基于极端梯度提升算法的西安市逐小时PM2.5浓度预报研究

doi:10.3969/j.issn.1673-503X.2023.01.006

Abstract

Abstract:

Based on the eXtreme Gradient Boosting (XGBoost) machine learning algorithm model, using hourly PM_2.5 concentration monitoring data and meteorological observation data in Xi'an from 2016 to 2021, the forecast test of hourly PM_2.5 concentration was carried out by selecting meteorological and time factors as the input features.The results showed that PM_2.5 concentration has a significant negative correlation with average temperature and visibility, and relative humidity and dew point temperature are significantly positively correlated with PM_2.5 concentration in winter.Easterly wind is more likely to produce heavily polluted weather.Generally, air pollution occurs frequently from the end of December to the beginning of January, but the PM_2.5 concentration is decreasing year by year.The PM_2.5 concentration in winter shows the most obvious bimodal diurnal variation with the highest values appearing in the early morning and around 11:00.Meanwhile, there is a "weekend effect" in the change of PM_2.5 concentration.The forecast model can truly reflect the changes of PM_2.5 concentration magnitude and trend with the determination coefficient of 0.77, the mean absolute error of 12.79 μg·m^-3 and the root mean square error of 18.68 μg·m^-3 between forecasted and observed values.The model has a relatively stable performance and better effect in forecasting PM_2.5 concentration in autumn and winter than in spring and summer but underestimates the extreme peaks.Besides, the forecast model has good interpretability and is clearly influenced by the visibility feature variables and the importance of feature variable such as dew point temperature, relative humidity, average temperature, and sea level pressure decrease in turn.Meanwhile, the time factors have a certain impact on the model.In addition, the forecast accuracy and efficiency of this model are higher than those of other statistical and machine-learning models.

Key words: PM_2.5 concentration, eXtreme Gradient Boosting, Machine learning, Meteorological factors, Forecast

CLC Number:

P456.8

Xu-ting ZHANG, Hui LIU, Rui-fang LIU, Fei JU, Jia-hui-min LIU, Xing-xing GAO, Shao-ni HUANG, Nan WANG. Study on hourly PM_2.5 concentration forecast based on XGBoost method in Xi'an city[J]. Journal of Meteorology and Environment, 2023, 39(1): 44-54.

