Daily and Annual PM2.5 Concentrations for the Contiguous United States, 1-km Grids, Version 1.10 (2000-2016)

The Daily and Annual PM2.5 Concentrations for the Contiguous United States, 1-km Grids, Version 1.10 (2000-2016) data set includes predictions of PM2.5 concentration in grid cells at a resolution of 1-km for the years 2000-2016. A generalized additive model was used that accounted for geographic difference to ensemble daily predictions of three machine learning models: neural network, random forest, and gradient boosting. The three machine learners incorporated multiple predictors, including satellite data, meteorological variables, land-use variables, elevation, chemical transport model predictions, several reanalysis data sets, and others. The annual predictions were calculated by averaging the daily predictions for each year in each grid cell. The ensembled model demonstrated better predictive performance than the individual machine learners with 10-fold cross-validated R-squared values of 0.86 for daily predictions and 0.89 for annual predictions. In version 1.10, the completeness of daily PM2.5 predictions have been enhanced by employing linear interpolation to impute missing values. Specifically, for days with small spatial patches of missing data with less than 100 grid cells, inverse distance weighting interpolation was used to fill the missing grid cells. Other missing daily PM2.5 predictions were interpolated from the nearest days with available data. Annual predictions were updated by averaging the imputed daily predictions for each year in each grid cell. These daily and annual PM2.5 predictions allow public health researchers to respectively estimate the short- and long-term effects of PM2.5 exposures on human health, supporting the U.S. Environmental Protection Agency (EPA) for the revision of the National Ambient Air Quality Standards for 24-hour average and annual average concentrations of PM2.5. The data are available in RDS and GeoTIFF formats for statistical research and geospatial analysis.

该数据集《美国大陆每日及年度PM2.5浓度，1公里网格，版本1.10（2000-2016）》包含了2000至2016年间，以1公里分辨率为网格单元的PM2.5浓度预测。采用了一种考虑地理差异的广义加性模型，以集成三种机器学习模型的每日预测：神经网络、随机森林和梯度提升。三种机器学习算法综合了多种预测因子，包括卫星数据、气象变量、土地利用变量、海拔、化学传输模型预测、多个再分析数据集以及其他因素。年度预测通过计算每个网格单元每年每日预测的平均值得出。集成模型在预测性能上优于单一机器学习模型，其10折交叉验证的R平方值分别为每日预测0.86和年度预测0.89。在版本1.10中，通过线性插值法补充缺失值，提升了每日PM2.5预测的完整性。具体而言，对于小于100个网格单元的小面积缺失数据，采用了逆距离加权插值法填充缺失的网格单元。其他缺失的每日PM2.5预测则通过最近有可用数据的日子的预测值进行插值。年度预测通过平均每个网格单元每年补充后的每日预测值进行更新。这些每日及年度PM2.5预测数据，使得公共卫生研究人员能够分别估算PM2.5暴露对人类健康的短期和长期影响，为美国环境保护署（EPA）修订国家环境空气质量标准中24小时平均和年度平均PM2.5浓度提供支持。数据以RDS和GeoTIFF格式提供，适用于统计分析与地理空间分析。