Search Results (4)

This study introduces an advanced ensemble methodology employing lightweight neural network models for identifying severe convective clouds from FY-4B geostationary meteorological satellite imagery. We have constructed a FY-4B based severe convective cloud dataset by a combination of algorithms and expert judgment. Through the ablation study of a model ensembling combination of multiple specialized lightweight architectures—ENet, ESPNet, Fast-SCNN, ICNet, and MobileNetV2—the optimal EFNet (ENet- and Fast-SCNN-based network) not only achieves real-time processing capabilities but also ensures high accuracy in severe weather detection. EFNet consistently outperformed traditional, heavier models across several key performance indicators: achieving an accuracy of 0.9941, precision of 0.9391, recall of 0.9201, F1 score of 0.9295, and computing time of 18.65 s over the test dataset of 300 images (~0.06 s per 512 × 512 pic). ENet shows high precision but misses subtle clouds, while Fast-SCNN has high sensitivity but lower precision, leading to misclassifications. EFNet’s ensemble approach balances these traits, enhancing overall predictive accuracy. The ensemble method of lightweight models effectively aggregates the diverse strengths of the individual models, optimizing both speed and predictive performance. Full article

Severe convection is a disastrous mesoscale weather system. The early detection of such systems is very important for saving peoples’ lives and properties. Previous studies address the issue mainly based on thresholding methods, which are not robust and accurate enough. In this paper, we propose a novel semantic segmentation method (Dugs-UNet) to solve the problem. Our method is based on the well-known U-Net framework. As convective clouds mimic fluids, its detection faces two important challenges. First, the shape and boundary features of clouds need to be carefully exploited. Second, the positive and negative samples for convection detection are very imbalanced. To address the two challenges, our method was carefully developed. Regarding the importance of the shape and boundary features for convective target detection, we introduce a shape stream module to extract these features. Also, a data-dependent upsample operation is adopted in the decoder of U-Net to effectively utilize the features. This is one of our contributions. To address the imbalance issue for convective target detection, the a focal loss function is employed to train our method, which is another contribution. Experimental results of 2018 Fengyun-4A satellite observations in China demonstrate the effectiveness of the proposed method. Compared to conventional thresholding-based methods and deep semantic segmentation algorithms such as SegNet, PSPNet, DeepLav-v3+ and U-Net, the proposed approach performs the best. Full article

Severe convective weather is hugely destructive, causing significant loss of life and social and economic infrastructure. Based on the U-Net network with the attention mechanism, the recurrent convolution, and the residual module, a new model is proposed named ARRU-Net (Attention Recurrent Residual U-Net) for the recognition of severe convective clouds using the cloud image prediction sequence from FY-4A data. The characteristic parameters used to recognize severe convective clouds in this study were brightness temperature values TBB₉, brightness temperature difference values TBB₉−TBB₁₂ and TBB₁₂−TBB₁₃, and texture features based on spectral characteristics. This method first input five satellite cloud images with a time interval of 30 min into the ARRU-Net model and predicted five satellite cloud images for the next 2.5 h. Then, severe convective clouds were segmented based on the predicted image sequence. The root mean square error (RMSE), peak signal-to-noise ratio (PSNR), and correlation coefficient (R²) of the predicted results were 5.48 K, 35.52 dB, and 0.92, respectively. The results of the experiments showed that the average recognition accuracy and recall of the ARRU-Net model in the next five moments on the test set were 97.62% and 83.34%, respectively. Full article

A squall line is a type of strongly organized mesoscale convective system that can cause severe weather disasters. Thus, it is crucial to explore the dynamic structure and hydrometeor distributions in squall lines. This study analyzed a squall line over Guangdong Province on 6 May 2016 that was observed using a Ka-band millimeter-wave cloud radar (CR) and an S-band dual-polarization radar (PR). Doppler spectral density data obtained by the CR were used to retrieve the vertical air motions and raindrop size distribution (DSD). The results showed the following: First, the CR detected detailed vertical profiles and their evolution before and during the squall line passage. In the convection time segment (segment B), heavy rain existed with a reflectivity factor exceeding 35 dBZ and a velocity spectrum width exceeding 1.3 m s⁻¹. In the PR detection, the differential reflectivity factor (Z_dr) was 1–2 dB, and the large specific differential phase (K_dp) also represented large liquid water content. In the transition and stratiform cloud time segments (segments B and C), the rain stabilized gradually, with decreasing cloud tops, stable precipitation, and a 0 °C layer bright band. Smaller K_dp values (less than 0.9) were distributed around the 0 °C layer, which may have been caused by the melting of ice crystal particles. Second, from the CR-retrieved vertical air velocity, before squall line passage, downdrafts dominated in local convection and weak updrafts existed in higher-altitude altostratus clouds. In segment B, the updraft air velocity reached more than 8 m s⁻¹ below the 0 °C layer. From segments C to D, the updrafts changed gradually into weak and wide-ranging downdrafts. Third, in the comparison of DSD values retrieved at 1.5 km and DSD values on the ground, the retrieved DSD line was lower than the disdrometer, the overall magnitude of the DSD retrieved was smaller, and the difference decreased from segments C to D. The standardized intercept parameter (N_w) and shape parameter (μ) of the DSD retrieved at 1.8 km showed good agreement with the disdrometer results, and the mass-weighted mean diameter (D_m) was smaller than that on the ground, but very close to the PR-retrieved D_m result at 2 km. Therefore, comparing with the DSD retrieved at around 2 km, the overall number concentration remained unchanged and D_m got larger on the ground, possibly reflecting the process of raindrop coalescence. Lastly, the average vertical profiles of several quantities in all segments showed that, first of all, the decrease of N_w and D_m with height in segments C and D was similar, reflecting the collision effect of falling raindrops. The trends were opposite in segment B, indicating that raindrops underwent intense mixing and rapid collision and growth in this segment. Then, PR-retrieved D_m profiles can verify the rationality of the CR-retrieved D_m. Finally, a vertical velocity profile peak generated a larger D_m especially in segments C and D. Full article