Objective Warning Signal Generation Method for Thunderstorm Gale in Jiangsu and Its Application for the 2023 Rainy Season
-
摘要: 为了实现江苏地区雷暴大风智能预警信号的自动生成以提升雷暴大风临近预报预警能力，在建立分钟级、公里级网格风场数据集的基础上，区分雷暴大风、系统性大风和混合性大风过程，并结合生成对抗网络（Generative Adversarial Network, GAN，针对雷暴大风建模）和具有物理约束架构的时空卷积神经网络（Physical Dynamics Network, PhyDNet，针对系统性大风和混合性大风建模），发展了基于深度学习的江苏地区0~2 h雷暴大风临近预报模型（Blending），实现面向区县的客观、分级预警信号自动生成技术，并对比探究了Blending和PhyDNet_ALL（不区分大风类型，直接采用PhyDNet建模）两种深度学习方法分别生成的客观预警信号与主观预警信号在2023年汛期的表现。结果表明：（1）相较于主观预警信号，由深度学习方法生成的客观预警信号能有效提升预警信号提前量；（2）深度学习方法可以提前预报出强对流大风天气及其演变过程；（3）由于Blending针对雷暴大风单独建模，确保对流系统强度以及中小尺度特征的维持，因此Blending可以更好地描述极端雷暴大风的演变特征，在橙色和红色预警信号提前量上显著高于PhyDNet_ALL。Abstract: To achieve the automatic generation of objective warning signals for thunderstorm gales in Jiangsu and to enhance nowcasting capabilities, a minute-scale and kilometer-scale wind field gridded dataset was established. This dataset distinguishes between different types of wind. By integrating a generative adversarial network for thunderstorm gale modeling and a PhyDNet for the modeling of wind associated with weather systems and mixed-type wind, we developed a deep-learning-based 0-2 hour nowcasting model (Blending) for thunderstorm gales in Jiangsu. Then, we compared the subjective and objective warning signals generated by the PhyDNet_ALL (which uses PhyDNet modeling without distinguishing wind types) and Blending for the 2023 rainy season. The results show that: (1) Compared to subjective warning signals, objective warning signals generated by deep learning methods effectively improved the lead time of warning signals. (2) Deep learning methods can predict thunderstorm gales and their evolution process in advance. (3) Blending, which models convective gales separately, ensures the maintenance of convection intensity and small-scale features, allowing it to better describe the evolution characteristics of extreme convective gales and significantly outperform PhyDNet_ALL in terms of improving the lead time of orange and red alerts.
-
Key words:
- thunderstorm gale /
- nowcasting /
- deep learning /
- warning signal
-
图 2 江苏各市县2023年4—9月Blending和PhyDNet_ALL方法生成的客观预警信号以及主观预警信号中黄色（第一行）、橙色（第二行）和红色（第三行）预警信号平均提前量。
图 4 2023年6月10日淮安雷暴大风过程（a. 观测，b. PhyDNet_ALL，c. Blending；a1—c4的下标数字1、2、3、4分别代表时间15:54、16:06、16:18、16:30）。两组深度学习试验的起报时间为15:48。
图 5 2023年6月10日南通雷暴大风过程（a. 观测，b. PhyDNet_ALL，c. Blending；a1—c4的下标数字1、2、3、4分别代表时间15:42、16:00、16:18、16:30）。两组深度学习试验的起报时间为15: 36。
图 6 2023年8月6日飑线过程（a. 观测，b. PhyDNet_ALL，c. Blending；a1—c4的下标数字1、2、3、4分别代表时间10:54、12:06、12:18、12:30）。两组深度学习试验的起报时间为10: 30。
表 1 联合加权MAE损失中各因子的权重设置
最大风速（S, m·s-1） 权重 组合反射率因子（CR/dBZ） 权重 闪电密度（L/6 min×5 km） 权重 S<5.5 0.5 CR<15 0.5 L<1 0.1 5.5≤S<8.0 1 15≤CR<25 1 1≤L<5 0.2 8.0≤S<13.9 2 25≤CR<35 2.5 5≤L<10 0.5 13.9≤S<17.2 10 35≤CR<45 5 10≤L<25 1 17.2≤S<20.8 20 45≤CR<50 10 L≥25 2 S≥20.8 30 CR≥50 15 
下载: 导出CSV
表 2 2023年6月10日盱眙县预警信号发布情况
盱眙县 黄色 黄色 红色 实况 15: 13 15: 25 16: 13 主观 14: 08（提前65 min） 14: 27（提前58 min） 漏报 PhyDNet_ALL 13: 56（提前79 min） 14: 18（提前91 min） 漏报 Blending 13: 48（提前85 min） 14: 18（提前91 min） 15: 48（提前25 min）
下载: 导出CSV
表 3 2023年6月10日海门区预警信号发布情况
海门区 黄色 橙色 红色 实况 15:40 15:58 16:29 主观 15: 17（提前23 min） 16: 12（滞后14 min） 漏报 PhyDNet_ALL 14: 27（提前73 min） 16: 00（滞后2 min） 漏报 Blending 14: 18（提前82 min） 14: 36（提前82 min） 15: 36（提前53 min）
