设为首页
加入收藏
Short-term Precipitation Correction Based on GRU Deep Learning
-
摘要: 为提高短时临近降水预报准确率, 提出一种订正广西对流尺度数值预报模式(GRAPES-GX)降水预报产品的深度学习方法。该方法通过神经网络对实况进行时空特征提取, 以门控循环网络(GRU)为基础框架, 针对降水产品进行改进, 并用于GRAPES-GX降水预报产品订正。在此基础上, 设计了大气物理规律适配模块, 通过物理条件匹配机制订正模式预报降水强度与落区的系统性误差, 增强训练样本中预报产品和实况的特征相关性, 并协同优化模型参数, 获得更优的订正效果。广西区域试验结果表明: 订正模型在各预报时效、各降水强度等级的TS(threat score)评分均得到正技巧, 总体TS技巧评分为2.21%。对于不低于0.1 mm·h-1、不低于2 mm·h-1、不低于7 mm·h-1、不低于15 mm·h-1、不低于25 mm·h-1和不低于40 mm·h-1降水强度预报TS技巧评分分别为5.67%、3.59%、2.18%、1.46%、1.01%和0.46%。0~2 h、2~4 h和4~6 h时效预报TS技巧评分分别为4.77%、1.28%和0.91%。Abstract: To improve the accuracy of short-term precipitation forecasts, a deep learning method is proposed to correct numerical model precipitation forecast products. This method extracts spatiotemporal features from numerical model forecasts and observations using a neural network and performs corrections based on a gated recurrent unit (GRU) framework. Additionally, an atmospheric physics adaptor module is meticulously designed to address systematic errors in the intensity and displacement of numerical model forecast by leveraging physical condition mechanisms. The module plays a crucial role within the overarching model framework, which consists of three integrated components: Feature network, recurrent-revising network, and physical adaptor. The feature network extracts precipitation intensity, distribution, motion characteristics and other related atmospheric physical features from precipitation in situ and numerical model forecast data, serving as input to the recurrent-revising network. Recurrent-revising network utilizes a recurrent neural network structure to adjust grid point forecast results on a time-step basis. Deep neural networks are used to extract spatiotemporal variation features from numerical model forecast data and historical observations, learning systematic errors in the evolution processes to correct the precipitation magnitude and distribution. The physical adaptor is an atmospheric physics adaptation module, which preprocesses numerical model forecast data using frequency distribution fitting and distribution deviation correction methods. In Guangxi convective-scale model precipitation forecast data, when there are significant differences between numerous samples and the precipitation in situ, the feature correlation is low, making it a challenge to capture systematic error characteristics during neural network training. By preprocessing the samples with the physical adaptor, differences between forecasts and observations are reduced, enhancing feature correlation between training input datasets and observations, thus facilitating better neural network training and achieving superior correction skills. This method not only adheres to but also integrates fundamental atmospheric physical laws governing precipitation evolution, thereby offering a robust and innovative approach for post-processing numerical model short-term precipitation forecasts. By incorporating these physical principles into the model framework, corrected forecasts not only reflect statistical patterns but also adhere closely to the underlying physical processes driving precipitation dynamics. Experimental results in Guangxi indicate that the model demonstrates positive threat score skills across various forecast times and precipitation intensities. Specifically, for different precipitation intensities (average of all times), threat score skills for 0.1 mm·h-1, 2 mm·h-1, 7 mm·h-1, 15 mm·h-1, 25 mm·h-1, and 40 mm·h-1 are 5.67%, 3.59%, 2.18%, 1.46%, 1.01%, and 0.46%, respectively; for different lead times, threat score skills for 0-2 h, 2-4 h, and 4-6 h are 4.77%, 1.28%, and 0.91%, respectively; and the overall average threat score skill across all precipitation intensities and times is 2.21%.