Figures/Tables 12

Table 1

Fig.1

Fig.2

Table 2

Fig.3

Fig.4

Fig.5

Fig.6

Table 3

Fig.7

Fig.8

Table 4

References 30

1	王庚辰, 王普才. 中国PM_2.5污染现状及其对人体健康的危害[J]. 科技导报, 2014, 32 (26): 72- 78. doi: 10.3981/j.issn.1000-7857.2014.26.010
2	Khaniabadi Y O , Goudarzi G , Daryanoosh S M , et al. Exposure to PM₁₀, NO₂, and O₃ and impacts on human health[J]. Environmental Science & Pollution Research, 2017, 24 (3): 2781- 2789.
3	张灿, 王静, 简高武, 等. 环境空气质量预报方法简介[J]. 环境与可持续发展, 2018, 43 (5): 139- 141. doi: 10.3969/j.issn.1673-288X.2018.05.036
4	Zhou G Q , Xu J M , Xie Y , et al. Numerical air quality forecasting over eastern China: An operational application of WRF-Chem[J]. Atmospheric Environment, 2017, 153, 94- 108. doi: 10.1016/j.atmosenv.2017.01.020
5	刘闽, 王帅, 林宏, 等. 沈阳市冬季环境空气质量统计预报模型建立及应用[J]. 中国环境监测, 2014, 30 (4): 10- 15. doi: 10.19316/j.issn.1002-6002.2014.04.002
6	沈劲, 钟流举, 何芳芳, 等. 基于聚类与多元回归的空气质量预报模型开发[J]. 环境科学与技术, 2015, 38 (2): 63- 66.
7	张自银, 赵秀娟, 熊亚军, 等. 基于动态统计预报方法的京津冀雾霾中期预报试验[J]. 应用气象学报, 2018, 29 (1): 57- 69.
8	李晓岚, 刘旸, 栾健, 等. 基于BP人工神经网络法沈阳市PM_2.5质量浓度集成预报试验[J]. 气象与环境学报, 2018, 34 (2): 100- 106.
9	戴李杰, 张长江, 马雷鸣. 基于机器学习的PM_2.5短期浓度动态预报模型[J]. 计算机应用, 2017, 37 (11): 3057- 3063.
10	曲悦, 钱旭, 宋洪庆, 等. 基于机器学习的北京市PM_2.5浓度预测模型及模拟分析[J]. 工程科学学报, 2019, 41 (3): 401- 407.
11	夏晓圣, 陈菁菁, 王佳佳, 等. 基于随机森林模型的中国PM_2.5浓度影响因素分析[J]. 环境科学, 2020, 41 (5): 2057- 2065.
12	任才溶, 谢刚. 基于随机森林和气象参数的PM_2.5浓度等级预测[J]. 计算机工程与应用, 2019, 55 (2): 213- 220.
13	夏润, 张晓龙. 基于改进集成学习算法的在线空气质量预测[J]. 武汉科技大学学报, 2019, 42 (1): 61- 67.
14	康俊锋, 黄烈星, 张春艳, 等. 多机器学习模型下逐小时PM_2.5预测及对比分析[J]. 中国环境科学, 2020, 40 (5): 1895- 1905.
15	高星星, 桂海林, 潘留杰, 等. 基于卫星遥感和地面观测资料的汾渭平原空气污染分析[J]. 气象与环境学报, 2020, 36 (4): 36- 44.
16	张文静, 王琦, 胡琳, 等. 西安市霾天气与清洁天气变化特征及影响因素分析[J]. 气象与环境学报, 2015, 31 (2): 31- 36.
17	白杨, 杨剑, 陈鹏, 等. 基于空间插值的西安市重空气污染期间主要污染物时空变化特征及相关性分析[J]. 环境科学研究, 2020, 33 (4): 809- 819.
18	黄蕾, 毕旭, 杨晓春, 等. 2014—2017年西安市PM_2.5污染特征及影响因子[J]. 干旱气象, 2020, 38 (3): 440- 447.
19	高星星, 桂海林, 王楠, 等. 汾渭平原一次沙尘和人为混合空气污染过程分析[J]. 气象与环境学报, 2021, 37 (1): 1- 8.
20	梁杰, 陈嘉豪, 张雪芹, 等. 基于独热编码和卷积神经网络的异常检测[J]. 清华大学学报(自然科学版), 2019, 59 (7): 523- 529.
21	Chen T Q, Guestrin C. XGBoost: A scalable tree boosting system[C]//Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Francisco, USA: ACM, 2016: 785-794.
22	齐甜方, 蒋洪迅, 石晓文. 面向多源数据沈阳市PM_2.5浓度预测研究及实证分析[J]. 系统工程, 2018, 36 (5): 104- 115.
23	Lundberg S M, Lee S. A unified approach to interpreting model predictions[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems (NIPS'17). California, USA: NIPS, 2017: 4768-4777.
24	Parsa A B , Movahedi A , Taghipour H , et al. Toward safer highways, application of XGBoost and SHAP for real-time accident detection and feature analysis[J]. Accident Analysis & Prevention, 2020, 136, 105405.
25	郑小华, 李明星, 刘慧, 等. 汾渭平原空气质量的时空特征及其与气象因子的关系[J]. 环境科学学报, 2020, 40 (11): 4113- 4121.
26	高维英, 李明, 李茜, 等. 关中地区重污染发生和减弱天气类型及气象要素定量特征[J]. 灾害学, 2020, 35 (2): 102- 109.
27	黄少妮, 孟金平, 杜莉丽, 等. 陕西关中两次霾重污染过程气象条件分析[J]. 气象科技, 2019, 47 (5): 809- 817.
28	赵卉伊褶, 周卫健, 牛振川, 等. 西安市2013—2017年大气PM_2.5的时空变化特征及影响因素分析[J]. 地球环境学报, 2019, 10 (3): 248- 256.
29	邓宏兵, 蹇雪莹. 长江经济带城市雾霾污染的周末效应及假期效应研究[J]. 华中师范大学学报(自然科学版), 2020, 54 (4): 536- 545.
30	张淑平, 韩立建, 周伟奇, 等. 冬季PM_2.5的气象影响因素解析[J]. 生态学报, 2016, 36 (24): 7897- 7907.

