. Incorporating high-resolution thermodynamic profiles into coupled atmospheric and chemical dispersion models can improve estimates of smoke injection height, complementing information on fire power. The results provide a basis for future developments related to understanding tropical pyro-convective clouds, showing how smoke dispersion may occur in the tropical region depending on the vertical structure of the atmosphere and fire intensity.

Acoustic precipitation enhancement (APE) is an emerging non-chemical weather-modification technique, yet quantitative three-dimensional evidence of its impact on rainy clouds remains scarce. This study investigates a stratiform precipitation event over the Bayinbuluke Basin in the central Tianshan Mountains of northwestern China, 29–30 August 2024, using an X-band phased-array weather radar (X-PAR) coordinated with an upward-directed acoustic source. Rapid volumetric scans and sector-aligned range-height indicators were combined to reconstruct the three-dimensional cloud structure before, during, and after acoustic operation. During acoustic operation, the results were stronger and more persistent than during the non-operation period, with localized values exceeding 40 dBZ. Within the 3 km influence zone, low-level reflectivity increased across all azimuthal sectors with clear directional dependence. Dual-ratio analysis showed statistically significant enhancement in the windward sector (247°, DR = 1.91, p = 0.0004) and the leeward sector (137°, DR = 1.51, p = 0.008), indicating that acoustic-induced responses extended beyond the primary radiation sector and propagated downstream with cloud advection. These results, based on a single stratiform precipitation case, demonstrate that volumetric X-PAR observations can detect localized cloud-structure responses during acoustic operation.

Much of the information for studying the processes of lightning discharge initiation and development is provided by studying thundercloud radio emissions in various frequency bands. The High-Frequency (HF) band better corresponds to the characteristic scales of lightning development but has been undeservedly forgotten and is used quite rarely. Based on observations carried out in the Upper Volga region it is shown that the intensity of HF radio emissions from lightning is high enough to be reliably recorded in nearby thunderstorms. It is found that at all stages of lightning development the intensity of its radio emission in the HF band up to 10 MHz exceeds the average background intensity significantly. The amplitudes of lightning pulses exceed the background level more significantly, up to 60 dB and more. The feasibility of using the HF band for lightning observations including tracing the direction of arrival of a radio emission is clearly demonstrated.

Abstract Agricultural sector forms the backbone of India's socioeconomy. Despite consistent efforts to develop the irrigation network for increased agricultural production, approximately 60% of India's cropped area still remains rain fed, thereby functional to the strength of the Indian summer monsoon (ISM). The probability of a weak monsoon is strongly correlated with the occurrence of droughts during the ISM season. The possible influence of atmospheric drivers in initiating these droughts is observed through the dynamics of low‐level jets (LLJ) over the northern Arabian Sea forming the core of LLJ during the ISM. Interestingly, since the past 72 years, from 1951 to 2022, the core of LLJ has become dry (increased saturation deficit by 17%) and weak (reduced wind speed by 5%). Additionally, the wind speed (saturation deficit) at the core shares close dependence (75%–80%) with the initiation of dry (wet) extremes exhibiting maximum correlation at a 2‐day lag. Furthermore, we observed a 50% (40%) increase in the dry (wet) extremes driven by the lower‐atmospheric dynamics of the LLJ core. Consequently, these dry (wet) extremes are characterized by a 6% (12%) enhancement in duration (intensity). Such conditions act as strong precursors for monsoon droughts.

Abstract Atmospheric rivers are episodic events that can advect relatively large quantities of moisture to Antarctica, contributing to both disproportionate precipitation and melting events. The Year of Polar Prediction, an international effort to improve weather prediction over the southern polar region, presents an opportunity to study the clouds and precipitation associated with winter atmospheric river events. This study uses enhanced surface, profile, and remote‐sensing observations from the Antarctic Peninsula (AP) during a Targeted Observing Period around 16 May 2022, when an event occurred with local warming similar to a warm front. We compare regional atmospheric simulations with the polar‐optimized version of the Weather Research and Forecasting Model to various in situ and remote‐sensing observations. The study emphasizes data from three stations: Escudero, Vernadsky, and Rothera. Mixed‐phase clouds were simulated at the three stations, with the precipitation being primarily rain at Escudero and primarily snow at Vernadsky and Rothera. The model produced reasonable simulations of the clouds and precipitation. Furthermore, modeled longwave cloud forcing at Escudero had small errors compared to observed values. A sensitivity test enhancing secondary ice production indicates mixed‐phase cloud sensitivity to the Hallett‐Mossop process, especially at Rothera.