-
Key words:
- forecast correction;
- deep learning;
- GRU
-
图 4 2023年8月27日07：36起报的不同时效降水强度预报与订正结果对比
Fig. 4 Comparison of precipitation intensity before and after correction for different lead times initiated at 0736 BT 27 Aug 2023
图 4 年8月27日07：36起报的不同时效降水强度预报与订正结果对比
Fig. 4 Comparison of precipitation intensity before and after correction for different lead times initiated at 0736 BT 27 Aug 2023
图 4 年8月27日07：36起报的不同时效降水强度预报与订正结果对比
Fig. 4 Comparison of precipitation intensity before and after correction for different lead times initiated at 0736 BT 27 Aug 2023
图 5 2023年9月11日02：24起报的不同时效降水强度预报与订正对比
Fig. 5 Comparison of precipitation intensity before and after correction for different lead times initiated at 0224 BT 11 Sep 2023
图 5 年9月11日02：24起报的不同时效降水强度预报与订正对比
Fig. 5 Comparison of precipitation intensity before and after correction for different lead times initiated at 0224 BT 11 Sep 2023
图 5 年9月11日02：24起报的不同时效降水强度预报与订正对比
Fig. 5 Comparison of precipitation intensity before and after correction for different lead times initiated at 0224 BT 11 Sep 2023
图 6 年7月17日11:00降水强度预报订正对比
Fig. 6 Comparison of precipitation intensity before and after correction at 1100 BT 17 Jul 2023
图 6 年7月17日11:00降水强度预报订正对比
Fig. 6 Comparison of precipitation intensity before and after correction at 1100 BT 17 Jul 2023
图 6 2023年7月17日11:00降水强度预报订正对比
Fig. 6 Comparison of precipitation intensity before and after correction at 1100 BT 17 Jul 2023
图 7 2023年7月18日08:00不同时效降水落区预报订正对比
Fig. 7 Comparison of precipitation coverage area before and after correction for different lead times at 0800 BT 18 Jul 2023
图 7 年7月18日08:00不同时效降水落区预报订正对比
Fig. 7 Comparison of precipitation coverage area before and after correction for different lead times at 0800 BT 18 Jul 2023
-
[1] 张博, 张芳华, 李晓兰, 等. "23·7" 华北特大暴雨数值预报检验评估. 应用气象学报, 2024, 35(1): 17-32. doi: 10.11898/1001-7313.20240102Zhang B, Zhang F H, Li X L, et al. Verification and assessment of "23·7" severe rainstorm numerical prediction in North China. J Appl Meteor Sci, 2024, 35(1): 17-32. doi: 10.11898/1001-7313.20240102 [2] 邢楠, 仲跻芹, 雷蕾, 等. 基于CMA-BJ的北京地区短时强降水预报试验. 应用气象学报, 2023, 34(6): 641-654. doi: 10.11898/1001-7313.20230601Xing N, Zhong J Q, Lei L, et al. A probabilistic forecast experiment of short-duration heavy rainfall in Beijing based on CMA-BJ. J Appl Meteor Sci, 2023, 34(6): 641-654. doi: 10.11898/1001-7313.20230601 [3] 黄丽萍, 邓莲堂, 王瑞春, 等. CMA-MESO关键技术集成及应用. 应用气象学报, 2022, 33(6): 641-654. doi: 10.11898/1001-7313.20220601Huang L P, Deng L T, Wang R C, et al. Key technologies of CMA-MESO and application to operational forecast. J Appl Meteor Sci, 2022, 33(6): 641-654. doi: 10.11898/1001-7313.20220601 [4] 许晨璐, 王建捷, 黄丽萍. 千米尺度分辨率下GRAPES-Meso4.0模式定量降水预报性能评估. 气象学报, 2017, 75(6): 851-876.Xu C L, Wang J J, Huang L P. Evaluation on QPF of GRAPES-Meso4.0 model at convection-permitting resolution. Acta Meteor Sinica, 2017, 75(6): 851-876. [5] 谢漪云, 王建捷. GRAPES千米尺度模式在西南复杂地区降水预报偏差与成因初探. 