空气质量级别	空气质量类别	PM_2.5浓度/(μg·m^-3)
一级	优	35
二级	良	75
三级	轻度污染	115
四级	中度污染	150
五级	重度污染	250
六级	严重污染	500

相关系数r	PM_2.5	海平面气压	平均气温	露点温度	相对湿度	平均风向	平均风速	能见度
PM_2.5	-	-0.2697^*	-0.0577^*	0.4363^*	0.4406^*	-0.0028	-0.2063^*	-0.6150^*
海平面气压	0.3613^*	-	-0.5399^*	-0.5398^*	-0.0881^*	0.0656^*	0.0322^*	0.3219^*
平均气温	-0.5040^*	-0.8896^*	-	0.3426^*	-0.4273^*	-0.0299^*	0.1096^*	0.0954^*
露点温度	-0.3976^*	-0.8080^*	0.8293^*	-	0.6757^*	0.0071	-0.1507^*	-0.6221^*
相对湿度	0.0883^*	0.0169^*	-0.1453^*	0.4163^*	-	0.0473^*	-0.2237^*	-0.6423^*
平均风向	0.0278^*	0.1226^*	-0.1071^*	-0.0873^*	0.0434^*	-	-0.2703^*	-0.0272^*
平均风速	-0.1682^*	-0.1236^*	0.1707^*	0.0625^*	-0.1750^*	-0.2811^*	-	0.2473^*
能见度	-0.5380^*	-0.3396^*	0.4807^*	0.1352^*	-0.5359^*	-0.0310^*	0.1879^*	-

季节	r	R²	MAE/(μg·m^-3)	RMSE/(μg·m^-3)
春季	0.76	0.58	12.53	16.14
夏季	0.65	0.42	7.05	9.17
秋季	0.88	0.77	11.25	15.48
冬季	0.85	0.72	20.46	28.61

方法	r	R²	MAE/(μg·m^-3)	RMSE/(μg·m^-3)	训练时长/s
线性回归	0.71	0.50	26.63	36.57	＜1
岭回归	0.71	0.50	26.54	36.41	＜1
支持向量机	0.71	0.50	19.45	28.22	190
随机森林	0.87	0.75	12.76	19.37	160
梯度提升决策树	0.87	0.75	13.42	19.38	16
本文方法	0.88	0.77	12.79	18.68	4