Abstract The climate impact of dust is still uncertain, partially due to poorly constrained dust physical and optical properties. Natural dust particles are known to have highly irregular shapes, but many models assume spheres when calculating the direct radiative effect (DRE). While the superior performance of non‐spherical shapes in remote sensing applications has been widely recognized, there has been no consensus about the importance of dust non‐sphericity in climate models. We assess the extent of the shape effect upon the dust optical properties and DRE at shortwave wavelengths within the NASA Goddard Institute for Space Studies ModelE2.1. We assume tri‐axial ellipsoids as an approximation to natural dust shapes that is suitable for model applications, and combine a widely used database of ellipsoidal single‐scattering properties with a recent shape distribution constructed from a comprehensive compilation of measurements. We find a shape‐induced enhancement of global dust extinction of , resulting in a global cooling increase of at the top of atmosphere and at the surface. The ellipsoidal shapes increase the total dust extinction per unit mass, improving our model representation of the dust optical depth compared to a semi‐observational constraint and ground‐based measurements. However, our analysis shows that the database of ellipsoidal scattering properties covers only one‐third of the observed distribution of shapes. This represents a major uncertainty in evaluating the dust shape effect in model applications.

Abstract Prolonged compound hot–dry events (CHDEs) have been linked to surging emergency medical demand, especially among vulnerable populations, making them a growing public health concern in a warming climate. However, the seasonal and regional drivers of high health–risk CHDEs (HHR–CHDEs) remain unclear, limiting effective public health responses. Here, we integrate a temperature–humidity-based health risk index, ambulance dispatch records, and distributed lag non-linear models to obtain high health-risk temperature–humidity thresholds. During HHR–CHDEs exposure, particularly higher emergency dispatch demands are observed among males, the elderly, and individuals suffering from trauma or alcohol poisoning We further use ERA5 reanalysis data to examine the dominant modes and physical mechanisms of HHR–CHDEs across China during the early and late warm seasons. In the early warm season, HHR–CHDEs are concentrated in Northwest China and are jointly driven by enhanced surface heating (51% contribution) and intensified moisture loss (37%), underpinned by the synergistic effect of an upstream Rossby wave source over Eastern Europe and anomalous Northwest Pacific SST warming. This coupling promotes a quasi-stationary high-pressure anomaly that blocks synoptic disturbances and reinforces regional heat–dryness feedbacks. By contrast, in the late warm season, HHR–CHDEs are centered over the middle–lower Yangtze River region and are dominated by persistent heat accumulation (65%) with a secondary drought contribution (25%). These events are driven by large-scale circulation anomalies resulting from upstream Rossby wave energy originating in the North Atlantic and enhanced by downstream SST warming over the subtropical Northwest Pacific. This Atlantic–Pacific coupling induces pronounced adjustments in the Walker and Hadley circulations, promoting subsidence and suppressed monsoonal moisture transport, and sustaining a health-threatening hot–dry atmosphere over densely populated regions. These findings reveal seasonally distinct remote forcing pathways and highlight the value of integrating large-scale diagnostics into climate–health early warning systems.

Abstract Fields of equatorial upper tropospheric circulation data are lag regressed against wavelet-filtered indexes of upper tropospheric zonal wind anomalies over a range of phase speeds at 50-day periods to study the propagation mechanisms of the upper tropospheric circulation signal of the Madden Julian oscillation (MJO) over the Indian Ocean. Results show that the MJO upper tropospheric zonal wind is accelerated by the geopotential gradient force in quadrature with the existing wind anomaly, yielding its eastward propagation. This effect is offset by Doppler advection of the MJO wind by the easterly background wind. Divergence of mass by the zonal wind propagates the associated geopotential height anomalies in concert with the winds. Results confirm that advection of background wind by the MJO wind amplifies MJO wind anomalies in phase with those anomalies where the background wind is zonally confluent and breaks it down in regions of diffluent background wind. Coincidence of the mass source driven by moist convection and zonally diffluent zonal wind anomalies with falling geopotential height relative to the regions east and west is consistent with planetary scale Kelvin wave zonal wind signals providing favorable conditions for convection. Circulation data at 100 hPa confirms approximately dry Kelvin wave dynamics driving MJO-associated equatorial circulations eastward.

