Estimation of surface ozone concentration over Jiangsu province using a high-performance deep learning model

摘要

Recently,the global background concentration of ozone(O_(3))has demonstrated a rising trend.Among various methods,groun-basedmonitoring of O_(3)concentrations is highly reliable for research analysis.To obtain information on the spatial characteristics of O_(3)concentrations,it is necessary that the groundmonitoring sites be constructed in sufficient density.In recent years,many researchers have used machine learning models to estimate surface O_(3)concentrations,which cannot fully provide the spatial and temporal information contained in a sample dataset.To solve this problem,the current study utilized a deep learning model called the Residual connection Convolutional Long Short-Term Memory network(RConvLSTM)to estimate daily maximum8-hr average(MDA8)O_(3)over Jiangsu province,China during 2020.In this research,the R-ConvLSTM model not only provides the spatiotemporal information ofMDA8 O_(3),but also involves residual connection to avoid the problem of gradient explosion and gradient disappearance with the deepening of network layers.We utilized the TROPOMI total O_(3)column retrieved fromSentinel-5 Precursor,ERA5 reanalysismeteorological data,and other supplementary data to build a pre-trained dataset.The R-ConvLSTM model achieved an overall sample-base cross-validation(CV)R^(2)of 0.955 with root mean square error(RMSE)of 9.372μg/m^(3).Model estimation also showed a city-based CV R^(2)of 0.896 with RMSE of 14.029μg/m^(3),the highest MDA8 O_(3)in spring being 122.60±31.60μg/m^(3)and the lowest in winter being 69.93±18.48μg/m^(3).