Pereira Gavin, Lee Hyung Joo, Bell Michelle, Regan Annette, Malacova Eva, Mullins Ben, Knibbs Luke D
School of Public Health, Curtin University, WA, Australia.
NASA Ames Research Center, CA, USA.
Environ Res. 2017 Nov;159:9-15. doi: 10.1016/j.envres.2017.07.044. Epub 2017 Jul 28.
Estimating exposure to particulate matter (PM) air pollution concentrations in Australia is challenging due to relatively few monitoring sites relative to the geographic distribution of the population. We modelled daily ground-level PM concentrations for the period 2006-2011 for Australia using linear mixed models with satellite remote-sensed AOD, land-use and geographical variables as predictors. The variation in daily PM explained by the model was 51% for Australia overall, and ranged from 51% for Tasmania to 78% for South Australia. Cross-validation indicated that the models were most suitable for prediction in New South Wales and Victoria and least suitable for prediction in Western Australia, the Australian Capital Territory and Tasmania. Most of the variation in PM concentrations was explained by temporal rather than spatial variation. The inclusion of AOD and other predictors did not substantially improve model performance. Temporal models were sufficient to account for daily PM variation recorded by statutory monitors.
由于相对于澳大利亚人口的地理分布而言，监测站点相对较少，因此估算澳大利亚空气中颗粒物（PM）污染浓度具有挑战性。我们使用线性混合模型，以卫星遥感气溶胶光学厚度（AOD）、土地利用和地理变量作为预测因子，对澳大利亚2006 - 2011年期间的每日地面PM浓度进行了建模。该模型解释的澳大利亚每日PM变化总体为51%，范围从塔斯马尼亚的51%到南澳大利亚的78%。交叉验证表明，这些模型最适合在新南威尔士州和维多利亚州进行预测，最不适合在西澳大利亚州、澳大利亚首都领地和塔斯马尼亚州进行预测。PM浓度的大部分变化是由时间变化而非空间变化所解释的。纳入AOD和其他预测因子并没有显著提高模型性能。时间模型足以解释法定监测器记录的每日PM变化。
