天气预报中的人工智能技术进展

doi:10.11867/j.issn.1001-8166.2020.053

以数值预报技术为主的天气预报由于存在大气运动规律认识和表达、观测资料同化应用、模式物理参数化等方面的不足，导致误差。人工智能技术基于大数据学习的优势为天气预报的改进和技术革命提供了新的可能。从人工智能的发展背景、人工智能技术在天气预报中的应用现状以及未来的发展趋势等方面进行了重点阐述，提出了人工智能技术与数值预报相融合的天气预报技术发展思路。特别指出，未来的天气预报人工智能算法需要着眼于导致预报不确定性的非线性、混沌性的大气运动特点，不仅要完善基于数据驱动的输入—输出映射的框架，更要从数学与物理学的本质出发，实现数学、物理的混合建模，在实现人工智能理论突破的同时，推动天气预报瓶颈问题的解决。

关键词: 人工智能, 天气预报, 数值预报

Numerical weather prediction, which is the major basis of current weather forecast, has some shortcomings, such as the understanding of the law of atmospheric motion, the assimilation and application of observation data, the expression of model physics, etc., leading to the forecast error of weather. The rapid development of artificial intelligence technology in recent years provides a new possibility for the advancement and innovation of weather forecast. In this paper, the background of the development of artificial intelligence, the current situation of the application of artificial intelligence technology to weather forecast and the future development trend are mainly described to account for this possibility. After that, the idea for development of weather forecast technology based on the integration of artificial intelligence and numerical forecast is put forward. Particularly, this study stresses that, in order to advance the AI algorithm of weather forecast in the future, it is requested to focus on the nonlinear and chaotic characteristics of atmospheric motion leading to the uncertainty of forecast. Starting from the essence of mathematics and physics, we need to realize the hybrid modeling of mathematics and physics, not only to establish the framework of input-output mapping, but also to provide solutions to the bottleneck problems of weather forecast.

Key words: Artificial intelligence, Weather forecast, Numerical weather prediction

中图分类号:

图1 “蓄水池”式的数据输入—输出（Input Layer-Output Layer）在深度学习中的应用^{[

30
]}

Fig.1 Application of reservoir-only prediction network (Input Layer-Output Layer) in Machine Learning^{[

30
]}

图2 覆盖美国大陆逐5分钟、空间分辨率5 km经偏差订正的智能手机地面气压观测^{[

31
]}

Fig.2 Surface pressure in horizontal resolution of 5 km observed by mobile phones over US^{[

31
]}

图3 上海中心气象台基于AI模型订正后的降水预报准确率评分与其他客观预报方法的比较
EC：ECMWF模式；GFS：NCEP/GFS模式；WARMS：华东区域WRF模式；机器学习订正：OBA模型;EC、WARMS频率订正：经频率订正后的模式预报结果

Fig.3 Comparison of accuracy for different objective prediction methods
EC：ECMWF model；GFS：NCEP/GFS model；WARMS：WRF model for East China；Correction with Machine Learning：OBA modual developed by Shanghai Central Meteorological Observatory;EC,WARMS frequency correction：Prediction results from frequency correction

图4 AI灾害天气预报模型优化流程
实线箭头表示主流程，虚线表示可选和循环过程

Fig.4 The optimized AI forecast flow chart for disastrous weather
Solid arrows for the main flow, dashed arrows for the optional and cycled flow

