Application of Deep Learning Bias Correction Method to Temperature Grid Forecast of 7-15 Days

Hu Yingying; Pang Lin; Wang Qiguang

doi:10.11898/1001-7313.20230404

Vol.34, NO.4, 2023

Turn off MathJax

Article Contents

Article Navigation > Journal of Applied Meteorological Science > 2023 > 34(4): 426-437

Hu Yingying, Pang Lin, Wang Qiguang. 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.

Citation:

Hu Yingying, Pang Lin, Wang Qiguang. 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.

Citation:

PDF( 2680 KB)

Application of Deep Learning Bias Correction Method to Temperature Grid Forecast of 7-15 Days

DOI: 10.11898/1001-7313.20230404

Hu Yingying^{1, 2},
Pang Lin³,
Wang Qiguang^{3
,
,}

1.
Chinese Academy of Meteorological Sciences, Beijing 100081
2.
University of Chinese Academy of Sciences, Beijing 100190
3.
China Meteorological Administration Training Center, Beijing 100081

Received Date: 2023-03-10
Rev Recd Date: 2023-06-06
Publish Date: 2023-07-31

Abstract

Abstract

The forecast error of numerical weather forecasting is inevitable, and there are still difficulties in temperature forecast of 7-15 days. To improve forecast accuracy and timeliness, the deviation correction technique is often used in operation. In recent years, deep learning methods have shown great potential in statistical post-processing of model forecasts. To improve the accuracy of Global Ensemble Prediction System of China Meteorological Administration (CMA-GEPS) for 7-15 days, error characteristics of 2 m temperature and 10 m wind products of CMA-GEPS control forecast provided by TIGGE data center from 25 December 2018 to 5 July 2022, and ERA5 data provided by ECMWF are analyzed. The U-Net model and residual connection model is used to conduct 2 m temperature lattice forecast error revision experiment for the lead time of 168-360 h in the region 15.75°-55.25°N, 73°-136.5°E. The experiments are designed with various data features to explore differences of the deep learning methods for longer lead time with different sample characteristics and model parameters, and performances of models are examined by comparing the bias, mean absolute bias and root mean square error. The results show that 2 m temperature forecast errors of 7-15 days become larger as the lead time increases. The model forecast skill gradually decreases, and in the target area, performance in eastern and southern marine and offshore areas is better than in western and northern the plateaus and mountains. The differences in the spatial distribution of errors are more prominent. Among the revised models, the effect of the U-Net model is better than that of the U-Net residual connection model, and adding the initial 2 m temperature data of ERA5 can greatly improves the performance, but the effect of adding CMA-GEPS control forecast 10 m wind product of CMA-GEPS control forecast is not apparent. For 9 lead times, the revised root mean square errors are reduced by 10%-25%, and the model can effectively reduce the large forecast errors for the northern Mongolian Plateau and the western Tibetan Plateau, and some mountainous areas in the target area.
- temperature grid forecast of 7-15 days;
- bias correction;
- deep learning method

FullText(HTML)

References(44)

Figures and Tables

Fig. 1 Root mean square error of CMA-GEPS model control forecast for 2 m temperature with different lead times from 8 Apr to 5 Jul in 2022

DownLoad: Full-Size Img PowerPoint

Fig. 2 Root mean square error of different lead times for 2 m temperature in the target area from 8 Apr to 5 Jul in 2022

DownLoad: Full-Size Img PowerPoint

图 3 CMA-GEPS模式控制预报与U-Net模型订正2 m气温平均绝对偏差

折线为基于数据集Ⅱ的订正结果相对CMA-GEPS控制预报的平均绝对偏差减小率

Fig. 3 Average absolute errors of 2 m temperature for CMA-GEPS model control forecast and U-Net model correction

the line denotes average absolute error reduction rate of U-Net model correction based on dataset Ⅱ relative to CMA-GEPS control forecast

