Tropical overshooting convection has a strong impact on both heat budget and moisture distribution in the upper troposphere and lower stratosphere, and it can pose a great risk to aviation safety. Cloud-top height is one of the essential concerns of overshooting convection for both the climate system and the aviation weather forecast. The main purpose of our work is to verify the application of the machine learning method, taking the random forest (RF) model as an instance, in overshooting cloud-top height retrieval from Himawari-8 data. By using collocated CloudSat observations as a reference, we utilize several infrared indicators of Himawari-8 that are commonly recognized to relate to cloud-top height, along with some temporal and geographical parameters (latitude, month, satellite zenith angle, etc.), as predictors to construct and validate the model. Analysis of variable importance shows that the brightness temperature of 6.2 um acts as the dominant predictor, followed by satellite zenith angle, brightness temperature of 13.3 um, latitude, and month. In the comparison between the RF model and the traditional single-channel interpolation method, retrievals from the RF model agree well with observation with a high correlation coefficient (0.92), small RMSE (222 m), and small MAE (164 m), while these metrics from traditional single-channel interpolation method shows lower skills (0.70, 1305 m, and 1179 m). This work presents a new sight of overshooting cloud-top height retrieval based on the machine learning method.
The magnitude of damage caused by hail depends on its size; however, direct observation or indirect estimation of hail size remains a significant challenge. One primary reason for estimations by proxy, such as through remote sensing methods, is that empirical relationships or statistical models established in one region may not apply to other areas. This study employs a machine learning method to build a hail size estimation model without assuming relations in advance. It uses FY-4A AGRI data to provide cloud-top information and ERA5 data to add vertical environment information. Before training the model, we conducted a principal component analysis (PCA) to analyze the highly influential factors on hail sizes. A total of 18 features, composed of four groups, namely brightness temperature (BT), the difference in BT (BTD), thermodynamics, and dynamics groups, were chosen from 29 original features. Dynamic and BTD features show superior performance in identifying large hail. Although the selected features are more closely correlated to hail sizes than unselected ones, the relationships are complicated and nonlinear. As a result, a two-layer regression back propagation neural network (BPNN) model with powerful fitting ability is trained with selected features to predict maximum hail diameter (MHD). The linear fitting R2 between predicted and observed MHDs is 0.52 on the test set, which signifies that our model performs well compared with other hail size estimation models. We also examine the model concerning all three hail cases in Shanghai, China, between 2019 and 2021. The model attained more satisfactory results than the radar-based maximum estimated hail size (MEHS) method, which overestimates the MHDs, thus further supporting the operational applications of our model.
Hail poses a significant meteorological hazard in China, leading to substantial economic and agricultural damage. To enhance the detection of hail and mitigate these impacts, this study presents an ensemble machine learning model (BPNN+Dtree) that combines a backpropagation neural network (BPNN) and a decision tree (Dtree). Using FY-4A satellite and ERA5 reanalysis data, the model is trained on geostationary satellite infrared data and environmental parameters, offering comprehensive, all-day, and large-area hail monitoring over China. The ReliefF method is employed to select 13 key features from 29 physical quantities, emphasizing cloud-top and thermodynamic properties over dynamic ones as input features for the model to enhance its hail differentiation capability. The BPNN+Dtree ensemble model harnesses the strengths of both algorithms, improving the probability of detection (POD) to 0.69 while maintaining a reasonable false alarm ratio (FAR) on the test set. Moreover, the model’s spatial distribution of hail probability more closely matches the observational data, outperforming the individual BPNN and Dtree models. Furthermore, it demonstrates improved regional applicability over overshooting top (OT)-based methods in the China region. The identified high-frequency hail areas correspond to the north-south movement of the monsoon rain belt and are consistent with the northeast-southwest belt distribution observed using microwave-based methods.
Cross-view geo-localization is a challenging task due to the significant changes in the appearance of target scenes from variable perspectives. Most existing methods primarily adopt Transformers or ConvNeXt as backbone models but often face high computational costs and accuracy degradation in complex scenarios. Therefore, this paper proposes a visual Mamba framework based on the state-space model (SSM) for cross-view geo-localization. Compared with the existing methods, Vision Mamba is more efficient in modeling and memory usage and achieves more efficient cross-view matching by combining the twin architecture of shared weights with multiple mixed losses. Additionally, this paper introduces Dice Loss to handle scale differences and imbalance issues in cross-view images. Extensive experiments on the public cross-view dataset University-1652 demonstrate that Vision Mamba not only achieves excellent performance in UAV target localization tasks but also attains the highest efficiency with lower memory consumption. This work provides a novel solution for cross-view geo-localization tasks and shows great potential to become the backbone model for the next generation of cross-view geo-localization.
Vertically developing convective clouds (VDCCs), characterized by cloud-top ascent and cooling, are critical precursors to severe convective weather due to their association with intense updrafts. However, existing studies are constrained by limited spatiotemporal resolution of data and tracking methodologies, hindering real-time and pixel-level capture of VDCC evolution. Furthermore, large-scale statistical analyses of VDCC spatiotemporal distribution remain scarce compared with mature convective systems, particularly in topographically complex regions like the Tibetan Plateau (TP). To address these challenges, we integrated an optical flow algorithm (for dense atmospheric motion vector (AMV) retrieval) with cloud-top cooling rates (CTCRs, as indicators of vertical development), leveraging the high spatiotemporal resolution and multispectral capabilities of the GEO-KOMPSAT-2A (GK2A) satellite. This approach achieved pixel-level VDCC detection at 10 min intervals across diurnal cycles, enabling comprehensive statistical analysis. Based on this technical foundation, the most important finding in the study was the distinct convective spatiotemporal distribution over the TP and East Asia (EA) by analyzing VDCC detection data in three summers (2021–2023). Specifically, VDCC diurnal peaks preceded precipitation by 2–3 h, confirming their precursor roles in both study regions. Regional comparisons revealed that topographic and thermal forcing strongly influenced VDCC distribution patterns. The TP exhibited earlier and more frequent daytime convection at middle-to-low levels than EA, driven by intense thermal forcing, yet vertical development was limited by moisture scarcity. In contrast, EA’s monsoonal moisture sustained deeper convection, with more VDCCs penetrating the upper troposphere. The detection and statistical studies of VDCCs offer new insights into convective processes over the TP and surrounding regions, offering potential improvements in severe weather monitoring and early warning systems.