基于时空混合效应模型的京津冀PM2.5浓度变化模拟

为揭示京津冀地区高精度PM_2.5的时空分布特征，以空间分辨率为1 km的MAIAC AOD数据为主要预测因子，以气象数据、植被指数、夜间灯光数、人口密度和海拔数据作为辅助因子，构建了一种新的时空混合效应模型（STLME）,在拟合最优次区域划分方案基础上对京津冀地区PM_2.5浓度进行预测分析.结果表明，基于STLME模型的ρ(PM_2.5)预测精度高于传统的线性混合效应模型（LME）,其十折交叉验证（CV）R²为0.91,明显高于LME模型的0.87,说明STLME模型在同时校正PM_2.5-AOD关系的时空异质性方面具有优势.最优次区域划分方案识别出PM_2.5-AOD关系的空间差异，并结合缓冲区平滑方法，提高了STLME模型预测精度.京津冀PM_2.5浓度时空变化差异显著，高值区主要分布在以石家庄、邢台和邯郸为中心的河北南部，低值区则位于燕山-太行山区；冬季PM_2.5污染最严重，其次是秋季和春季，夏季污染最轻.STLM...

Modeling of PM2.5 Concentrations in the Beijing-Tianjin-Hebei Region Using a Space-time Linear Mixed Effects Model

To reveal the spatiotemporal distribution characteristics of high-precision PM_2.5 concentrations in the Beijing-Tianjin-Hebei (BTH) region, a space-time linear mixed effects (STLME) model was developed in this study. The MAIAC AOD at a 1 km spatial resolution and the meteorological material, vegetation index, light quantity at night, population density, and altitude data were employed as the main and auxiliary predictive factors in the STLME model, respectively, to estimate the ground-level PM_2.5 concentrations on the BTH region by optimizing the sub-regional division scheme. The results indicated that the STLME model with the CV R² valued at 0.91 outperformed traditional linear-mixed effects (LME) with a CV R² of 0.87, indicating the superiority of the STLME model in simultaneously correcting the spatiotemporal heterogeneity of the PM_2.5-AOD relationship. The optimal sub-region partitioning scheme identified the spatial difference in the PM_2.5-AOD relationship and, combined with the buffer smoothing method, improved the prediction accuracy of the STLME model. The PM_2.5 levels in the BTH region exhibited strong spatiotemporal variations. The areas with higher PM_2.5 concentrations were mainly located in the southern Hebei province centered in the Shijiazhuang, Xingtai, and Handan cities, whereas the Yanshan-Taihangshan mountainous areas were the regions with lower values. In addition, the most heavily polluted season was winter, followed by autumn and spring, and summer was the cleanest season. The spatiotemporal distribution prediction results of high-precision PM_2.5 concentrations provided by the STLME model provide a scientific basis for the health risk assessment of PM_2.5 pollution in the BTH region and also provide a scientific reference for the identification of air pollution sources.