Centro de Recursos Naturais e Ambiente, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais, 1049-001, Lisbon, Portugal.
Centro de Recursos Naturais e Ambiente, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais, 1049-001, Lisbon, Portugal; CE3C, Centre for Ecology, Evolution and Environmental Changes, Faculdade de Ciências, Universidade de Lisboa, Campo Grande Bloco C2 5° Piso, 1749-016 Lisbon, Portugal.
Sci Total Environ. 2016 Aug 15;562:740-750. doi: 10.1016/j.scitotenv.2016.04.081. Epub 2016 Apr 22.
In most studies correlating health outcomes with air pollution, personal exposure assignments are based on measurements collected at air-quality monitoring stations not coinciding with health data locations. In such cases, interpolators are needed to predict air quality in unsampled locations and to assign personal exposures. Moreover, a measure of the spatial uncertainty of exposures should be incorporated, especially in urban areas where concentrations vary at short distances due to changes in land use and pollution intensity. These studies are limited by the lack of literature comparing exposure uncertainty derived from distinct spatial interpolators. Here, we addressed these issues with two interpolation methods: regression Kriging (RK) and ordinary Kriging (OK). These methods were used to generate air-quality simulations with a geostatistical algorithm. For each method, the geostatistical uncertainty was drawn from generalized linear model (GLM) analysis. We analyzed the association between air quality and birth weight. Personal health data (n=227) and exposure data were collected in Sines (Portugal) during 2007-2010. Because air-quality monitoring stations in the city do not offer high-spatial-resolution measurements (n=1), we used lichen data as an ecological indicator of air quality (n=83). We found no significant difference in the fit of GLMs with any of the geostatistical methods. With RK, however, the models tended to fit better more often and worse less often. Moreover, the geostatistical uncertainty results showed a marginally higher mean and precision with RK. Combined with lichen data and land-use data of high spatial resolution, RK is a more effective geostatistical method for relating health outcomes with air quality in urban areas. This is particularly important in small cities, which generally do not have expensive air-quality monitoring stations with high spatial resolution. Further, alternative ways of linking human activities with their environment are needed to improve human well-being.
在大多数将健康结果与空气污染相关联的研究中，个人暴露评估是基于空气质量监测站的测量值，而这些测量值与健康数据位置并不重合。在这种情况下，需要插值器来预测未采样位置的空气质量并分配个人暴露值。此外，还应纳入暴露的空间不确定性度量，特别是在城市地区，由于土地利用和污染强度的变化，浓度在短距离内会发生变化。这些研究受到缺乏比较来自不同空间插值器的暴露不确定性的文献的限制。在这里，我们使用两种插值方法（回归克里金法（RK）和普通克里金法（OK））来解决这些问题。这些方法用于使用地质统计学算法生成空气质量模拟。对于每种方法，地质统计学不确定性都来自广义线性模型（GLM）分析。我们分析了空气质量与出生体重之间的关联。个人健康数据（n=227）和暴露数据于 2007 年至 2010 年在葡萄牙锡尼什收集。由于城市中的空气质量监测站没有提供高空间分辨率的测量值（n=1），我们使用地衣数据作为空气质量的生态指标（n=83）。我们发现，GLM 与任何地质统计学方法的拟合都没有显著差异。然而，RK 的模型往往更频繁地拟合得更好，更频繁地拟合得更差。此外，地质统计学不确定性结果显示，RK 的平均值和精度略高。与地衣数据和高空间分辨率的土地利用数据相结合，RK 是一种更有效的地质统计学方法，可将健康结果与城市地区的空气质量联系起来。这在小城市中尤为重要，小城市通常没有昂贵的具有高空间分辨率的空气质量监测站。此外，还需要替代的方法将人类活动与其环境联系起来，以提高人类的福祉。