气象学报, 2021, 79(5): 732-749.Xie Y Y, Wang J J. Preliminary study on the deviation and cause of precipitation prediction of GRAPES kilometer scale model in southwest complex terrain area. Acta Meteor Sinica, 2021, 79(5): 732-749. [6] 王伯民, 刘小宁. 中国云的分布. 北京: 气象出版社, 2009.Wang B M, Liu X N. Distribution of China Cloud. Beijing: China Meteorological Press, 2009. [7] 莫建飞, 钟仕全, 陈燕丽, 等. 极端降水事件下广西流域洪涝社会经济暴露度分析. 灾害学, 2018, 33(2): 83-88. doi: 10.3969/j.issn.1000-811X.2018.02.016Mo J F, Zhong S Q, Chen Y L, et al. Study on social-economic exposure degree model of basin flood hazard of extreme precipitation events in Guangxi. J Catastrophology, 2018, 33(2): 83-88. doi: 10.3969/j.issn.1000-811X.2018.02.016 [8] 林振敏, 黄荣, 戚云枫, 等. 广西对流尺度数值天气预报模式系统建设与效益评估. 气象研究与应用, 2022, 43(2): 105-110.Lin Z M, Huang R, Qi Y F, et al. Construction and benefit evaluation of convective scale numerical weather prediction model system in Guangxi. J Meteor Res Appl, 2022, 43(2): 105-110. [9] Boeing G. Visual analysis of nonlinear dynamical systems: Chaos, fractals, self-similarity and the limits of prediction. Systems, 2016, 4(4). DOI: 10.3390/systems4040037. [10] 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. 气象学报, 2022, 80(5): 649-667.Yang X, Dai K, Zhu Y J. Progress and challenges of deep learning techniques in intelligent grid weather forecasting. Acta Meteor Sinica, 2022, 80(5): 649-667. [11] 智协飞, 赵忱. 基于集合成员订正的强降水多模式集成预报. 应用气象学报, 2020, 31(3): 303-314. doi: 10.11898/1001-7313.20200305Zhi X F, Zhao C. Heavy precipitation forecasts based on multi-model ensemble members. J Appl Meteor Sci, 2020, 31(3): 303-314. doi: 10.11898/1001-7313.20200305 [12] 王娟怀, 李清泉, 汪方, 等. 基于DERF2.0的华南前汛期降水订正. 应用气象学报, 2021, 32(1): 115-128. doi: 10.11898/1001-7313.20210110Wang J H, Li Q Q, Wang F, et al. Correction of precipitation forecast predicted by DERF2.0 during the pre-flood season in South China. J Appl Meteor Sci, 2021, 32(1): 115-128. doi: 10.11898/1001-7313.20210110 [13] 谢舜, 孙效功, 张苏平, 等. 基于SVD与机器学习的华南降水预报订正方法. 应用气象学报, 2022, 33(3): 293-304. doi: 10.11898/1001-7313.20220304Xie S, Sun X G, Zhang S P, et al. Precipitation forecast correction in South China based on SVD and machine learning. J Appl Meteor Sci, 2022, 33(3): 293-304. doi: 10.11898/1001-7313.20220304 [14] 吴启树, 韩美, 刘铭, 等. 基于评分最优化的模式降水预报订正算法对比. 应用气象学报, 2017, 28(3): 306-317. doi: 10.11898/1001-7313.20170305Wu Q S, Han M, Liu M, et al. A comparison of optimal-score-based correction algorithms of model precipitation prediction. J Appl Meteor Sci, 2017, 28(3): 306-317. doi: 10.11898/1001-7313.20170305 [15] 李丹, 林文, 刘群, 等. 机器学习在人工增雨效果统计检验中的应用. 