摘要
文章选取MODIS数据,利用暗像元算法反演得到气溶胶光学厚度,利用BP神经网络算法通过网络训练和验证,得出PM10浓度遥感监测模型。利用该模型反演得到贵州省2014年3、7、10、12四个典型月份的PM10浓度值。结果表明模型训练和验证PM10浓度模拟值与实测值相关性系数(r)分别为0.76和0.62,利用此模型监测贵州省PM10近地面浓度是可行的;贵州省夏、秋季PM10浓度较低,春、冬季PM10浓度较高;贵州省的PM10浓度整体较低,空气质量较好。
This article selects the MODIS data using Dark Pixel Method to get the aerosol optical thickness.Then use the BP neural network to promote PM10 remote sensing monitoring model by network training and validation.This model was used to in-verse PM10 concentration in March, July, October and December four typical months of 2014 in Guizhou Province.Research re-sults show that the measured value are 0.76 and 0.62 in both training and validation.Above all, the model fits for Guizhou PM 10 monitoring.In summer and autumn, PM10 concentration is lower than in spring and winter.The concentration of PM10 in Guizhou Province is relatively low, and the air quality is better.
出处
《环境科学与管理》
CAS
2015年第8期114-118,共5页
Environmental Science and Management
基金
贵州省重大科技专项<"数字环保"关键技术研究及应用示范>项目(黔科合重大专项字[2012]6007)
