Abstract：
BACKGROUND: Exposure to green space can protect against poor health through a variety of mechanisms. However, there is heterogeneity in methodological approaches to exposure assessments which makes creating effective policy recommendations challenging.
OBJECTIVE: Critically evaluate the use of a satellite-derived exposure metric, the Enhanced Vegetation Index (EVI), for assessing access to different types of green space in epidemiological studies.
METHODS: We used Landsat 5-8 (30 m resolution) to calculate average EVI for a 300 m radius surrounding 1.4 million households in Wales, UK for 2018. We calculated two additional measures using topographic vector data to represent access to green spaces within 300 m of household locations. The two topographic vector-based measures were total green space area stratified by type and average private garden size. We used linear regression models to test whether EVI could discriminate between publicly accessible and private green space and Pearson correlation to test associations between EVI and green space types.
RESULTS: Mean EVI for a 300 m radius surrounding households in Wales was 0.28 (IQR = 0.12). Total green space area and average private garden size were significantly positively associated with corresponding EVI measures (β = < 0.0001, 95% CI: 0.0000, 0.0000; β = 0.0001, 95% CI: 0.0001, 0.0001 respectively). In urban areas, as average garden size increases by 1 m2, EVI increases by 0.0002. Therefore, in urban areas, to see a 0.1 unit increase in EVI index score, garden size would need to increase by 500 m2. The very small β values represent no \'measurable real-world\' associations. When stratified by type, we observed no strong associations between greenspace and EVI.
CONCLUSIONS: It is a widely implemented assumption in epidiological studies that an increase in EVI is equivalent to an increase in greenness and/or green space. We used linear regression models to test associations between EVI and potential sources of green reflectance at a neighbourhood level using satellite imagery from 2018. We compared EVI measures with a \'gold standard\' vector-based dataset that defines publicly accessible and private green spaces. We found that EVI should be interpreted with care as a greater EVI score does not necessarily mean greater access to publicly available green spaces in the hyperlocal environment.
OBJECTIVE: Critically evaluate the use of a satellite-derived exposure metric, the Enhanced Vegetation Index (EVI), for assessing access to different types of green space in epidemiological studies.
METHODS: We used Landsat 5-8 (30 m resolution) to calculate average EVI for a 300 m radius surrounding 1.4 million households in Wales, UK for 2018. We calculated two additional measures using topographic vector data to represent access to green spaces within 300 m of household locations. The two topographic vector-based measures were total green space area stratified by type and average private garden size. We used linear regression models to test whether EVI could discriminate between publicly accessible and private green space and Pearson correlation to test associations between EVI and green space types.
RESULTS: Mean EVI for a 300 m radius surrounding households in Wales was 0.28 (IQR = 0.12). Total green space area and average private garden size were significantly positively associated with corresponding EVI measures (β = < 0.0001, 95% CI: 0.0000, 0.0000; β = 0.0001, 95% CI: 0.0001, 0.0001 respectively). In urban areas, as average garden size increases by 1 m2, EVI increases by 0.0002. Therefore, in urban areas, to see a 0.1 unit increase in EVI index score, garden size would need to increase by 500 m2. The very small β values represent no \'measurable real-world\' associations. When stratified by type, we observed no strong associations between greenspace and EVI.
CONCLUSIONS: It is a widely implemented assumption in epidiological studies that an increase in EVI is equivalent to an increase in greenness and/or green space. We used linear regression models to test associations between EVI and potential sources of green reflectance at a neighbourhood level using satellite imagery from 2018. We compared EVI measures with a \'gold standard\' vector-based dataset that defines publicly accessible and private green spaces. We found that EVI should be interpreted with care as a greater EVI score does not necessarily mean greater access to publicly available green spaces in the hyperlocal environment.
摘要：
背景：暴露于绿色空间可以通过多种机制防止不良健康。然而,暴露评估的方法存在异质性，这使得创建有效的政策建议具有挑战性。
目标：严格评估使用卫星衍生的暴露度量，增强植被指数（EVI），在流行病学研究中评估对不同类型绿色空间的访问。
方法：我们使用Landsat5-8（30m分辨率）计算了威尔士140万个家庭周围300m半径的平均EVI，2018年英国。我们使用地形矢量数据计算了另外两个度量，以表示访问家庭位置300m内的绿色空间。两种基于地形矢量的措施是按类型和平均私人花园大小分层的总绿地面积。我们使用线性回归模型来测试EVI是否可以区分公共可访问和私人绿地以及Pearson相关性来测试EVI和绿地类型之间的关联。
结果：威尔士300m半径周围家庭的平均EVI为0.28（IQR=0.12）。总绿地面积和平均私人花园面积与相应的EVI指标显着相关（β=<0.0001，95％CI：0.0000，0.0000；β=0.0001，95％CI：0.0001，0.0001）。在城市地区,随着平均花园面积增加1平方米，EVI增加0.0002。因此,在城市地区，要看到EVI指数得分增加0.1个单位，花园大小需要增加500平方米。非常小的β值表示没有“可测量的真实世界”关联。当按类型分层时，我们没有观察到绿色空间和EVI之间的强烈关联。
结论：这是一个在表观学研究中广泛实施的假设，即EVI的增加等同于绿色和/或绿色空间的增加。我们使用线性回归模型，使用2018年的卫星图像测试EVI与邻居水平的潜在绿色反射率来源之间的关联。我们将EVI措施与“黄金标准”基于矢量的数据集进行了比较，该数据集定义了可公开访问和私人绿色空间。我们发现，EVI应谨慎解释为EVI得分更高并不一定意味着可以在超本地环境中更多地访问公共可用的绿色空间。
目标：严格评估使用卫星衍生的暴露度量，增强植被指数（EVI），在流行病学研究中评估对不同类型绿色空间的访问。
方法：我们使用Landsat5-8（30m分辨率）计算了威尔士140万个家庭周围300m半径的平均EVI，2018年英国。我们使用地形矢量数据计算了另外两个度量，以表示访问家庭位置300m内的绿色空间。两种基于地形矢量的措施是按类型和平均私人花园大小分层的总绿地面积。我们使用线性回归模型来测试EVI是否可以区分公共可访问和私人绿地以及Pearson相关性来测试EVI和绿地类型之间的关联。
结果：威尔士300m半径周围家庭的平均EVI为0.28（IQR=0.12）。总绿地面积和平均私人花园面积与相应的EVI指标显着相关（β=<0.0001，95％CI：0.0000，0.0000；β=0.0001，95％CI：0.0001，0.0001）。在城市地区,随着平均花园面积增加1平方米，EVI增加0.0002。因此,在城市地区，要看到EVI指数得分增加0.1个单位，花园大小需要增加500平方米。非常小的β值表示没有“可测量的真实世界”关联。当按类型分层时，我们没有观察到绿色空间和EVI之间的强烈关联。
结论：这是一个在表观学研究中广泛实施的假设，即EVI的增加等同于绿色和/或绿色空间的增加。我们使用线性回归模型，使用2018年的卫星图像测试EVI与邻居水平的潜在绿色反射率来源之间的关联。我们将EVI措施与“黄金标准”基于矢量的数据集进行了比较，该数据集定义了可公开访问和私人绿色空间。我们发现，EVI应谨慎解释为EVI得分更高并不一定意味着可以在超本地环境中更多地访问公共可用的绿色空间。
