A Method for Estimating the Total Retrieved Ground Ozone Based on Deep Neural Network

In view of the problems of time-consuming and poor accuracy in the estimation of the total amount of ground ozone inversion by the traditional estimation method, a ground ozone inversion total amount estimation method based on depth neural network is proposed. The total amount of surface ozone inversion is preliminarily estimated. First, satellite observations are simulated to determine the amount of radiation transfer. Through the forward modeling and inversion process, the simulation results are compared with the satellite observation results to obtain the total ozone preliminary valuation. Based on the deep neural network model, a training model for estimating the total amount of ground ozone inversion is constructed, and the training effect of the network model is improved by adaptively changing the learning efficiency, and then the correct total amount of ozone is obtained through the iterative process. The experimental results show that the estimated time of this method is 4.5 s when the total amount of ozone recovered on the ground is 120 m3/hr. In five experiments, the average estimation accuracy of the total surface ozone retrieved by this method is 94.5%. It is proved that the estimation method of total surface ozone based on depth neural network inversion is effective.