DownLoad: Full-Size Img PowerPoint

图 4 同图 3，但为均方根误差

折线为基于数据集Ⅱ的订正结果相对CMA-GEPS控制预报的均方根误差减小率

Fig. 4 The same as in Fig. 3, but for root mean square error

the line denotes root mean square error reduction rate of U-Net model correction based on dataset Ⅱ relative to CMA-GEPS control forecast

DownLoad: Full-Size Img PowerPoint

Fig. 5 Daily root mean square error of 2 m temperature for 168 h lead time from 8 Apr to 5 Jul in 2022 by CMA-GEPS control forecast and U-Net model correction

DownLoad: Full-Size Img PowerPoint

Fig. 6 The same as in Fig. 5, but for 360 h lead time

DownLoad: Full-Size Img PowerPoint

Fig. 7 Root mean square error of 2 m temperature in the target area for 360 h lead time by CMA-GEPS control forecast and U-Net model correction

DownLoad: Full-Size Img PowerPoint

Table 1 Evaluation indicaters of CMA-GEPS model control forecast of 2 m temperature for 240 h lead time and deep learning model correction based on dataset Ⅰ (unit:℃)

预报与订正结果	评价指标
预报与订正结果	偏差	平均绝对偏差	均方根误差
CMA-GEPS控制预报产品	0.41	2.57	3.56
U-Net模型订正	0.51	2.40	3.16
U-Net残差连接模型订正	0.10	2.73	3.68

DownLoad: Download CSV

Table 2 The same as in Table 1, but based on dataset Ⅱ (unit:℃)

预报与订正结果	评价指标
预报与订正结果	偏差	平均绝对偏差	均方根误差
CMA-GEPS控制预报产品	0.41	2.57	3.56
U-Net模型订正	-0.21	1.67	2.29
U-Net残差连接模型订正	-0.03	1.77	2.44

