Explore the vast range of titles available.

Large language models (LLMs) based on artificial intelligence have found applications across various sectors—including medicine, education, science, literature, and marketing. Although they offer considerable opportunities, their limitations also raise important concerns. This study evaluates several AI tools in the context of meteorology and climatology. The tools examined include ChatGPT o3-mini, o1, 4.o, 4.0; Gemini Advanced 1.5 and 2.0; Copilot; Perplexity; DataAnalyst; Consensus; ScholarGPT; SciSpace; Claude; and DeepSeek. The evaluation tasks comprised cloud recognition and classification from photographs, gap-filling in literature reviews, map creation based on provided datasets, comparative interpretation of maps, and archival data retrieval from line graphs converted to numerical data. Each task was rated on a 0–5 scale. Conducted between February 2024 and February 2025, the study found that ChatGPT o3-mini excelled in cloud classification; ChatGPT4.o and ScholarGPT produced high-quality maps; Claude 3.5 Sonnet and SciSpace provided the most detailed map descriptions; and Consensus and ChatGPT o1 were the most effective for literature review support. However, all tools performed poorly in regards to archival data retrieval, with Claude 3.5 Sonnet yielding the smallest errors. Overall, substantial progress was observed over the study period.

With the advancement of oblique photography technology, dense matching point clouds have found widespread application. As a key step in point cloud data processing, point cloud classification has gained continuous attention in 3D computer vision and spatial information processing. Addressing the challenges posed by the high-dimensional and redundant nature of point cloud data, which often leads to low classification accuracy, this study proposes a point cloud classification method based on feature selection and ensemble learning. First, we use a combination of Pearson correlation coefficient and random forest feature importance methods to screen point cloud data features, thereby eliminating redundant features to enhance the expression of critical features. Next, we construct a stacked ensemble model, KNN-SVM-RF-XG, using K-Nearest Neighbors, Support Vector Machine, Random Forest, and XGBoost as base models, with logistic regression as the meta-model. The experimental results demonstrated that the feature selection methods significantly enhanced the classification accuracy of point clouds, achieving peak accuracy of 96.03%. KNN-SVM-RF-XG model attained classification accuracy and precision of 96.20% and 96.19%, respectively, surpassing the performance of individual classifiers. Furthermore, the proposed model exhibited superior classification capabilities compared to deep learning approaches such as PointNet, PointNet++, and Transformer-based architectures, showcasing robust generalization ability. This research provides technical support for urban data renewal, land-use surveys, and urban planning management.

A convective and stratiform cloud classification method for weather radar is proposed based on the density-based spatial clustering of applications with noise (DBSCAN) algorithm. To identify convective and stratiform clouds in different developmental phases, two-dimensional (2D) and three-dimensional (3D) models are proposed by applying reflectivity factors at 0.5° and at 0.5°, 1.5°, and 2.4° elevation angles, respectively. According to the thresholds of the algorithm, which include echo intensity, the echo top height of 35 dBZ (ET), density threshold, and ε neighborhood, cloud clusters can be marked into four types: deep-convective cloud (DCC), shallow-convective cloud (SCC), hybrid convective-stratiform cloud (HCS), and stratiform cloud (SFC) types. Each cloud cluster type is further identified as a core area and boundary area, which can provide more abundant cloud structure information. The algorithm is verified using the volume scan data observed with new-generation S-band weather radars in Nanjing, Xuzhou, and Qingdao. The results show that cloud clusters can be intuitively identified as core and boundary points, which change in area continuously during the process of convective evolution, by the improved DBSCAN algorithm. Therefore, the occurrence and disappearance of convective weather can be estimated in advance by observing the changes of the classification. Because density thresholds are different and multiple elevations are utilized in the 3D model, the identified echo types and areas are dissimilar between the 2D and 3D models. The 3D model identifies larger convective and stratiform clouds than the 2D model. However, the developing convective clouds of small areas at lower heights cannot be identified with the 3D model because they are covered by thick stratiform clouds. In addition, the 3D model can avoid the influence of the melting layer and better suggest convective clouds in the developmental stage.

