CNN-SVR用于MAX-DOAS预测对流层NO_(2)廓线  

Prediction of Tropospheric NO_(2) Profile Using CNN-SVR-Based MAX-DOAS

作　　者：潘屹峰 田鑫 谢品华[2,3,4] 李昂 徐晋 任博 黄骁辉 田伟 王子杰 Pan Yifeng;Tian Xin;Xie Pinhua;Li Ang;Xu Jin;Ren Bo;Huang Xiaohui;Tian Wei;Wang Zijie(Information Materials and Intelligent Sensing Laboratory of Anhui Province,Institutes of Physical Science and Information Technology,Anhui University,Hefei 230601,Anhui,China;Key Laboratory of Environmental Optics and Technology,Anhui Institute of Optics and Fine Mechanics,Chinese Academy of Sciences,Hefei 230031,Anhui,China;CAS Center for Excellence in Regional Atmospheric Environment,Xiamen 361021,Fujian,China;School of Environmental Science and Optoelectronic Technology,University of Science and Technology of China,Hefei 230026,Anhui,China)

机构地区：[1]安徽大学物质科学与信息技术研究院安徽省信息材料与智能传感实验室,安徽合肥230601 [2]中国科学院安徽光学精密机械研究所环境光学与技术重点实验室,安徽合肥230031 [3]中国科学院区域大气环境研究卓越创新中心,福建厦门361021 [4]中国科学技术大学环境科学与光电技术学院,安徽合肥230026

出　　处：《光学学报》2022年第24期1-9,共9页Acta Optica Sinica

基　　金：国家自然科学基金(42105132,U19A2044,42030609);安徽省自然科学基金(2008085QD183,2008085QD182)。

摘　　要：提出一种基于卷积神经网络(CNN)和支持向量回归机(SVR)的多轴差分光学吸收光谱(MAX-DOAS)对流层NO_(2)垂直分布预测方法。将2019年南京站点采集的原始MAX-DOAS数据通过QDOAS软件拟合获取O_(4)和NO_(2)差分斜柱浓度,结合基于最优估算的气溶胶和痕量气体廓线反演算法——PriAM算法反演了对流层NO_(2)廓线,并将其作为预测模型的输出。此外,通过平均影响值方法进行预测模型输入变量的选择,确定了MAX-DOAS数据、温度、气溶胶光学厚度和低云覆盖率为模型的最佳输入变量。通过实验优化网络结构和参数,最终建立预测模型在测试集与PriAM的平均百分比误差仅为9.14%,与单独建立的CNN、SVR、反向传播模型相比,平均百分比误差分别降低了8.22%、6.00%、32.28%。因此,CNN-SVR能够利用MAX-DOAS数据对对流层NO_(2)廓线进行有效预测。This study proposes a method based on a convolutional neural network(CNN)and support vector regression machine(SVR)for predicting the vertical distribution of NO_(2)in the troposphere by multi-axis differential optical absorption spectroscopy(MAX-DOAS)technology.Taking the Nanjing site as an example,we obtain the O_(4) and NO_(2)differential slant column density(dSCD)according to the raw MAX-DOAS data collected by QDOAS fitting in 2019,invert the tropospheric NO_(2)profile by combining the optimal estimation-based aerosol and trace gas profile inversion algorithmPriAM,and use the profile as the output of the prediction model.In addition,the input variables of the prediction model are selected by the mean impact value method,with MAX-DOAS data,temperature,aerosol optical thickness,and low cloud coverage finally identified as the optimal input variables for the model.Furthermore,the network structure and parameters are optimized through experiments,and the average percentage error of the final CNN-SVR prediction model in the test set with PriAM is only 9.14%,which is 8.22%,6.00%,and 32.28%lower than that of the separately constructed CNN,SVR,and backpropagation models,respectively.Therefore,CNN-SVR can effectively predict tropospheric NO_(2)profiles by using MAX-DOAS data.

关 键 词：大气光学 卷积神经网络 支持向量回归机 多轴差分吸收光谱 对流层NO_(2)廓线 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]