DownLoad: Download CSV

References

[1]	Dai K, Cao Y, Qian Q F, et al. Situation and tendency of operation technologies in short-and medium-range weather forecast. Meteor Mon, 2016, 42(12): 1445-1455. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXX201612003.htm
[2]	Li Z C, Bi B G, Jin R H, et al. The development and application of the modern weather forecast in China for the recent 10 years. Acta Meteor Sinica, 2014, 72(6): 1069-1078. doi: 10.3969/j.issn.1004-4965.2014.06.007
[3]	Duan Y H, Jin R H. The status quo of modern weather operations in China and its future. Adv Meteor Sci Tech, 2012, 2(5): 6-11. https://www.cnki.com.cn/Article/CJFDTOTAL-QXKZ201205004.htm
[4]	Shen X S, Wang J J, Li Z C, et al. China's independent and innovative development of numerical weather prediction. Acta Meteor Sinica, 2020, 78(3): 451-476. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXB202003008.htm
[5]	Chou J F. An innovative road to numerical weather prediction—From initial value problem to inverse problem. Acta Meteor Sinica, 2007, 65(5): 673-682. doi: 10.3321/j.issn:0577-6619.2007.05.003
[6]	Zeng Q C. Weather forecast—From empirical to physicomathematical theory and super-computing system engineering. Physics, 2013, 42(5): 300-314. https://www.cnki.com.cn/Article/CJFDTOTAL-WLZZ201305001.htm
[7]	Zhang D Q, Zheng Z H, Chen L J, et al. Advances on the predictability and prediction methods of 10-30 d extended range forecast. J Appl Meteor Sci, 2019, 30(4): 416-430. doi: 10.11898/1001-7313.20190403
[8]	Kang Z M, Bao Y Y, Zhou N F. Current situation and development of medium-range and extended-range weather forecast in China. Adv Meteor Sci Tech, 2013, 3(1): 18-24. https://www.cnki.com.cn/Article/CJFDTOTAL-QXKZ201301008.htm
[9]	Jin R H, Ma J, Ren H C, et al. Advances and development countermeasures of 10-30 days extended-range forecasting technology in China. Adv Earth Science, 2019, 34(8): 814-825. https://www.cnki.com.cn/Article/CJFDTOTAL-DXJZ201908007.htm
[10]	Shen X S, Chen Q Y, Sun J, et al. Development of operation global medium-range forecast system in National Meteorological Centre. Meteor Mon, 2021, 47(6): 645-654. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXX202106001.htm
[11]	Wu M L, Lu Z Y, Wang Y. The Method for Mid-range Forecast in Future 8-17 Days//Proceedings of the 2006 Annual Conference of the Chinese Meteorological Society on Mesoscale Weather Dynamics, Numerical Simulation and Prediction, 2006: 929-935.
[12]	Shen X S, Su Y, Hu J L, et al. Development and operation transformation of GRAPES global middle-range forecast system. J Appl Meteor Sci, 2017, 28(1): 1-10. doi: 10.11898/1001-7313.20170101
[13]	Yin S, Li Y, Ma J, et al. Preliminary study on bias correction for the extended-range temperature forecast. Meteor Mon, 2020, 46(3): 412-419. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXX202003012.htm
[14]	Bauer P, Thorpe A, Brunet G. The quiet revolution of numerical weather prediction. Nature, 2015, 525(7567): 47-55.
[15]	Lorenz E N. Deterministic nonperiodic flow. J Atmos Sci, 1963, 20(2): 130-141.
[16]	Zhao S R, Zhao C G, Zhao R X, et al. The development of objective meteorological element forecast in China. Adv Meteor Sci Tech, 2012, 2(5): 12-21. https://www.cnki.com.cn/Article/CJFDTOTAL-QXKZ201205005.htm
[17]	Ji L R, Chen J B, Zhang D M, et al. Review of some numerical aspects of the dynamic framework of NWP model. Chinese J Atmos Sci, 2005, 29(1): 120-130. https://www.cnki.com.cn/Article/CJFDTOTAL-DQXK200501013.htm
[18]	Huang J Y. Meteorological Statistical Analysis and Forecasting Methods. Beijing: China Meteorological Press, 1990.
[19]	Shi N. Meteorological Statistical Forecasting. Beijing: China Meteorological Press, 2009.
[20]	Shi J E, Qu D G, Sun S Z, et al. Studies on long-range forecasting by statistical methods: Part Ⅰ—Application of stepwise multiple regression technique. Acta Meteor Sinica, 1964, 34(4): 507-518. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXB196404010.htm
[21]	Zhang Y T, Tong H, Sun J. Application of a bias correction method to meteorological forecast for the Pyeongchang Winter Olympic Games. J Appl Meteor Sci, 2020, 31(1): 27-41. doi: 10.11898/1001-7313.20200103
[22]	Klein W H, Lewis B M, Enger I. Objective prediction of five-day mean temperatures during winter. J Meteor, 1959, 16(6): 672-682.
