研究报告
郎艺超,肖璐,George Christakos.基于SARIMA模型和普通Kriging法对杭州市主城区PM2.5短期预测和制图[J].环境科学学报,2018,38(1):62-70
基于SARIMA模型和普通Kriging法对杭州市主城区PM2.5短期预测和制图
- Forecasting concentrations of PM2.5 in main urban area of Hangzhou and mapping using SARIMA model and ordinary Kringing method
- 基金项目：国家自然科学基金（No.41671399）
- 作者
- 单位
- 郎艺超
- 浙江大学海岛与海岸带研究所, 舟山 316021
- 肖璐
- 浙江大学海岛与海岸带研究所, 舟山 316021
- George Christakos
- 浙江大学海岛与海岸带研究所, 舟山 316021
- 摘要：基于SARIMA模型对杭州市主城区内的7个站点和周边3个站点的PM2.5浓度进行4 h平均的短期预报，并利用普通Kriging法对杭州市主城区PM2.5进行空间插值和制图.在建立SARIMA模型时，用批量自动化的方法，使用R语言编程对夏季和冬季各360期的数据进行SARIMA（p，d，q）×（P，D，Q）6模型的参数的确定和拟合，来预测下一期的值.在10个站点分别进行120期的预测，做出真实与预测的时间序列图，在精度分析中，夏季和冬季PM2.5浓度总体的平均绝对误差（MAE）分别为8.4 μg·m-3和14.8 μg·m-3.在ArcGIS中，对每期的数据使用普通Kriging法，用球面模型拟合半变异函数，决定各个站点的权重，来对杭州市区内未知空间样点进行插值，最后生成完整的杭州市主城区PM2.5短期预测图.本研究创新性地将SARIMA模型广义化运用到小尺度时间序列中，预报效果较好，并且批量自动化预测和制图的方法，可为今后的预测制图产品化提供技术支持.
- Abstract：The average four-hour short-time forecasting of PM2.5 concentrations of 7 stations in main urban area of Hangzhou and the surrounding 3 stations based on the SARIMA model, and the ordinary Kriging method was used to interpolate and mapping. When establishing the SARIMA model, we use the batch automation method by R language to fit the parameters of the SARIMA(p,d,q)×(P,D,Q)6 model during respective 360 periods in summer and winter to predict the next value. The PM2.5 concentrations of 120 periods in each 10 stations were forecasted and then were made into time series charts of the true and predictive value. In precision analysis, the MAE was 8.4 μg·m-3 and 14.8 μg·m-3 in summer and winter respectively. In ArcGIS, the ordinary Kriging method was used for interpolating each period of data by means of spherical model to fit semivariogram and determine the weight of each station for interpolating unknown space in Hangzhou. Finally, the short-term PM2.5 concentrations forecasting maps in main urban area of Hangzhou were completed. This research innovatively generalized the SARIMA model into small scale time series with good prediction precision, and the batch automation method could potentially be employed in the operational forecasting and mapping.
摘要点击次数： 1060 全文下载次数： 1565