1	Bjerknes V. Dynamic Meteorology and Hydrograph, Part II. Kinematics[M].New York: Camegie Institute, Gibson Bros,1911.
2	Key J, Maslanik J, Schweiger A. Classification of merged AVHRR and SMMR Arctic data with neural networks[J]. Photogrammetric Engineering and Remote Sensing, 1989, 55(9):1 331-1 338.
3	Bankert R. Cloud classification of AVHRR imagery in maritime regions using a probabilistic neural network[J]. Journal of Applied Meteorology, 1994,33(8):909-918.
4	Marzban C, Stumpf G. A neural network for tornado prediction based on Doppler radar-derived attributes[J]. Journal of Applied Meteorology, 1996, 35(5):617-626.
5	Lakshmanan V, Stumpf G, Witt A. A neural network for detecting and diagnosing tornadic circulations using the mesocyclone detection and near storm environment algorithms[C]//21st International Conference on Information Processing Systems, San Diego, American Meteorological Society, J5.2, 2005.
6	Marzban C, Stumpf G. A neural network for damaging wind prediction[J]. Weather and Forecasting, 1998, 13 (1):151-163.
7	Marzban C, Witt A. A Bayesian neural network for severe-hail size prediction[J]. Weather and Forecasting, 2001, 16(5):600-610.
8	Manzato A. Hail in northeast Italy: A neural network ensemble forecast using sounding-derived indices[J]. Weather and Forecasting, 2013, 28(1):3-28.
9	Anagnostou E. A convective/stratiform precipitation classification algorithm for volume scanning weather radar observations[J]. Meteorological Applications, 2004, 11(4):291-300.
10	Lakshmanan V, Karstens C, Krause J, et al. Quality control of weather radar data using polarimetric variables[J]. Journal of Atmospheric and Oceanic Technology, 2014, 31(6):1 234-1 249.
11	Lakshmanan V, Rabin R, DeBrunner V. Identifying and tracking storms in satellite images[C]//Second Artificial Intelligence Conference, Long Beach, CA, American Meteorological Society, 90-95, 2000.
12	Lakshmanan V, Fritz A, Smith T, et al. An automated technique to quality control radar reflectivity data[J]. Journal of Applied Meteorology, 2007, 46(3): 288-305.
13	Newman J, Lakshmanan V, Heinselman P L, et al. Range- correcting azimuthal shear in Doppler radar data[J]. Weather and Forecasting, 2013, 28:194-211.
14	Trafalis T B, Ince H, Richman M B. Tornado detection with support vector machines[C]//International Conference on Computational Science. Berlin, Heidelberg：Springer, 2003: 289-298.
15	Adrianto I, Trafalis T, Lakshmanan V. Support vector machines for spatiotemporal tornado prediction[J]. International Journal of General Systems, 2007, 38(7): 759-776.
16	Dai Lijie, Zhang Changjiang, Xue Licheng, et al. Eyed tropical cyclone intensity objective estimation model based on infrared satellite image and relevance vector machine[J]. Journal of Remote Sensing, 2018, 22(4): 581-590.
	戴李杰,张长江,薛利成, 等. 红外卫星云图和相关向量机的有眼热带气旋客观定强模型[J]. 遥感学报，2018, 22(4):581-590.
17	Chisholm D, Ball J, Veigas K, et al. The diagnosis of upper-level humidity[J]. Journal of Applied Meteorology, 1968, 7(4):613-619.
18	Williams J D, Dettling A S, Steiner M. Combining observations and model data for short-term storm forecasting[C]//Remote Sensing Applications for Aviation Weather Hazard Detection and Decision Support, SanDiego, CA, International Society for Optics and Photonics, 2008.
19	Williams J, Sharman R, Craig J, et al. Remote detection and diagnosis of thunderstorm turbulence[C]//Remote Sensing Applications for Aviation Weather Hazard Detection and Decision Support, San Diego, CA, International Society for Optics and Photonics. 2008.
20	Gagne II D J, McGovern A, Brotzge J. Classification of convective areas using decision trees[J]. Journal of Atmospheric and Oceanic Technology, 2009, 26(7): 1 341-1 353.
21	McGovern A, Gagne II D J, Williams J K, et al. Enhancing understanding and improving prediction of severe weather through spatiotemporal relational learning[J]. Machine Learning, 2014, 95(1): 27-50.
22	McGovern A, Gagne II D J, Basara J. Solar energy prediction: An international contest to initiate interdisciplinary research on compelling meteorological problems[J]. Bulletin of the American Meteorological Society, 2015, 96(8):1 388-1 395.
23	Elmore K L, Grams H. Using mping data to generate random forests for precipitation type forecasts[C]//14th Conference on Artificial and Computational Intelligence and its Applications to the Environmental Sciences, 4.2, 2016.
24	Breiman L. Random forests[J]. Machine Learning, 2001, 45(1):5-32.
25	Friedman J H. Stochastic gradient boosting[J]. Computational Statistics and Data Analysis, 2002, 38(4): 367-378.
26	Schapire R E. The boosting approach to machine learning: An overview［M］//Denison D D，Hansen M H，Holmes C，et al.Nonlinear Estimation and Classification. New York： Springer, 2003.
27	Baldwin M, Kain J, Lakshmivarahan S. Development of an automated classification procedure for rainfall systems[J]. Monthly Weather Review, 2005, 133(4):844-862.
28	Lakshmanan V, Smith T. Evaluating a storm tracking algorithm[C]//26th Conference on Interactive Information Processing Systems. Atlanta, GA: American Meteorological Society, 2010.
29	Miller M L, Lakshmanan V, Smith T M. An automated method for depicting meso-cyclone paths and intensities[J]. Weather and Forecasting, 2013, 28: 570-585.
30	Pathak J, Wikner A, Fussell R, et al. Hybrid forecasting of chaotic processes: Using machine learning in conjunction with a knowledge-based model[J]. Chaos: An Interdisciplinary Journal of Nonlinear Science, 2018, 28(4): 041101.
31	McNicholas C. Smartphone pressure analysis with machine learning and kriging[C]//Oral Presentation， 1B.2. The 100th American Meteorological Society Annual Meeting, Boston, MA, USA, 2020.
32	Chakraborty T. Creating bias-corrected global radiation datasets from climate reanalysis products using supervised learning[C]//Oral Presentation, 1A.3. The 100th American Meteorological Society Annual Meeting, Boston, MA, USA, 2020.
33	Lakshmanan V, Herzog B, Kingfield D. A method for extracting postevent storm tracks[J]. Journal of Applied Meteorology and Climatology, 2015, 54(2): 451-462.
34	Adams-Selin R D, Ziegler C L. Forecasting hail using a one-dimensional hail growth model within WRF[J]. Monthly Weather Review, 2016, 144: 4 919-4 939.
35	Sobash R A, Schwartz C S, Romine G S, et al. Severe weather prediction using storm surrogates from an Ensemble Forecasting System[J]. Weather Forecasting, 2016, 31(1): 255-271.
36	Smith T M, Lakshmanan V, Stumpf G J, et al. Multi-radar multi-sensor (mrms) severe weather and aviation products: Initial operating capabilities[J]. Bulletin of the American Meteorological Society, 2016, 97(9):1 617-1 630.
37	Witt A, Eilts M, Stumph G, et al. An enhanced hail detection algorithm for the WSR-88D[J]. Weather Forecasting, 1998, 13: 286-303.
38	Clark A J, MacKenzie A, McGovern A, et al. An automated, multi-parameter dryline identification algorithm[J]. Weather and Forecasting, 2015, 30(6)： 1 781-1 794.
39	Ma Leiming, Bao Xuwei, Research progress on physical parameterization schemes in numerical weather prediction models[J]. Advances in Earth Science, 2017, 32(7): 679-687.
	马雷鸣，鲍旭炜. 数值天气预报模式物理过程参数化方案的研究进展[J]. 地球科学进展，2017, 32(7): 679-687.
40	Ma L M，Tan Z M. Improving the behavior of the cumulus parameterization for tropical cyclone prediction convection trigger[J]. Atmospheric Research， 2009， 92(2)： 190-211.
41	Ma Leiming. Research progress on major structures of tropical cyclone boundary layer[J]. Progress in Geophysics, 2013, 28(3):1 259-1 268.
	马雷鸣.热带气旋边界层关键结构研究进展[J]. 地球物理学进展, 2013, 28(3): 1 259-1 268.
42	Ma L M, Bao X W. Parametrization of planetary boundary-layer height with helicity and verification with tropical cyclone prediction[J]. Boundary Layer Meteorology, 2016, 160(3):569-593.
43	Glahn H R, Lowry D A. The use of model output statistics (mos) in objective weather forecasting[J]. Journal of Applied Meteorology,1972, 11(8): 1 203-1 211.
44	Ma L M. Convection forecast enhanced by the deep learning of radar observation and numerical prediction[C]//Oral Presentation, 1B.1. The 100th American Meteorological Society Annual Meeting, Boston, MA, USA, 2020.