[1]	Rui-yang HUA,Wei LI,Cheng-zhi YE,Chang-jian NI. Variations of ozone concentration and its influence factors in Changsha from 2016 to 2019 [J]. Journal of Meteorology and Environment, 2023, 39(2): 44-52.
[2]	Zheng-hua TAN, Zhong-yan LU, Hai-feng LIN, An-qi NIE, Fang-da TENG. Verification for precipitation forecasted by NWP models in Liaoning province during the summer of 2020 [J]. Journal of Meteorology and Environment, 2023, 39(1): 10-16.
[3]	Xue-yi XUN,Bao-cheng YANG,Ying-hua HU,Min ZHANG. Verification on the summertime Eurasian middle-high latitude ensemble forecasting circulation system [J]. Journal of Meteorology and Environment, 2023, 39(1): 17-25.
[4]	Jing-li WANG,Tian-jiao GAO,Rong-ping LI,Peng-cheng YU,Yun-tao MA,Na MI,Ri-hong WEN,Kai ZHANG. Research on soil moisture retrieval model based on optical remote sensing and microwave remote sensing [J]. Journal of Meteorology and Environment, 2023, 39(1): 73-82.
[5]	Mei-hui QU,Wen-gang ZHU,Shu-hong ZHAO,Xiao-tong ZHU,Gang TU,Zong-ting GAO. Influence of conventional observation assimilation on numerical prediction of a short-time heavy precipitation process in Jilin province [J]. Journal of Meteorology and Environment, 2022, 38(6): 10-19.
[6]	Jing LIU, Chuan-lei CHEN, Ying WANG, Kui-zhi CAI, Chuan REN, Wei DONG. Comparison of neighborhood verification of radar echo forecast in Liaoning province using four high-resolution models [J]. Journal of Meteorology and Environment, 2022, 38(5): 25-33.
[7]	Nan FANG, Guo-quan XIE, Xiao-jian RUAN, Chen-ping REN, Shu-jie JIANG, Wei-wei ZHANG. Application of long short-term memory neural network (LSTM) model in low visibility forecast [J]. Journal of Meteorology and Environment, 2022, 38(5): 34-41.
[8]	Dong-dong WANG, Li SUN, Lei YANG, Li-du SHEN, Shu WANG, Yu CHEN. Multi-model comparison and evaluation of forecast statistics and spatial verification for different precipitation types caused by Typhoon Bavi [J]. Journal of Meteorology and Environment, 2022, 38(4): 37-46.
[9]	Tao WANG, Yi-shu WANG, Chun-yu ZHAO, Xiao-tao WANG, Mei-ou QIN, Yu-min SHEN, Yi-ling HOU, Jian-yun ZHAO. Prediction model of first-frost date in Liaoning province using machine learning methods [J]. Journal of Meteorology and Environment, 2022, 38(4): 47-56.
[10]	Sheng-jie CHEN,Xin-ru TIAN,Ruan YAO,Lin JIANG. Forecast of electricity consumption in Nanjing based on meteorological factors in winter and summer [J]. Journal of Meteorology and Environment, 2022, 38(4): 127-135.
[11]	Hui-meng BAO,Da-feng GUO,Wei LI. Correction tests of ECMWF modeling quantitative precipitation using frequency matching method in Jiangxi provice [J]. Journal of Meteorology and Environment, 2022, 38(2): 12-20.
[12]	Ying-ying MENG, Shu-ling LI, Ling ZHAO, Xue-mei BAI, Min-hui YAN. Validation of three forecast products of air temperature in Heilongjiang province [J]. Journal of Meteorology and Environment, 2022, 38(2): 21-30.
[13]	Jia-qi ZHAO. Spatial and temporal characteristics of climatic comfort and analysis of key meteorological factors in Hohhot-Baotou-Ordos region [J]. Journal of Meteorology and Environment, 2022, 38(2): 78-88.
[14]	Qiang-jun LIU, Jun-fang SONG, Cheng-yu NIU, Jun ZHAO, Yu-jun ZHANG. Characteristics of O₃ concentration and its relationships with meteorological factors in the southern foot of Taihang Mountains from 2015 to 2019 [J]. Journal of Meteorology and Environment, 2021, 37(4): 48-53.
[15]	Bo ZHANG,Chao YU,Wei HUANG,Chuan-lei CHEN. Effect of persistent blocking in Eurasian on temperature and precipitation in June in Greater Khingan Mountains and the precursor signal of temperature [J]. Journal of Meteorology and Environment, 2021, 37(4): 63-69.

Study on hourly PM_2.5 concentration forecast based on XGBoost method in Xi'an city

RichHTML

PDF (PC)

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 12

References 30

Related Articles 15

Metrics

Comments

Recommended 10

Study on hourly PM2.5 concentration forecast based on XGBoost method in Xi'an city

RichHTML

PDF (PC)

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 12

References 30

Related Articles 15

Metrics

Comments

Recommended 10

Study on hourly PM_2.5 concentration forecast based on XGBoost method in Xi'an city