Shi Yan, Liu Rui-Mei, Luo Yi, Yang Kun
School of Information Science and Technology, Yunnan Normal University, Kunming 650500, China.
GIS Technology Engineering Research Centre for West-China Resources and Environment, Ministry of Education, Yunnan Normal University, Kunming 650500, China.
Huan Jing Ke Xue. 2020 Jan 8;41(1):1-13. doi: 10.13227/j.hjkx.201905109.
We use measured aerosol fine particulate matter (PM) data, moderate resolution imaging spectroradiometer (MODIS) aerosol optical depth (AOD) data, and meteorological parameters (temperature, wind speed, wind direction, boundary layer height, and relative humidity) from the Chinese national control monitoring network, to consider seasonal and regional differences in the relationship between AOD and PM. We propose a two-stage combined estimation model of PM concentrations based on the -support vector regression (-SVR/Epsilon-SVR) and the Mind Evolutionary Computation-BP neural network (MEC-BP) for analyzing spatiotemporal variations in PM concentrations in China between 2000 and 2017. The results showed that the two-stage combined estimation model provided a reliable estimation of the monthly ground-level PM concentrations at a spatial resolution of 1°×1° during 2000-2017 in China. This effectively offsets the time and space gaps in the current data sets of the ground monitoring network (=0.838, root mean square errors (RMSE)=11.512 μg·m, mean absolute percentage error (MAPE)=14.905%, mean squared percentage error (MSPE)=0.243%, mean absolute error (MAE)=6.476 μg·m, mean squared error (MSE)=132.519 μg·m). The preliminary spatiotemporal analysis results showed that:① Over the period 2000-2017, 2014 represented an important demarcation point for the annual PM concentration, as its trend changed from one of continuous increase to one of rapid decrease. The PM concentration decreases more rapidly in areas with high concentrations of PM in particular, including the northern coastal area, the eastern coastal area, and the middle reaches of the Changjiang River. ② During the studied period, the annual average PM concentration exceeded the second level criterion of the Chinese national air quality standard (35 μg·m) over more than 65% of China. Although the PM pollution situation in China improved to a certain extent in the latter years of the studied period, the air pollution situation remained poor.
我们使用了来自中国国家环境监测网的实测气溶胶细颗粒物（PM）数据、中分辨率成像光谱仪（MODIS）气溶胶光学厚度（AOD）数据以及气象参数（温度、风速、风向、边界层高度和相对湿度），以研究AOD与PM关系中的季节和区域差异。我们提出了一种基于支持向量回归（-SVR/ε-SVR）和思维进化计算-BP神经网络（MEC-BP）的PM浓度两阶段组合估计模型，用于分析2000年至2017年中国PM浓度的时空变化。结果表明，该两阶段组合估计模型能够可靠地估计2000 - 2017年中国1°×1°空间分辨率下的月均地面PM浓度。这有效地弥补了地面监测网络现有数据集在时间和空间上的差距（相关系数=0.838，均方根误差（RMSE）=11.512μg·m，平均绝对百分比误差（MAPE）=14.905%，均方百分比误差（MSPE）=0.243%，平均绝对误差（MAE）=6.476μg·m，均方误差（MSE）=132.519μg·m）。初步的时空分析结果显示：①在2000 - 2017年期间，2014年是年PM浓度的一个重要分界点，其趋势从持续上升转变为快速下降。特别是在PM高浓度地区，包括北部沿海地区、东部沿海地区和长江中游地区，PM浓度下降更为迅速。②在研究期间，中国超过65%的地区年平均PM浓度超过了中国国家空气质量标准的二级标准（35μg·m）。尽管在研究后期中国的PM污染状况有一定程度的改善，但空气污染状况仍然严峻。
Zotero 插件