基于传统和深度学习技术的黄渤海域大风预报方法研究-海洋预报

基于传统和深度学习技术的黄渤海域大风预报方法研究

单位：1. 天津市西青区气象局, 天津 300380;

2. 天津海洋中心气象台, 天津 300074

分类号：P732.4

出版年·卷·期（页码）：2022·39·第六期（34-43）

摘要：

基于黄渤海域站点风速观测资料以及TIGGE资料，选取欧洲数值预报中心（EC）、中国（CMA）、美国（NCEP）、加拿大（ECCC）4家集合预报产品，在综合评估各家性能的基础上，构建、优化和对比了海上大风集成平均（EM）、动态权重（WEM）、变权偏差订正（BCWEM）3类传统集成方法和长短期记忆神经网络（LSTM）方法。结果表明：LSTM在大风集成预报中性能最优。对于黄渤海域10m风速预报，EC综合表现最好，NCEP在6级及以上大风段优势明显。各家预报误差均具有显著日变化特征，夜间预报能力弱于白天。优化训练期长度和去除表现较差成员可显著改善WEM和BCWEM的大风预报能力。相对EM的预报结果，WEM无明显改进，BCWEM和LSTM则有显著提升，后两者在全风速段和大风风速段上的预报误差均下降10%以上，且在夜间时段更为明显。BCWEM有效订正了EM和WEM方法对弱风速的预报偏差，LSTM则进一步减小了对强风速的预报误差，并提高了对大风站次的命中数和ETS评分。大风个例分析也表明，LSTM有效弥补了传统方法对低涡东移型大风漏报的问题，提升了对冷高压型大风的预报能力，优势明显。

Based on the wind speed observation data of stations in the Yellow Sea and Bohai Sea and the ensemble forecast products of the European Centre for Medium-Range Weather Forecasts (EC), China (CMA), the United States (NCEP) and Canada (ECCC) in the THORPEX Interactive Grand Global Ensemble (TIGGE) data, three traditional integration methods, including ensemble mean (EM), dynamic weight ensemble mean (WEM), bias correction weighted ensemble mean (BCWEM), and the long short-term memory neural network (LSTM) methods are constructed, optimized, and compared on the basis of comprehensive evaluation of the performance of the ensemble forecast products. The results show that LSTM has the best performance in sea gale integrated prediction. For the 10-m wind speed forecast in the Yellow Sea and Bohai Sea, EC has the best comprehensive performance, while NCEP has significant advantages in the gale prediction that is equal or greater than level 6. The diurnal variations of the forecast errors in four products are significant, and the prediction abilities of all products at nighttime are weaker than that at daytime. In the traditional methods, the gale prediction ability of WEM and BCWEM can be significantly improved by optimizing the length of training period and removing the members with poor performance. Compared with EM, WEM shows no significant improvement, while BCWEM and LSTM shows a significant improvement with a decrease in forecast error by more than 10% for both full wind speed and strong wind speed, which is more remarkable at nighttime. BCWEM effectively corrects the prediction bias of the EM and WEM methods for moderate and weak wind speed, while LSTM further reduces the prediction error for strong wind speed and improved the hit number of gale stations and ETS score. The cases analysis of gale also shows that LSTM effectively compensates for the missing report problem of gale in low vortex eastward type by traditional methods, and improves the prediction ability of gale in cold and high pressure type with significant advantage.