下载: 导出CSV
表 4 2023年8月6日宝应县预警信号发布情况
宝应县 黄色 橙色 红色 实况 11:38 12:13 —— 主观 漏报 12:36（滞后23 min） —— PhyDNet_ALL 9:48（提前110 min） 漏报 —— Blending 9:42（提前116 min） 10:30（提前103 min） ——
下载: 导出CSV
-
[1] 吴海英, 孙继松, 慕瑞琪, 等. 一次强对流过程中两种不同类型风暴导致的极端对流大风分析[J]. 气象学报, 2023, 81(2): 205-217. [2] 郑永光, 周康辉, 盛杰, 等. 强对流天气监测预报预警技术进展[J]. 应用气象学报, 2015, 26(6): 641-657. [3] HAIDEN T, KANN A, WITTMANN C, et al. The integrated nowcasting through comprehensive analysis (INCA) system and its validation over the eastern Alpine region[J]. Wea Forecasting, 2011, 26(2): 166-183. [4] SEITY Y, BROUSSEAU P, MALARDEL S, et al. The AROME-France convective-scale operational model[J]. Mon Wea Rev, 2011, 139(3): 976-991. [5] TANG Y, LEAN H W, BORNEMANN J. The benefits of the Met Office variable resolution NWP model for forecasting convection[J]. Meteorol Appl, 2013, 20(4): 417-426. [6] ZÄNGL G, REINERT D, RIPODAS P, et al. The ICON (ICOsahedral Non-hydrostatic) modelling framework of DWD and MPI ‐ M: Description of the non‐hydrostatic dynamical core[J]. Quart J Roy Meteor Soc, 2015, 141(687): 563-579. [7] MA Z, HAN W, ZHAO C, et al. A case study of evaluating the GRAPES_Meso V5. 0 forecasting performance utilizing observations from South China Sea Experiment 2020 of the"Petrel Project"[J]. Atmos Res, 2022, 280: 106437. [8] 俞小鼎, 周小刚, 王秀明. 雷暴与强对流临近天气预报技术进展[J]. 气象学报, 2012, 70(3): 311-337. [9] 庄潇然, 郑玉, 王亚强, 等. 基于深度学习的融合降水临近预报方法及其在中国东部地区的应用研究[J]. 气象学报, 2023, 81(2): 286-303. [10] MCGOVERN A, ELMORE K L, GAGNE II D J, et al. Using artificial intelligence to improve realtime decision-making for high-impact weather[J]. Bull Amer Meteor Soc, 2017, 98(10): 2 073-2 090. [11] AYZEL G, HEISTERMANN M, SOROKIN A, et al. All convolutional neural networks for radar-based precipitation nowcasting[J]. Procedia Comput Sci, 2019, 150: 186-192. [12] 陈元昭, 林良勋, 王蕊, 等. 基于生成对抗网络GAN的人工智能临近预报方法研究[J]. 大气科学学报, 2019, 42(2): 311-320. [13] 郭瀚阳, 陈明轩, 韩雷, 等. 基于深度学习的强对流高分辨率临近预报试验[J]. 气象学报, 2019, 77(4): 715-727. [14] 杨绚, 代刊, 朱跃建. 深度学习技术在智能网格天气预报中的应用进展与挑战[J]. 气象学报, 2022, 80(5): 649-667. [15] REICHSTEIN M, CAMPS-VALLS G, STEVENS B, et al. Deep learning and process understanding for datadriven Earth system science[J]. Nature, 2019, 566(7743): 195-204. [16] SHI X J, CHEN Z R, WANG H, et al. Convolutional LSTM network: A machine learning approach for precipitation nowcasting[C]//Proceedings of the 28th International Conference on Neural Information Processing Systems. Montreal: MIT Press, 2015: 802-810. [17] WANG Y B, LONG M S, WANG J M, et al. PredRNN: Recurrent neural networks for predictive learning using spatiotemporal LSTMs[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach: Curran Associates Inc., 2017: 879-888. [18] WANG Y B, GAO Z F, LONG M S, et al. PredRNN++: Towards a resolution of the deep-in-time dilemma in spatiotemporal predictive learning[C]∥Proceedings of the 35th International Conference on Machine Learning. Stockholm: PLMR, 2018: 5 110-5 119. [19] LE GUEN V, THOME N. Disentangling physical dynamics from unknown factors for unsupervised video prediction[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle: IEEE, 2020: 11 471-11 481. [20] RONNEBERGER O, FISCHER P, BROX T. U-Net: