灾害学2025，Vol.40Issue(4)：37-43,70,8.DOI:10.3969/j.issn.1000-811X.2025.04.007
基于GNSS层析水汽与大气污染物的PM2.5浓度预测
Prediction of PM2.5 Concentration Based on GNSS Tomography of Precipitable Water Vapor and Atmospheric Pollutants
摘要
Abstract
To accurately carry out spatiotemporal prediction of PM2.5 concentration,FFT and ConvLSTM models are used to predict PM2.5 concentration based on GNSS tomography of precipitable water vapor(PWV)and atmospheric pollutants.For GNSS observation data,the tomographic equation system is composed of tomo-graphic observation matrix and constraint matrix components,and algebraic reconstruction method is selected for equation solving to obtain GNSS tomographic PWV.The correlation between various elements and PM2.5 con-centration is analyzed.Based on ConvLSTM grid model,spatiotemporal prediction of PM2.5 concentration for the next 24 hours is achieved by using FFT to provide the common variation period of each element.It is found that the PWV values from the first to seventh layer combinations account for over 90%of the total PWV values of the entire layer,and each layer's PWV is positively correlated with PM2.5 concentration.The prediction effect of PM2.5 concentration using GNSS layered PWV from the first to seventh layer is better than that using PWV from the entire layer.Using PWV data within the atmospheric pressure range of 1000 hPa to 500 hPa can better reflect the impact of meteorological processes on PWV,thereby improving the accuracy of predicting PM2.5 concentration.关键词
GNSS层析/分层水汽/FFT/ConvLSTM/PM2.5浓度Key words
GNSS tomography/layered water vapor/FFT/ConvLSTM/PM2.5 concentration分类
资源环境引用本文复制引用
刘严萍,张培晅,吴文坛,王勇,李隆熙,刘春宏..基于GNSS层析水汽与大气污染物的PM2.5浓度预测[J].灾害学,2025,40(4):37-43,70,8.基金项目
天津市轨道交通导航定位及时空大数据技术重点实验室开放课题"基于深度学习的多因素融合短临突发降雨预报系统"(TKL2025A02) （TKL2025A02）
天津市科技计划项目"面向青少年群体的北斗卫星导航系统专题科普宣传(C)"(21KPHDRC00070) （C）