Cloud classification is not only important for weather forecasts, but also for radiation budget studies. Although cloud mask and classification procedures have been proposed for Himawari-8 Advanced Himawari Imager (AHI), their applicability is still limited to daytime imagery. The split window algorithm (SWA), which is a mature algorithm that has long been exploited in the cloud analysis of satellite images, is based on the scatter diagram between the brightness temperature (BT) and BT difference (BTD). The purpose of this research is to examine the usefulness of the SWA for the cloud classification of both daytime and nighttime images from AHI. We apply SWA also to the image data from Moderate Resolution Imaging Spectroradiometer (MODIS) onboard Aqua and Terra to highlight the capability of AHI. We implement the cloud analysis around Japan by employing band 3 (0.469 μm) of MODIS and band 1 (0.47 μm) of AHI for extracting the cloud-covered regions in daytime. In the nighttime case, the bands that are centered at 3.9, 11, 12, and 13 µm are utilized for both MODIS and Himawari-8, with somewhat different combinations for land and sea areas. Thus, different thresholds are used for analyzing summer and winter images. Optimum values for BT and BTD thresholds are determined for the band pairs of band 31 (11.03 µm) and 32 (12.02 µm) of MODIS (SWA31-32) and band 13 (10.4 µm) and 15 (12.4 µm) of AHI (SWA13-15) in the implementation of SWA. The resulting cloud mask and classification are verified while using MODIS standard product (MYD35) and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) data. It is found that MODIS and AHI results both capture the essential characteristics of clouds reasonably well in spite of the relatively simple scheme of SWA based on four threshold values, although a broader spread of BTD obtained with Himawari-8 AHI (SWA13-15) could possibly lead to more consistent results for cloud-type classification than SWA31-32 based on the MODIS sensors.

Studies have implicated the importance of longwave (LW) cloud‐radiative forcing (CRF) in facilitating or accelerating the upscale development of tropical moist convection. While different cloud types are known to have distinct CRF, their individual roles in driving upscale development through radiative feedback is largely unexplored. Here we examine the hypothesis that CRF from stratiform regions has the greatest positive effect on upscale development of tropical convection. We do so through numerical model experiments using convection‐permitting ensemble WRF (Weather Research and Forecasting) simulations of tropical cyclone formation. Using a new column‐by‐column cloud classification scheme, we identify the contributions of five cloud types (shallow, congestus, and deep convective; and stratiform and anvil clouds). We examine their relative impacts on longwave radiation moist static energy (MSE) variance feedback and test the removal of this forcing in additional mechanism‐denial simulations. Our results indicate the importance stratiform and anvil regions in accelerating convective upscale development. Plain Language Summary Infrared or longwave radiation and its interaction with clouds is important in the formation of tropical storms. Given the different shapes and distributions of distinct cloud types, we hypothesize that they interact with longwave radiation differently, and therefore exert different impacts on the organization of tropical convection. This issue has largely been unexplored. To address this gap, we tested our hypothesis by analyzing numerical model simulations of the formation of two tropical cyclones. Further, we developed a new cloud classification scheme based on cloud properties that identifies five distinct cloud types. Using this classification, we examined the impact of radiative interactions with different cloud types on the development of tropical storms by turning off this feedback in specific cloud types. Our results indicate that light‐raining regions, such as stratiform and anvil clouds, contribute dominantly to longwave cloud‐radiative trapping and the moistening of convective regions. This is due to both these cloud types' strong greenhouse trapping effect and their extensive areal coverage, which spreads this effect over large regions of a developing storm. Key Points A new column‐by‐column cloud microphysical classification scheme is developed for application with numerical models Radiative feedback due to stratiform and anvil clouds is a leading driver of tropical convective upscale development The local radiative forcing by deep convective regions is similar in magnitude to stratiform but its impact is limited by its smaller area

