Study of 2 m temperature variation correction during transitional processes of temperature in Sichuan

doi:10.11755/j.issn.1006-7639(2023)-01-0164

Abstract

Abstract:

Based on the daily 2 m maximum and minimum temperature data from 1990 to 2019 in Sichuan Province, the temperature transitional weather processes have been analyzed statistically. Then a correction model of temperature change during transitional processes of temperature has been performed by using of NCEP/NCAR (National Center for Environmental Prediction/National Center for Atmospheric Research) daily reanalysis data and the LightGBM (Light Gradient Boosting Machine) algorithm.The results show that the area with the most temperature transitional processes is the slope transition zone between the plateau and the basin, while the least is in the basin. The number of temperature transitional processes in each region has an obviously seasonal differences with the most in spring and the least in winter, and the temperature transitional processes in spring is significantly more than those in the other three seasons. For the training set from 1990 to 2019，the LightGBM model has good performances with an overall accuracy of 78.64% and a mean absolute error of 1.35 ℃. For the independent testing set in 2020，the LightGBM model has an overall accuracy of 53.60% and a mean absolute error of 2.19 ℃, which are better than those of ECMWF (European Centre for Medium-Range Weather Forecasting), SCMOC and SPCO models.

Key words: transitional process of temperature, LightGBM algorithm, machine learning

摘要：

基于四川地区1990—2019年的逐日2 m最高、最低温度站点实况数据，对气温转折天气过程进行统计和分析，在此基础上，应用LightGBM（Light Gradient Boosting Machine）算法及NCEP/NCAR（National Center for Environmental Prediction/National Center for Atmospheric Research）逐日再分析资料，构建气温转折天气过程变温订正模型。结果表明：（1）出现气温转折过程最多的区域是高原与盆地的边坡过渡区，最少的是盆地；（2）各区域的气温转折过程具有明显的季节差异，均表现为春季最多、冬季最少，且春季的气温转折过程明显多于其他3季；（3）在1990—2019年验证集中，LightGBM订正模型表现较好，准确率为78.64%，平均绝对误差为1.35 ℃。（4）在2020年的独立样本测试中，LightGBM订正模型的准确率为53.60%，平均绝对误差为2.19 ℃，整体订正效果优于ECMWF模式（European Centre for Medium-Range Weather Forecasting）、中央台城镇预报指导报（SCMOC）及四川省气象台数值预报客观释用城镇预报指导报（SPCO）的预报。

关键词: 气温转折过程, LightGBM算法, 机器学习

CLC Number:

P457.3

FENG Liangmin, ZHOU Qiuxue, CAO Pingping, WANG Jiajin. Study of 2 m temperature variation correction during transitional processes of temperature in Sichuan[J]. Journal of Arid Meteorology, 2023, 41(1): 164-172.

冯良敏, 周秋雪, 曹萍萍, 王佳津. 四川地区气温转折过程2 m温度变化订正研究[J]. 干旱气象, 2023, 41(1): 164-172.

Figures/Tables 9

Fig.1 Partitioning of Sichuan region based on altitude （the color shaded， Unit： m）（A： the basin， B： the slope transition zone of plateau and basin，C： the mountains of southwestern Sichuan， D： the plateau in western Sichuan）

Fig.2 The box plots of absolute values of daily variation of the maximum and minimum 2 m temperature in four regions of Sichuan Province from 1990 to 2019

Fig.3 Distribution of temperature transitional processes in four regions of Sichuan Province during 1990-2019

Fig.4 Seasonal distribution of temperature transitional processes in four regions of Sichuan Province during 1990-2019

Fig.5 The distribution of characteristic factors with the absolute value of correlation coefficients over 0.3 between maximum temperature change and characteristic factors in different regions of Sichuan Province （The horizontal axis δH represents the potential height variation field, δT represents the temperature variation field, RH and δRH represent the humidity field and its variation field, respectively, u, v and δu, δv represent the wind fields and their variation fields, respectively, the numbers after the short line represent the geopotential height, δSLP represents the sea level pressure variation in 24 hours）

Tab.1 Main parameters of the model

n-estimators	learning-rate	num-leaves	boosting-type
100	0.10	192	GBDT

Tab.2 The accuracy and mean absolute error of the LightGBM model in validation set in different regions of Sichuan Province

盆地			高原与盆地的边坡过渡区			川西南山地			川西高原			全省
R/%	MAE/℃		R/%	MAE/℃		R/%	MAE/℃		R/%	MAE/℃		R/%	MAE/℃
86.59	1.07		83.03	1.19		64.31	1.87		65.14	1.83		78.64	1.35

Tab.3 The comparison of accuracy and mean absolute error of the LightGBM， ECMWF，SCMOC and SPCO model in the test set

