【摘要】 以京津冀2020年318个地面监测站点的PM2.5数据为估算因子,构建了时空线性混合效应模型(STLME)和时空嵌套线性混合效应模型(STNLME),为AOD数据的补值研究提供了一种新方法.结果表明:在有AOD-PM2.5匹配数据的日期,上述两个模型估算精度相近,交叉验证后决定系数R2分别为0.868和0.874,均方根误差RMSE分别为0.112和0.109;在无AOD-PM2.5匹配数据的日期,嵌套模型估算精度明显高于非嵌套模型,交叉验证后决定系数R2分别为0.63和0.26.经过模型补值后,研究区监测站点所在网格AOD数据空间维有效比率从原始数据的44.35%提高到99.35%,时间维有效比率从87.94%提高到100%;同时,每个站点的年均AOD值都有明显提高,弥补了高PM2.5浓度条件下缺失的AOD数据,可以减少空气污染和健康研究中暴露评估的偏差.更多还原

【Abstract】 A spatiotemporal linear mixed effect model(STLME) and a spatiotemporal nested linear mixed effect model(STNLME) were presented using the PM2.5 measurements of 318 ground monitoring stations in Beijing-Tianjin-Hebei(BTH) in 2020 to fill the missing data of AOD. The results indicated that the STLME and STNLME models in the days with AOD-PM2.5 matchups showed similar performance with the cross-validation(CV) R2 valued at 0.868 and 0.874, and the root mean square error(RMSE) valued at 0.112 and 0.109, respectively. However, the STNLME model with the CV R2 valued at 0.63 outperforms STLME with the CV R2 of 0.26 in the days without PM2.5-AOD matchups. After models filling, the spatial valid value ratio of AOD data in the grid where the monitoring stations are located was increased from 44.35% to 99.35%, and the temporal valid value ratio was increased from 87.94%to 100%. Meanwhile, the annual mean AOD value of each station had increased significantly, and the missing AOD were filled under the condition of high PM2.5 level, which could reduce the biases of exposure assessment in air pollution and health studies.更多还原