The cloud classification algorithm widely used in the International Satellite Cloud Climatology Project (ISCCP) tends to underestimate low clouds over the Tibetan Plateau (TP), often mistaking water clouds for high‐level clouds. To address this issue, we propose a new algorithm based on cloud‐top temperature and optical thickness, which we apply to TP using Advanced Himawari Imager (AHI) geostationary satellite data. Compared with Clouds and the Earth's Radiant Energy System cloud‐type products and ISCCP results obtained from AHI data, this new algorithm markedly improved low‐cloud detection accuracy and better aligned with cloud phase results. Validation with lidar cloud‐type products further confirmed the superiority of this new algorithm. Diurnal cloud variations over the TP show morning dominance shifting to afternoon high clouds and evening mid‐level clouds. Winter is dominated by high clouds, summer by mid‐level clouds, spring by daytime low clouds and nighttime high clouds, and autumn by low and mid‐level clouds. Plain Language Summary The accurate identification of low clouds over the Tibetan Plateau (TP) is crucial for climate regulation, ecosystems, aviation safety, research, and modeling. However, satellite‐based methods often miss these clouds, misclassifying them as high‐level clouds. To remedy this, we developed a new algorithm using cloud‐top temperature and optical thickness, applied to Advanced Himawari Imager data. This significantly improves low‐cloud detection, better aligning with actual cloud phases. Simultaneously, we analyzed diurnal cloud variations over the TP with the new algorithm. Cloud types at different altitudes in the TP exhibit strong seasonality. The dominant cloud types in winter and summer are high and mid‐level, respectively. In spring, low clouds dominate during the day (2:00–10:00 UTC), transitioning to high clouds at night (10:00–18:00 UTC), with mid‐level clouds prevailing at other times. In autumn, low clouds dominate during the day, transitioning to mid‐level clouds at other times, with fewer occurrences of high clouds. Key Points Employing cloud‐top temperature instead of pressure resolves classification‐phase inconsistencies for clouds in the Tibetan Plateau (TP) Lidar validation shows new algorithm's low cloud detection outperforms the conventional International Satellite Cloud Climatology Project algorithm for both TP and plains The study reveals significant diurnal and seasonal variations in low clouds over the TP

Semantic segmentation of ground-based all-sky images (ASIs) can provide high-resolution cloud coverage information of distinct cloud types, applicable for meteorology-, climatology- and solar-energy-related applications. Since the shape and appearance of clouds is variable, and there is high similarity between cloud types, a clear classification is difficult. Therefore, most state-of-the-art methods focus on the distinction between cloudy and cloud-free pixels without taking into account the cloud type. On the other hand, cloud classification is typically determined separately at the image level, neglecting the cloud's position and only considering the prevailing cloud type. Deep neural networks have proven to be very effective and robust for segmentation tasks; however they require large training datasets to learn complex visual features. In this work, we present a self-supervised learning approach to exploit many more data than in purely supervised training and thus increase the model's performance. In the first step, we use about 300 000 ASIs in two different pretext tasks for pretraining. One of them pursues an image reconstruction approach. The other one is based on the DeepCluster model, an iterative procedure of clustering and classifying the neural network output. In the second step, our model is fine-tuned on a small labeled dataset of 770 ASIs, of which 616 are used for training and 154 for validation. For each of them, a ground truth mask was created that classifies each pixel into clear sky or a low-layer, mid-layer or high-layer cloud. To analyze the effectiveness of self-supervised pretraining, we compare our approach to randomly initialized and pretrained ImageNet weights using the same training and validation sets. Achieving 85.8 % pixel accuracy on average, our best self-supervised model outperforms the conventional approaches of random (78.3 %) and pretrained ImageNet initialization (82.1 %). The benefits become even more evident when regarding precision, recall and intersection over union (IoU) of the respective cloud classes, where the improvement is between 5 and 20 percentage points. Furthermore, we compare the performance of our best model with regards to binary segmentation with a clear-sky library (CSL) from the literature. Our model outperforms the CSL by over 7 percentage points, reaching a pixel accuracy of 95 %.