参考文献：

[1] 傅赐福, 李涛, 刘仕潮, 等. 1909号台风"利奇马" 引发渤海湾风暴潮特征及无人机灾害调查[J]. 海洋预报, 2021, 38(5):17-23. FU C F, LI T, LIU S C, et al. Characteristics of the storm surge and UAV disaster investigation caused by the typhoon Lekima (No. 1909) in the Bohai Bay[J]. Marine Forecasts, 2021, 38(5):17-23.
[2] LEITH C E. Theoretical skill of Monte Carlo forecasts[J]. Monthly Weather Review, 1974, 102(6):409-418.
[3] LORENZ E N. Atmospheric predictability as revealed by naturally occurring analogues[J]. Journal of the Atmospheric Sciences, 1969, 26(4):636-646.
[4] EPSTEIN E S. A scoring system for probability forecasts of ranked categories[J]. Journal of Applied Meteorology, 1969, 8(6):985-987.
[5] PARK Y Y, BUIZZA R, LEUTBECHER M. TIGGE:preliminary results on comparing and combining ensembles[J]. Quarterly Journal of the Royal Meteorological Society, 2008, 134(637):2029-2050.
[6] KRISHNAMURTI T N, KISHTAWAL C M, LAROW T E, et al. Improved weather and seasonal climate forecasts from multimodel superensemble[J]. Science, 1999, 285(5433):1548-1550.
[7] 智协飞, 黄闻. 基于卡尔曼滤波的中国区域气温和降水的多模式集成预报[J]. 大气科学学报, 2019, 42(2):197-206. ZHI X F, HUANG W. Multimodel ensemble forecasts of surface air temperature and precipitation over China by using Kalman filter[J]. Transactions of Atmospheric Sciences, 2019, 42(2):197-206.
[8] ZHU S P, ZHI X F, GE F, et al. Subseasonal forecast of surface air temperature using superensemble approaches:experiments over Northeast Asia for 2018[J]. Weather and Forecasting, 2021, 36(1):39-51.
[9] 危国飞, 刘会军, 吴启树, 等. 多模式降水分级最优化权重集成预报技术[J]. 应用气象学报, 2020, 31(6):668-680. WEI G F, LIU H J, WU Q S, et al. Multi-model consensus forecasting technology with optimal weight for precipitation intensity levels[J]. Journal of Applied Meteorological Science, 2020, 31(6):668-680.
[10] 智协飞, 赵忱. 基于集合成员订正的强降水多模式集成预报[J]. 应用气象学报, 2020, 31(3):303-314. ZHI X F, ZHAO C. Heavy precipitation forecasts based on multimodel ensemble members[J]. Journal of Applied Meteorological Science, 2020, 31(3):303-314.
[11] 郭蓉, 余晖, 漆梁波, 等. 台风路径多模式集成预报技术研究[J]. 气象科学, 2019, 39(6):839-846. GUO R, YU H, QI L B, et al. A study on multi-model ensemble forecast technique for Typhoon track[J]. Journal of the Meteorological Sciences, 2019, 39(6):839-846.
[12] 高松, 徐江玲, 刘桂艳, 等. 基于机器学习的青岛市区近岸海雾集成预报方法[J]. 海洋科学, 2021, 45(3):33-42. GAO S, XU J L, LIU G Y, et al. Ensemble forecast of sea fog in Qingdao coastal area based on machine learning[J]. Marine Sciences, 2021, 45(3):33-42.
[13] 智协飞, 林春泽, 白永清, 等. 北半球中纬度地区地面气温的超级集合预报[J]. 气象科学, 2009, 29(5):569-574. ZHI X F, LIN C Z, BAI Y Q, et al. Superensemble forecasts of the surface temperature in Northern Hemisphere middle latitudes[J]. Scientia Meteorologica Sinica, 2009, 29(5):569-574.
[14] 张玉涛, 佟华, 孙健. 一种偏差订正方法在平昌冬奥会气象预报的应用[J]. 应用气象学报, 2020, 31(1):27-41. ZHANG Y T, TONG H, SUN J. Application of a bias correction method to meteorological forecast for the Pyeongchang Winter Olympic Games[J]. Journal of Applied Meteorological Science, 2020, 31(1):27-41.
[15] JI L Y, ZHI X F, SIMMER C, et al. Multimodel ensemble forecasts of Precipitation based on an object-based diagnostic evaluation[J]. Monthly Weather Review, 2020, 148(6):2591-2606.
[16] 祁海霞, 彭涛, 林春泽, 等. 清江流域降水的多模式BMA概率预报试验[J]. 气象, 2020, 46(1):108-118. QI H X, PENG T, LIN C Z, et al. Probabilistic forecasting of precipitation over the Qingjiang River basin using BMA multimodel ensemble technique[J]. Meteorological Monthly, 2020, 46(1):108-118.