[1]	端义宏. 登陆台风精细结构的观测、预报与影响评估[J]. 地球科学进展, 2015, 30(8): 847-854.
[2]	白晓平,李红,方栋,胡非,FrancescaCostabile,王芬娟. 资料同化在空气质量预报中的应用[J]. 地球科学进展, 2007, 22(1): 66-73.
[3]	胡永云. 关于平流层异常影响对流层天气系统的研究进展[J]. 地球科学进展, 2006, 21(7): 713-720.
[4]	李耀辉;;赵建华;薛纪善;陈德辉;沈学顺;王红;陈勇. 基于GRAPES的西北地区沙尘暴数值预报模式及其应用研究[J]. 地球科学进展, 2005, 20(9): 999-1011.
[5]	宋淑丽;朱文耀;廖新浩. 地基GPS气象学研究的主要问题及最新进展[J]. 地球科学进展, 2004, 19(2): 250-259.
[6]	陈俊勇. 大地测量学的新进展——国际大地测量协会（IAG）2001年科学大会综述[J]. 地球科学进展, 2002, 17(3): 305-310.
[7]	谢志辉，丑纪范. 大气动力学方程组全局分析的研究进展[J]. 地球科学进展, 1999, 14(2): 133-139.
[8]	钱正安,焦彦军. 青藏高原气象学的研究进展和问题[J]. 地球科学进展, 1997, 12(3): 207-216.
[9]	肖义越; 戴桦; 陈辉. 专家系统在地球科学中的应用[J]. 地球科学进展, 1993, 8(1): 49-54.

阅读次数

全文

摘要

Development of Artificial Intelligence Technology in Weather Forecast