区域	LightGBM		ECMWF		SCMOC		SPCO
区域	R/%	MAE/℃	R/%	MAE/℃	R/%	MAE/℃	R/%	MAE/℃
盆地	50.18	2.90	47.88	2.51	32.16	3.94	39.58	2.95
高原与盆地的边坡过渡区	58.44	2.13	50.77	2.44	40.50	3.22	50.30	2.33
川西南山地	45.58	2.79	47.73	2.59	39.86	3.25	50.36	2.54
川西高原	57.60	2.13	42.00	2.81	52.80	2.68	49.20	2.33
全省	53.60	2.19	48.32	2.53	39.58	3.36	47.26	2.54

Fig.6 The difference distribution of accuracy between the LightGBM model and ECMWF （a），SCMOC （b） and SPCO （c） in the test set （Unit： %）

References 22

[1]	曹渝昆, 朱萌, 2019. 基于主成分分析和LightGBM的风电场发电功率超短期预测[J]. 上海电力学院学报, 35(6): 562-566.
[2]	曹萍萍, 肖递祥, 徐栋夫, 等, 2018. 基于概率匹配的西南区域模式定量降水订正试验[J]. 气象科技, 46(1): 102-111.
[3]	符娇兰, 代刊, 2016. 基于CRA空间检验技术的西南地区东部强降水EC模式预报误差分析[J]. 气象, 42(12): 1 456-1 464.
[4]	冯慧敏, 智协飞, 崔慧慧, 等, 2016. 基于多模式集成技术的地面气温精细化预报[J]. 气象与环境科学, 39(4): 73-79.
[5]	胡玉荣, 2008. 新形势下气象科技服务发展的思考[J]. 气象软科学, 1(3): 20-24.
[6]	何珊珊, 蓝盈, 戚云枫, 2021. GRAPES-GFS模式2 m温度预报的最优时窗滑动订正方法[J]. 气象科技, 49(5): 746-753.
[7]	刘新伟, 黄武斌, 蒋盈沙, 等, 2021. 基于LightGBM算法的强对流天气分类识别研究[J]. 高原气象, 40(4): 909-918. DOI
[8]	刘军中, 2021. 基于机器学习的最大风速预测系统研究与实现[D]. 南昌: 南昌大学.
[9]	孙敏, 袁慧玲, 杜予罡, 2018. 上海地区春季最高气温预报失败案例分析[J]. 气象, 44(1): 65-79.
[10]	盛裴轩, 毛节泰, 李建国, 等, 2003. 大气物理学[M]. 北京: 北京大学出版社, 122-154.
[11]	佟华, 姚明明, 王雨, 等, 2006. T213L31全球中期数值天气预报系统2 m温度预报误差源分析[J]. 气象, 32(2): 52-57.
[12]	谭江红, 陈伟亮, 王珊珊, 2018. 一种机器学习方法在湖北定时气温预报中的应用试验[J]. 气象科技进展, 8(5): 46-50.
[13]	吴启树, 韩美, 郭弘, 等, 2016. MOS温度预报中最优训练期方案[J]. 应用气象学报, 27(4): 426-434.
[14]	王志宇, 2019. 基于LightGBM框架的上海市大气能见度预报订正研究[D]. 上海: 华东师范大学.
[15]	王丹, 高红燕, 张宏芳, 等, 2015. 一种逐时气温预报方法[J]. 干旱气象, 33(1): 89-97. DOI
[16]	肖明静, 隋明, 范苏丹, 等, 2012. 3种数值模式温度预报产品在山东应用的误差分析与订正[J]. 干旱气象, 30(3): 472-477.
[17]	薛志磊, 张书余, 2012. 气温预报方法研究及其应用进展综述[J]. 干旱气象, 30(3): 451-458.
[18]	薛谌彬, 陈娴, 张瑛, 等, 2019. ECMWF高分辨率模式2 m温度预报误差订正方法研究[J]. 气象, 45(6): 831-842.
[19]	肖红茹, 龙柯吉, 伍清, 等, 2020. 1980—2017年四川盆地寒潮及其气温变化特征[J]. 高原山地气象研究, 40(4): 47-52.
[20]	张丹峰, 2018. 基于LightGBM, XGBoost, ERT混合模型的风机叶片结冰预测研究[D]. 上海: 上海师范大学.
[21]	BAUER P, THORPE A, BRUNET G, 2015. The quiet revolution of numerical weather prediction[J]. Nature, 525 (7567): 47-55. DOI URL
[22]	BONAVITA M, HOLM E, ISAKSEN L, et al, 2016. The evolution of the ECMWF hybrid data assimilation system[J]. Quarterly Journal of the Royal Meteorological Society, 142 (694): 287-303.