[17] 门晓磊, 焦瑞莉, 王鼎, 等. 基于机器学习的华北气温多模式集合预报的订正方法[J]. 气候与环境研究, 2019, 24(1):116-124. MEN X L, JIAO R L, WANG D, et al. A temperature correction method for multi-model ensemble forecast in North China based on machine learning[J]. Climatic and Environmental Research, 2019, 24(1):116-124.
[18] HINTON G E, OSINDERO S, TEH Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006, 18(7):1527-1554.
[19] CHENG L L, ZANG H X, DING T, et al. Ensemble recurrent neural network based probabilistic wind speed forecasting approach[J]. Energies, 2018, 11(8):1958.
[20] LI H C, YU C, XIA J J, et al. A model output machine learning method for grid temperature forecasts in the Beijing area[J]. Advances in Atmospheric Sciences, 2019, 36(10):1156-1170.
[21] 刘岳明, 杨晓梅, 王志华, 等. 基于深度学习RCF模型的三都澳筏式养殖区提取研究[J]. 海洋学报, 2019, 41(4):119-130. LIU Y M, YANG X M, WANG Z H, et al. Extracting raft aquaculture areas in Sanduao from high-resolution remote sensing images using RCF[J]. Haiyang Xuebao, 2019, 41(4):119-130.
[22] 孙健, 曹卓, 李恒, 等. 人工智能技术在数值天气预报中的应用[J]. 应用气象学报, 2021, 32(1):1-11. SUN J, CAO Z, LI H, et al. Application of artificial intelligence technology to numerical weather prediction[J]. Journal of Applied Meteorological Science, 2021, 32(1):1-11.
[23] 智协飞, 王田, 季焱. 基于深度学习的中国地面气温的多模式集成预报研究[J]. 大气科学学报, 2020, 43(3):435-446. ZHI X F, WANG T, JI Y. Multimodel ensemble forecasts of surface air temperature over China based on deep learning approach[J]. Transactions of Atmospheric Sciences, 2020, 43(3):435-446.
[24] 王国松, 王喜冬, 侯敏, 等. 基于观测和再分析数据的LSTM深度神经网络沿海风速预报应用研究[J]. 海洋学报, 2020, 42(1):67-77. WANG G S, WANG X D, HOU M, et al. Research on application of LSTM deep neural network on historical observation data and reanalysis data for sea surface wind speed forecasting[J]. Haiyang Xuebao, 2020, 42(1):67-77.
[25] 范书鸣, 储鏖, 蒋勤. 再分析风场修正及其在渤海湾典型温带风暴潮模拟中的应用[J]. 海洋预报, 2021, 38(4):61-68. FAN S M, CHU A, JIANG Q. Optimization of reanalysis wind field and its application in the simulation of extratropical storm surge in the Bohai Bay[J]. Marine Forecasts, 2021, 38(4):61-68.
[26] 许立兵, 王安喜, 汪纯阳, 等. 基于机器学习的海洋环境预报订正方法研究[J]. 海洋通报, 2020, 39(6):695-704. XU L B, WANG A X, WANG C Y, et al. Research on correction method of marine environment prediction based on machine learning[J]. Marine Science Bulletin, 2020, 39(6):695-704.
[27] 司鹏, 梁冬坡, 朱男男, 等. 黄渤海海域16个石油平台站风速资料的初步质量检测[J]. 海洋预报, 2020, 37(1):43-49. SI P, LIANG D P, ZHU N N, et al. Preliminary quality verification of wind speed data observed by 16 oil platform monitoring stations in the Yellow Sea and Bohai Sea[J]. Marine Forecasts, 2020, 37(1):43-49.
[28] 严明良, 缪启龙, 沈树勤. 基于超级集合思想的数值预报产品变权集成方法探讨[J]. 气象, 2009, 35(6):19-25, 129-130. YAN M L, MIAO Q L, SHEN S Q. Exploration on ensemble model of numerical forecasting based on variable-weight superensemble method[J]. Meteorological Monthly, 2009, 35(6):19-25, 129-130.
[29] 智协飞, 李刚, 彭婷. 基于贝叶斯理论的单站地面气温的概率预报研究[J]. 大气科学学报, 2014, 37(6):740-748. ZHI X F, LI G, PENG T. On the probabilistic forecast of 2 meter temperature of a single station based on Bayesian theory[J]. Transactions of Atmospheric Sciences, 2014, 37(6):740-748.
[30] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8):1735-1780.
[31] 陈圣劼, 刘梅, 张涵斌, 等. 集合预报产品在江苏省暴雨预报中的应用评估[J]. 气象, 2019, 45(7):893-907. CHEN S J, LIU M, ZHANG H B, et al. Evaluation on forecasting heavy rainfall over Jiangsu region using ensemble forecast techniques and products[J]. Meteorological Monthly, 2019, 45(7):893-907.

服务与反馈：

【文章下载】【发表评论】【查看评论】【加入收藏】