应用气象学报, 2024, 35(1): 118-128. doi: 10.11898/1001-7313.20240110Li D, Lin W, Liu Q, et al. Application of machine learning to statistical evaluation of artificial rainfall enhancement. J Appl Meteor Sci, 2024, 35(1): 118-128. doi: 10.11898/1001-7313.20240110 [16] Zhang Y C, Long M S, Chen K Y, et al. Skilful nowcasting of extreme precipitation with NowcastNet. Nature, 2023, 619(7970): 526-532. doi: 10.1038/s41586-023-06184-4 [17] Sayeed A, Choi Y, Jung J, et al. A deep convolutional neural network model for improving WRF simulations. IEEE Trans Neural Netw Learn Syst, 2023, 34(2): 750-760. doi: 10.1109/TNNLS.2021.3100902 [18] Zhang C J, Zeng J, Wang H Y, et al. Correction model for rainfall forecasts using the LSTM with multiple meteorological factors. Meteor Appl, 2020, 27(1). DOI: 10.1002/met.1852. [19] Han L, Chen M X, Chen K K, et al. A deep learning method for bias correction of ECMWF 24-240 h forecasts. Adv Atmos Sci, 2021, 38(9): 1444-1459. doi: 10.1007/s00376-021-0215-y [20] Hochreiter S, Schmidhuber J. Long short-term memory. Neural Comput, 1997, 9: 1735-1780. doi: 10.1162/neco.1997.9.8.1735 [21] 米前川, 高西宁, 李玥, 等. 深度学习方法在干旱预测中的应用. 应用气象学报, 2022, 33(1): 104-114. doi: 10.11898/1001-7313.20220109Mi Q C, Gao X N, Li Y, et al. Application of deep learning method to drought prediction. J Appl Meteor Sci, 2022, 33(1): 104-114. doi: 10.11898/1001-7313.20220109 [22] Chung J, Gülçehre Ç, Cho K H, et al. Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling. Eprint ArXiv, 2014. DOI: 10.48550/arXiv.1412.3555. [23] 卓健, 廖胜石, 苏传程, 等. MLP在雷达定量降水估测中的应用. 热带气象学报, 2023, 39(3): 289-299.Zhuo J, Liao S S, Su C C, et al. Application of MLP in radar quantitative precipitation estimation. J Trop Meteor, 2023, 39(3): 289-299. [24] 卓健, 陈少斌, 周冬静, 等. 快速动态分级法进行雷达定量估测降水方法研究. 热带气象学报, 2018, 34(6): 856-864.Zhuo J, Chen S B, Zhou D J, et al. An approach for radar quantitative precipitation estimate based on fast dynamic categorical method. J Trop Meteor, 2018, 34(6): 856-864. [25] 胡莹莹, 庞林, 王启光. 基于深度学习的7~15 d温度格点预报偏差订正. 应用气象学报, 2023, 34(4): 426-437. doi: 10.11898/1001-7313.20230404Hu Y Y, Pang L, Wang Q G. Application of deep learning bias correction method to temperature grid forecast of 7-15 days. J Appl Meteor Sci, 2023, 34(4): 426-437. doi: 10.11898/1001-7313.20230404 [26] 张林, 吴蕾, 李峰, 等. 基于深度学习的天气雷达异常数据识别技术. 应用气象学报, 2023, 34(6): 694-705. doi: 10.11898/1001-7313.20230605Zhang L, Wu L, Li F, et al. Indentification of weather radar abnormal data based on deep learning. J Appl Meteor Sci, 2023, 34(6): 694-705. doi: 10.11898/1001-7313.20230605 [27] Shi W Z, Caballero J, Huszár F, et al. Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network. 2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2016: 1874-1883. [28] 张秀芝. Weibull分布参数估计方法及其应用. 气象学报, 1996, 54(1): 108-116.Zhang X Z. Parameter estimation method of Weibull distribution and its application. Acta Meteor Sinica, 1996, 54(1): 108-116. [29] Laroche S, Zawadzki I. A variational analysis method for retrieval of three-dimensional wind field from single-Doppler radar data. J Atmos Sci, 1994, 51(18): 2664-2682. [30] Wilks D S. Statistical Methods in the Atmospheric Sciences. New York: Academic Press, 2019: 373-374. -
下载:
计量
- 摘要浏览量: 825
- HTML全文浏览量: 97
- PDF下载量: 145
- 被引次数: 0
