Zhang Yuhuan

Using support vector regression to predict PM10 and PM2.5

Support vector machine (SVM), as a novel and powerful machine learning tool, can be used for the prediction of PM10 and PM2.5 (particulate matter less or equal than 10 and 2.5 micrometer) in the atmosphere. This paper describes the development of a successive over relaxation support vector regress (SOR-SVR) model for the PM10 and PM2.5 prediction, based on the daily average aerosol optical depth (AOD) and meteorological parameters (atmospheric pressure, relative humidity, air temperature, wind speed), which were all measured in Beijing during the year of 2010–2012. The Gaussian kernel function, as well as the k-fold crosses validation and grid search method, are used in SVR model to obtain the optimal parameters to get a better generalization capability. The result shows that predicted values by the SOR-SVR model agree well with the actual data and have a good generalization ability to predict PM10 and PM2.5. In addition, AOD plays an important role in predicting particulate matter with SVR model, which should be included in the prediction model. If only considering the meteorological parameters and eliminating AOD from the SVR model, the prediction results of predict particulate matter will be not satisfying.

Remote sensing of atmospheric PM2.5 from high spatial resolution image of Chinese environmental satellite HJ-1/ CCD data

Due to lack of the middle infrared wavelength, it is difficult to employ the classic dark target method to retrieve aerosol optical depth (AOD) from Chinese Environmental satellite HJ1-CCD data. Therefore, focusing on extracting weak information from mixed satellite signals of atmosphere and land surface, we developed the Multi-wavelength, Multi-sensor, Multi-day (3M) approach in order to utilize maximally the observation information, and with the consideration of a prior knowledge, e.g. the surface property changes quickly with location but slowly with time. We present the AOD retrieval algorithm based on HJ1-CCD blue and green bands, based on the look-up table approach constructed using 6S radiative transfer model to provide the simultaneous determination of AOD and the ground reflectance. The aerosol and particle size information obtained from ground-based sun-sky radiometer are then used to estimate PM2.5 on the surface level, while air humidity and height of planetary boundary layer from reanalysis data are employed to improve the correlation between AOD and PM2.5. Validation of the retrieval results with different spatial resolutions (300, 500 and 1000 m), are also performed by comparison with ground-based measurements at three sites of North China regions.

基于地理加权模型的我国冬季PM 2.5 遥感估算方法研究

为了分析冬季我国区域范围内近地面PM2.5质量浓度时空分布特征,根据卫星遥感反演PM2.5质量浓度的基本原理,综合考虑我国不同地区的PM2.5污染特征的空间差异性,基于卫星遥感、气象模式资料及同期地面观测的PM2.5质量浓度数据采用地理加权模型进行回归分析,研究构建了我国区域范围内近地面PM2.5遥感反演模型.结果表明:在冬季暗像元反演AOD算法受限制的情况下,深蓝算法产品可以一定程度上弥补暗像元算法的不足,将二者有效融合能同时提高AOD产品的精度和空间覆盖度;利用地理加权回归模型进行全国区域PM2.5遥感估算,既能体现全国PM2.5时空分布的全局变化特性,又能从局部体现全国PM2.5组分、污染程度及垂直分布结构特征的空间差异特性,基于地理加权回归模型的PM2.5遥感反演结果(R2=0.7)明显优于多元线性回归模型(R2=0.56);2013年12月-2014年2月份全国PM2.5空间分布呈现明显的区域特征,PM2.5浓度较高的地方主要分布在华北南部、长三角中部和北部、华中东部及四川东部等地,西部和北部地区PM2.5污染相对较轻;从时间变化来看,全国冬季12月份PM2.5污染最重,1月份次之,2月份相对最低.这可为全国PM2.5区域联防联控提供有力的信息支撑.

Assessment of aerosol models to AOD retrieval from HJ1 Satellites

The Chinese environmental satellites HJ1 A and B can play a significant role in the aerosol retrieval due to their high spatial and temporal resolution. The current Aerosol Optical Depth (AOD) retrieval methods from HJ1-CCD are almost based on the LUT (Look-Up Table), by selecting the best fitting result to determine the AOD. However, aerosol model selection has an important impact on the retrieval results when creating the lookup table; inappropriate choice of aerosol model will significantly affect the accuracy and applicability of the method. This paper determined the local aerosol physical properties (such as complex refractive index, and size distribution) based on the observational data, thus we defined the aerosol type and retrieved the AOD of the local aerosol. Furthermore we compared the results retrieved from the measurement aerosol model with those retrieved from the inherent aerosol model in the radiative transfer model and then evaluate its effect on the aerosol type.