[23]	Glahn H R, Lowry D A. The use of model output statistics (MOS) in objective weather forecasting. J Appl Meteor, 1972, 11(8): 1203-1211.
[24]	Ding S S. The advance of model output statistics method in China. Acta Meteor Sinica, 1985, 43(3): 332-338. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXB198503008.htm
[25]	Chen Y Y, Chen X G, Ma J R, et al. A study on subtle MOS forecasting method of wind. Sci Meteor Sinica, 2006, 26(2): 210-216. https://www.cnki.com.cn/Article/CJFDTOTAL-QXKX200602013.htm
[26]	Luo J Y, Zhou J S, Yan Y C. Local temperature MOS forecast method based on numerical forecast products and superior guidance. Meteor Sci Technol, 2014, 42(3): 443-450. https://www.cnki.com.cn/Article/CJFDTOTAL-QXKJ201403016.htm
[27]	Tong H, Guo P W, Zhu Y J, et al. Bias correction and statistical downscaling meteorological parameters forecast technique based on large-scale numerical model products. Meteor Mon, 2014, 40(1): 66-75. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXX201401008.htm
[28]	Sun J, Cao Z, Li H, et al. Application of artificial intelligence technology to numerical weather prediction. J Appl Meteor Sci, 2021, 32(1): 1-11. doi: 10.11898/1001-7313.20210101
[29]	Xu X F. From physical model to intelligent analysis: A new exploration to reduce the uncertainty of weather forecast. Meteor Mon, 2018, 44(3): 341-350. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXX201803001.htm
[30]	Ma L M. Development of artificial intelligence technology in weather forecast. Adv Earth Science, 2020, 35(6): 551-560. https://www.cnki.com.cn/Article/CJFDTOTAL-DXJZ202006001.htm
[31]	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.
[32]	Zhang Y B, Chen M X, Han L, et al. Multi-element deep learning fusion correction method for numerical weather prediction. Acta Meteor Sinica, 2022, 80(1): 153-167. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXB202201011.htm
[33]	Han N F, Yang L, Chen M X, et al. Machine learning correction of wind, temperature and humidity elements in Beijing-Tianjin-Hebei Region. J Appl Meteor Sci, 2022, 33(4): 489-500. doi: 10.11898/1001-7313.20220409
[34]	Zhao L N, Lu S, Qi D, et al. Daily maximum air temperature forecast based on fully connected neural network. J Appl Meteor Sci, 2022, 33(3): 257-269. doi: 10.11898/1001-7313.20220301
[35]	Yin X Y, Hu Z Q, Zheng J F, et al. Filling in the dual polarization radar echo occlusion based on deep learning. J Appl Meteor Sci, 2022, 33(5): 581-593. doi: 10.11898/1001-7313.20220506
[36]	Jin Z Q, Wang X M, Bao Y S, et al. Squall line identification method based on convolution neural network. J Appl Meteor Sci, 2021, 32(5): 580-591. doi: 10.11898/1001-7313.20210506
[37]	Chen Y Y, Yu X D, Gao X H, et al. A new method for non-linear classify and non-linear regression Ⅰ: Introduction to support vector machine. J Appl Meteor Sci, 2004, 15(3): 345-354. http://qikan.camscma.cn/article/id/20040344
[38]	Feng H Z, Chen Y Y. A new method for non-linear classify and non-linear regression Ⅱ: Application of support vector machine to weather forecast. J Appl Meteor Sci, 2004, 15(3): 355-365. http://qikan.camscma.cn/article/id/20040345
[39]	Zhang Y B, Song L Y, Chen M X, et al. A study of error correction for high-resolution gridded forecast based on a convolutional neural network in the Beijing-Tianjing-Hebei Region. Trans Atmos Sci, 2022, 45(6): 850-862. https://www.cnki.com.cn/Article/CJFDTOTAL-NJQX202206005.htm
[40]	Chen Y W, Huang X M, Li Y, et al. Ensemble learning for bias correction of station temperature forecast based on ECMWF products. J Appl Meteor Sci, 2020, 31(4): 494-503. doi: 10.11898/1001-7313.20200411
[41]	Jin R H, Dai K, Zhao R X, et al. Progress and challenge of seamless fine gridded weather forecasting technology in China. Meteor Mon, 2019, 45(4): 445-457. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXX201904001.htm
[42]	Richardson D, Buizza R, Hagedom R. First Workshop on the THORPEX Interactive Grand Global Ensemble(TIGGE). WMO/TD-No. 1273, WWRP/THORPEX-No. 05, 2005.
[43]	Li J C. Study on Runoff Forecasting of Honganjian River Basin Based on TIGGE Data. Zhengzhou: North China University of Water and Electric Power, 2019.
[44]	Ronneberger O, Fischer P, Brox T. U-Net: Convolutional networks for biomedical image segmentation. Medical Image Computing and Computer-Assisted Intervention(MICCAI), 2015, 9351: 234-241.