Abstract. To what extent the new particle formation (NPF) contributed to the cloud condensation nuclei (CCN) remained unclear, especially at the boundary layer top (BLT) in polluted atmosphere. Based on measurements at a mountain-top background site in southeastern China during spring 2024, this study systematically investigates the nucleation mechanism and subsequent growth dynamics of NPF events under contrasting air masses, and quantifies their role as a source of CCN. Eight NPF events were observed, and three of them occurred in the polluted conditions (NPF-P) which associated with regional transportation while the rest five events appeared in the clean conditions (NPF-C). The average formation rate (J2.5: 2.4 cm-3s-1 vs. 0.7 cm-3s-1) and growth rate (GR: 6.8 nm h−1 vs. 5.5 nm h−1) were significantly higher in NPF-P events than in NPF-C events, alongside elevated concentrations of sulfuric acid and ammonia. The correlation between log J3 and [H2SO4], as well as theoretical simulations with the MALTE_BOX model, indicates that the enhanced nucleation in polluted conditions can be attributed to the participation of ammonia in stabilizing sulfuric acid-based clusters. In addition, much higher CCN enhancement factor was observed in NPF-P (EFCCN: 1.6 vs. 0.7 in NPF-C) due to the regional transported of anthropogenic pollutants from the urban cluster regions and their secondary transformation under enhanced atmospheric oxidation capacity. Furthermore, the duration of NPF-to-CCN conversion was quantified using a “Time Window (τ)”, revealing that polluted conditions accelerated the conversion by 17.0 % (τ = 16.4 h vs. 19.8 h). Nitrate played an important role in maintaining a rapid particle growth rate, thereby shortening τ and enhancing CCN production from NPF – a process that can ultimately influence cloud microphysical properties by increasing the potential cloud droplet number concentration. These findings reveal that polluted air masses enhance both the efficiency and speed of CCN production at the BLT through elevated atmospheric oxidation capacity.

Abstract. The radiative properties of clouds depend partially on the cloud droplet number concentration, which is determined by the concentration of cloud condensation nuclei (CCN) when the clouds are formed. In turn, CCN concentrations are determined by the atmospheric particle size distribution and their chemical composition. We use a novel Lagrangian modelling framework to examine the origins and history of gas and aerosol components observed at the boreal forest measurement site SMEAR II, and their potential to act as CCN. This framework combines: (a) global emission datasets, (b) backward trajectories from the FLEXible PARTicle dispersion model (FLEXPART) airmass dispersion model, (c) a detailed description of atmospheric chemistry and aerosol dynamics from the Model to Simulate the Concentration of Organic Vapours, Sulphuric Acid and Aerosol Particles (SOSAA). We apply this SOSAA-FP (FP standing for FLEXPART) framework to simulate a period from March to October 2018 with 1 h time resolution, focusing on the concentrations of CCN between 0.1 % and 1.2 % maximum supersaturation as calculated by the κ-Köhler theory (with respective dry particle diameter of activation ca. 175–35 nm). We find that the model PM1 fraction of primary particles, sulfates and secondary organic aerosol correlate well with the observed organic aerosol and sulfate trends and explain most of the observed organic aerosol and sulfate PM1 mass. Our results show that primary particle emissions play a considerable role in CCN concentrations even at a rural site such as SMEAR II. Changes in atmospheric cluster formation rates had a relatively weak impact on the CCN concentrations in the sensitivity runs. Enhanced cluster formation increased (decreased) the CCN concentrations for the highest (lowest) maximum supersaturation. Without any cluster formation our modelled median CCN concentrations changed by −48 % and +23 % for supersaturations of 1.2 % and 0.1 %, respectively, whereas omitting primary particle emissions had a decreasing effect in all calculated CCN supersaturation classes (−82 % and −33 % decrease in median CCN of 1.2 % and 0.1 % supersaturation, respectively). While the enhancing effect of cluster formation to high supersaturation (i.e., small diameter) CCN concentrations is unsurprising, the weak sensitivity to cluster formation rates and the decreasing effect to lowest supersaturation CCN was unexpected, as was the strong influence of anthropogenic primary emissions. The Lagrangian model framework showed its power, as it was possible to trace down the causes behind the unexpected outcomes by comparing how the particle population evolved along the trajectories in different sensitivity tests.