Marine cold air outbreaks (CAOs) frequently occur in the Arctic when cold air moves over the relatively warm ocean, resulting in large turbulent fluxes, instability and cloud formation. Given the high frequency of CAOs during the Arctic winter, the associated clouds have a large impact on the region's radiative balance. Due to Arctic warming, the prevalence of CAOs and their clouds may change, impacting the Arctic radiative balance and potentially amplifying or mitigating local and global warming. To better understand how CAO clouds respond to Arctic warming, this study has developed a phenomenological CAO cloud classification tool that utilizes machine learning methods to identify closed and open cell clouds in CAOs from MODIS satellite imagery. This new approach achieves better performance in identifying CAO clouds compared to the marine cold air outbreak index calculated using MERRA-2 reanalysis, with accuracies of 85.4 % and 78.0 %, respectively. The new approach has revealed frequent CAO cloud formation in regions of high sea surface temperatures, with occurrence maxima along the Norwegian coast and the Northern Atlantic region south of Iceland. Furthermore, the approach reveals trends in CAO cloud cover that suggest a shortening of the CAO season, characterized by an approximate 10 %, increase in cloud coverage during winter and a nearly 20 % decrease during the shoulder months over the past 25 years. These trends suggest a positive radiative feedback during winter in response to climate change, underscoring the importance of further investigating these clouds to understand the trajectory of future Arctic climate.

Three supervised neural network cloud classification routines are applied to daytime MODIS Aqua imagery and compared for the year 2018 over the North Atlantic Ocean. Routines surveyed here include the Morphology Identification Data Aggregated over the Satellite-era (MIDAS), which specializes in subtropical stratocumulus (Sc) clouds; sugar, gravel, flowers, and fish (SGFF), which is focused on shallow cloud systems in the tropical trade winds; and the community record of marine low-cloud mesoscale morphology supported by the NASA Making Earth System Data Records for Use in Research Environments (MEaSUREs) dataset, which is focused on shallow clouds globally. Comparisons of co-occurrence and vertical and geographic distribution show that morphologies are classified in geographically distinct regions; shallow suppressed and deeper aggregated and disorganized cumulus are seen in the tropical trade winds. Shallow Sc types are frequent in subtropical subsidence regions. More vertically developed solid stratus and open- and closed-cell Sc are frequent in the mid-latitude storm track. Differing classifier routines favor noticeably different distributions of equivalent types. Average scene albedo is more strongly correlated with cloud albedo than cloud amount for each morphology. Cloud albedo is strongly correlated with the fraction of optically thin cloud cover. The albedo of each morphology is dependent on latitude and location in the mean anticyclonic wind flow over the North Atlantic. Strong rain rates are associated with middling values of albedo for many cumuliform types, hinting at a complex relationship between the presence of heavily precipitating cores and cloud albedo. The presence of ice at cloud top is associated with higher albedos. For a constant albedo, each morphology displays a distinct set of physical characteristics.

Automatic recognition of ground-based clouds is crucial for meteorology and especially for the operational safety of Unmanned Aerial Vehicles (UAVs), but it is challenged by variable cloud shapes, complex lighting, and background interference. This paper introduces ALGA-DenseNet, an improved DenseNet model with a multi-scale attention mechanism. The model employs Color Jitter to enhance image robustness and improve learning of intra-class variations and inter-class differences. It incorporates Adaptive Local and Global Attention (ALGA) to merge features, enhancing feature selection. Additionally, it integrates mixed and depthwise separable convolutions to optimize multi-scale feature extraction, reducing parameters and computational complexity. Furthermore, integrating a Vision Transformer (ViT) and Dynamic Multi-head Attention (DMA) enhances representation of complex cloud features. Experimental results show recognition accuracies of 97.94% on the TJNU (Tianjin Normal University) Ground-based Cloud Dataset (GCD) and 97.25% on the Cirrus Cumulus Stratus Nimbus (CCSN) dataset. This indicates the model’s capability for fine-grained, multi-scale extraction of cloud textures, shapes, and color features, along with strong generalization performance.