Convolutional networks for biomedical image segmentation[C]//Proceedings of the 18th International Conference on Medical Image Computing and Computer-Assisted Intervention. Munich: Springer, 2015: 234-241. [21] ZHOU K H, ZHENG Y G, DONG W S, et al. A deep learning network for cloud-to-ground lightning nowcasting with multisource data[J]. J Atmos Ocean Technol, 2020, 37(5): 927-942. [22] TREBING K, STAǸCZYK T, MEHRKANOON S. SmaAt-UNet: Precipitation nowcasting using a small attentionUNet architecture[J]. Pattern Recognition Letters, 2021, 145: 178-186. [23] HU Y, CHEN L, WANG Z B, et al. Towards a more realistic and detailed deep-learning-based radar echo extrapolation method[J]. Remote Sens, 2022: 14(1): 24. [24] 张延彪, 宋林烨, 陈明轩, 等. 基于卷积神经网络的京津冀地区高分辨率格点预报偏差订正试验[J]. 大气科学学报, 2022, 45(6): 850-862. [25] 马志峰, 张浩, 刘劼. 基于深度学习的短临降水预报综述[J]. 计算机工程与科学, 2023, 45(10): 1 731-1 753. [26] HAN L, LIANG H, CHEN H, et al. Convective precipitation nowcasting using U-Net Model[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 60: 1-8. [27] LI Y, LIU Y B, SUN R F. et al. Convective storm VIL and lightning nowcasting using satellite and weather radar measurements based on multi-task learning models[J]. Adv Atmos Sci, 2023, 40: 887-899. [28] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[J]. Advances in neural information processing systems, 2017, 30. [29] GAO Z, SHI X, WANG H, et al. Earthformer: Exploring space-time transformers for earth system forecasting[J]. Advances in Neural Information Processing Systems, 2022, 35: 25 390-25 403. [30] GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial networks[J]. Communications of the ACM, 2020, 63 (11): 139-144. [31] RAVURI S, LENC K, WILLSON M, et al. Skilful precipitation nowcasting using deep generative models of radar[J]. Nature, 2021, 597 (7878): 672-677. [32] HESS P, DRÜKE M, PETRI S, et al. Physically constrained generative adversarial networks for improving precipitation fields from Earth system models[J]. Nature Machine Intelligence, 2022, 4(10): 828-839. [33] 肖海霞, 张峰, 王亚强, 等. 基于生成对抗网络和卫星数据的云图临近预报[J]. 应用气象学报, 2023, 34(2): 220-233. [34] 曾强胜, 郭敬天, 任鹏, 等. 基于分层生成对抗网络的短临降水预报方法研究[J]. 中国海洋大学学报(自然科学版), 2024, 54(2): 23-32. [35] WANG C, TANG G, GENTINE P. PrecipGAN: Merging microwave and infrared data for satellite precipitation estimation using generative adversarial network[J]. Geophysical Research Letters, 2021, 48(5): e2020GL092032. [36] PRUDDEN R, ADAMS S, KANGIN D, et al. A review of radar-based nowcasting of precipitation and applicable machine learning techniques[J]. 2020, arXiv preprint arXiv: 2005.04988. [37] 李显风, 师春香, 胡佳军, 等. CLDAS数据质量在线评估系统的设计与实现[J]. 气象科技, 2017, 45(6): 1 116-1 124. [38] 张涛, 苗春生, 王新. LAPS与STMAS地面气温融合效果对比试验[J]. 高原气象, 2014, 33(03): 743-752. [39] GODDEN, D AND LURMANN F. Development of the PLMSTAR model and its application to ozone episode conditions in the South Coast Air Basin[R]. Environmental Research and Tecnology, Inc., Westlake Village, CA, 1983. [40] XIAO H, WANG Y, ZHENG Y, et al. Convetive-gust nowcasting based on radar reflectivity and a deep learning algorithm[J]. Geosci Model Dev, 2023, 16(12): 3 611-3 628. -
点击查看大图
图(6) / 表(4)
计量
- 文章访问数: 12
- HTML全文浏览量: 4
- PDF下载量: 3
- 被引次数: 0
粤公网安备 4401069904700003号