Application of ARIMA Model for Mid- and Long-term Forecasting of Ozone Concentration
Application of ARIMA Model for Mid- and Long-term Forecasting of Ozone Concentration
复制标题
DOI：
10.13227/j.hjkx.202011237
复制
发表时间：
2021-07-15
期刊：
影响因子：
--
通讯作者：
Ma Zhi-qiang
中科院分区：
文献类型：
--
作者：
Li Ying-ruo;Han Ting-ting;Ma Zhi-qiang
臭氧污染最近成为京津冀地区严重的空气质量问题。由于缺乏前体排放清单和臭氧生成的物理化学机制的复杂性，数值模拟在臭氧预报中仍存在较大偏差。时间序列分析模型简单、计算成本低，可有效地应用于臭氧污染预测。我们对上店子、保定和天津三个站点的臭氧浓度进行了时间序列分析。建立了季节性和动态ARIMA模式进行中长期臭氧预报。季节ARIMA模式臭氧月平均预测值与实测值的相关系数R可达0.951，RMSE仅为10.2 μ g·m（-3）。动态ARIMA模式预测的8 h臭氧平均值预测值与实测值的相关系数R由0.296 ~ 0.455提高到0.670 ~ 0.748，RMSE得到有效降低。
Ozone pollution has recently become a severe air quality issue in the Beijing-Tianjin-Hebei region. Due to the lack of a precursor emission inventory and complexity of physical and chemical mechanism of ozone generation, numerical modeling still exhibits significant deviations in ozone forecasting. Owing to its simplicity and low calculation costs, the time series analysis model can be effectively applied for ozone pollution forecasting. We conducted a time series analysis of ozone concentration at Shangdianzi, Baoding, and Tianjin sites. Both seasonal and dynamic ARIMA models were established to perform mid- and long-term ozone forecasting. The correlation coefficient R between the predicted and observed value can reach 0.951, and the RMSE is only 10.2 mu g.m(-3) for the monthly average ozone prediction by the seasonal ARIMA model. The correlation coefficient R between the predicted and observed value increased from 0.296-0.455 to 0.670-0.748, and RMSE was effectively reduced for the 8-hour ozone average predicted by the dynamic ARIMA model.