摘要
针对常规的PM_(2.5)浓度预测模型难以同时处理PM_(2.5)分布的空间非平稳性和空间自相关性问题,结合空气污染物、气象数据、全球导航卫星系统(GNSS)天顶对流层延迟,分析PM_(2.5)与这3类因子的相关性;选取其中与PM_(2.5)浓度显著相关的因子作为建模的辅助因子,建立了顾及多因子影响的地理加权回归克里金(GWRK)模型,进行中国区域PM_(2.5)浓度插值。结果表明,GWRK模型的平均绝对误差(MAE)、均方根误差(RMSE)、平均相对误差(MRE)、相关系数(r)分别为6.49μg/m^(3)、13.64μg/m^(3)、15.92%、0.85;除RMSE略逊于地理加权回归模型外,其余精度指标比反距离加权(IDW)、克里金、地理加权回归方法均有比较明显的提高。
Aiming at the question that conventional prediction models were difficult to simultaneously deal with the spatial nonstationarity and spatial autocorrelation of PM_(2.5)concentration distribution, by combining air pollutants, meteorological data and Global Navigation Satellite System(GNSS) zenith tropospheric delay, the correlations between PM_(2.5)and these three types of factors were analyzed. By selecting the factors which significantly correlated with PM_(2.5)concentration as the modeling cofactor, the Geographically Weighted Regression Kriging(GWRK) model considering multifactor influence was constructed to interpolate the PM_(2.5)concentration in China. The results shows that the Mean Absolute Error(MAE), Root Mean Square Error(RMSE), Mean Relative Error(MRE) and correlation coefficient of GWRK model are 6.49 μg/m^(3), 13.64 μg/m^(3), 15.92%, 0.85, respectively. Except that RMSE is slightly inferior to Geographically Weighted Regression(GWR) model, the remaining accuracy indicators are obviously improved compared with Inverse Distance Weighted(IDW), Kriging, and GWR methods.
作者
谢劭峰
周志浩
潘梦清
黄良珂
刘立龙
Xie Shaofeng;Zhou Zhihao;Pan Mengqing;Huang Liangke;Liu Lilong(College of Geomatics and Geoinformation,Guilin University of Technology,Guilin 541006,China;Land Reorganization and Reserve Center of Luohu District,Shenzhen City,Shenzhen 518023,China;Guilin Institute of Surveying and Mapping,Guilin 541100,China)
出处
《科技通报》
2022年第3期14-19,共6页
Bulletin of Science and Technology
基金
国家自然科学基金项目(41864002)。
关键词
PM_(2.5)空间插值
天顶对流层延迟
气象因子
地理加权回归克里金
PM_(2.5)
spatial interpolation
zenith tropospheric delay
meteorological factors
geographically weighted regression Kriging
