目的 拟合最优预测模型,探讨乌鲁木齐市可吸入颗粒物(inhalable particles,PM10)污染时间分布特征及趋势变化,为推进大气治理提供参考依据.方法 以2016-2022年乌鲁木齐市颗粒物PM10监测资料为基础,构建颗粒物PM10月均浓度数据库,应用自回归集成移动平均模型(autoregressive integrated moving average model,ARIMA)分别拟合颗粒物PM10预测模型及对其分布特征进行分析,并预测2023-2024年颗粒物PM10浓度趋势变化.结果 2016-2022年乌鲁木齐市颗粒物PM10月均浓度数值比较,差异有统计学意义(P＜0.01);最优模型为ARIMA(0,0,1)(1,1,0)12;月均浓度值逐年减小,每年的1、2、12月浓度值达到最大;经预测,2023-2024年乌鲁木齐市颗粒物PM10月均浓度变化趋势与2016-2022年一致.结论 最优模型为ARIMA(0,0,1)(1,1,0)12;乌鲁木齐市颗粒物PM10月均浓度每年呈现秋冬季节高,呈逐年降低的趋势变化,该模型可对乌鲁木齐市颗粒物PM10月均浓度进行有效的短期预测分析.
Analysis on characteristics and trend of PM10 pollution in Urumqi City based on ARIMA model
Objective To fit the optimal prediction model and explore the temporal distribution characteristics and trend changes of inhalable particles(PM10)pollution in Urumqi City,providing reference for promoting atmospheric governance.Methods Based on the monitoring data of PM10 in Urumqi City from 2016 to 2022,the database of PM10 monthly average concentration was constructed,and the autoregressive integrated moving average model(ARIMA)was used to fit the prediction model of PM10 and analyze its distribution characteristics,and forecast the trend change of PM10 concentration from 2023 to 2024.Results There was a statistically significant difference in the monthly average concentration of PM10 in Urumqi City from 2016 to 2022(P＜0.01),and the optimal model was ARIMA(0.0,1)(1,1,0)12.The monthly average concentration values of PM10 in Urumqi City had been decreasing year by year,reaching their maximum values in January,February and December.According to the prediction,the monthly average concentration of particulate matter PM10 in Urumqi City from 2023 to 2024 was consistent with that from 2016 to 2022.Conclusion The optimal model is ARIMA(0,0,1)(1,1,0)12.The monthly average concentration of PM10 in Urumqi City shows a trend of increasing in autumn and winter,and decreasing year by year.This model can effectively predict and analyze the monthly average concentration of PM10 in Urumqi City in the short term.
Atmospheric particulate matterTime series analysisPrediction
NETL
NSTL
万方数据
