北半球基于近地面能见度的PM2.5浓度（1959-2024）

1）数据内容：1959-2024年北半球5000多个站点逐日的PM2.5浓度数据。2）数据来源及加工方法：基于地面能见度及其他气象观测数据，使用机器学习的方法估计每日PM2.5浓度。其中，每个站点的PM2.5浓度的时间序列的前80%的数据用于训练，剩下的20%的数据用于测试。3）数据质量描述：训练结果显示估计的PM2.5与浓度监测的PM2.5浓度之间的斜率（95%置信区间）为0.955±0.0002，R2为0.95，RMSE为7.2μg/m3，MAE为3.2μg/m3。测试结显示预测的PM2.5浓度与监测的PM2.5浓度之间的斜率（95%置信区间）为0.864±0.0010，R2为0.79，RMSE为14.8μg/m3，MAE为7.6μg/m3。 4）数据应用成果及前景：补充PM2.5的历史数据，为研究大气环境、人类健康和气候变化以及模式同化提供支持。

1) Data Content: Daily PM2.5 concentration data from more than 5,000 monitoring sites in the Northern Hemisphere spanning from 1959 to 2022.
2) Data Source and Processing Method: Daily PM2.5 concentrations were estimated using machine learning methods based on ground visibility and other meteorological observation data. Specifically, the first 80% of the time series PM2.5 concentration data at each site were used for model training, while the remaining 20% were reserved for model testing.
3) Data Quality Description: The training results show that the slope (95% confidence interval) between the estimated PM2.5 concentrations and the in-situ monitored PM2.5 concentrations is 0.955±0.0002, with an R² of 0.95, a RMSE of 7.2 μg/m³, and a MAE of 3.2 μg/m³. The test results indicate that the slope (95% confidence interval) between the predicted PM2.5 concentrations and the monitored PM2.5 concentrations is 0.864±0.0010, with an R² of 0.79, a RMSE of 14.8 μg/m³, and a MAE of 7.6 μg/m³.
4) Data Application Achievements and Prospects: This dataset supplements historical PM2.5 data, providing support for research on atmospheric environment, human health, climate change, and model assimilation.

搜集汇总

背景与挑战

背景概述

该数据集提供了1959-2022年北半球5000多个站点逐日的PM2.5浓度数据，基于地面能见度及其他气象观测数据，使用机器学习方法估计每日PM2.5浓度。数据质量较高，训练和测试结果显示估计的PM2.5浓度与监测值之间的相关性良好，适用于研究大气环境、人类健康和气候变化等领域。