Research On Prediction Methods For Atmospheric Vortex Intensity Based On Deep Learning

The intensity variation of vortex intensity plays an important role in the atmosphere and ocean,The intensity variation of South Asian high（SAH）and polar vortex play an important role in the formation and maintenance of many mesoscale systems.These two kinds of vortex are important factors affecting extreme weather,precipitation and atmospheric chemical and physical processes in the Northern Hemisphere and even the world.However,due to the nonlinear characteristics of vortex system,they are difficult to predict,and previous studies focus on the variation of SAH and polar vortex daily intensity is poorly understood.This study verified the significance of vortex prediction technology by analyzing the impact of SAH intensity on chemical distribution.In order to predict the variation of SAH intensity,the SAH intensity time series data set are constructed and then pre-trained by combining the Ensemble Empirical Mode Decomposition with Adaptive Noise（CEEMDAN）method which has the ability to deal with nonlinear and unstable single with Permutation Entropy（PE）method which can extract the SAH intensity feature of IMF decomposed by CEEMDAN,and the Convolution Based Gated Recurrent Neural Network（Conv GRU）model is used to train and predict the intensity of SAH,and it has been compared with various kinds of efficient time series prediction models at last.The prediction results show that the combination of CEEMDAN and Conv GRU have a higher accuracy and more stable prediction ability than the traditional deep learning model,which proves that the method has good applicability for the prediction of nonlinear systems in the atmosphere.By analyzing the distribution characteristics of O₃ and CO in SAH strong and weak cycles（SP and WP）,both O₃ and CO are found to be more concentrated within the SAH at 100 h Pa during SP.The analysis shows that the O₃-poor air within the anticyclone is relatively hard to escape from the inside SAH during SPs.These results indicate that the intraseasonal variation of the SAH intensity can significantly affect the chemical distribution in the upper troposphere and lower stratosphere region during the Asian summer monsoon season.Ultimately,we established a deep learning model for multi-day and long-time intensity prediction of the polar vortex.Focusing on the winter period with the strongest polar vortex intensity,reanalysis data set is used to construct polar vortex anomaly distribution images and polar vortex intensity time series.Then we propose a new Convolution Neural Network with Long Short-term Memory Based on Gaussian Smoothing（GSCNN-LSTM）model which not only can accurately predict the variation characteristics of polar vortex intensity from day to day,but also can produce a skillful forecast for lead times of up to 20 days.Moreover,the innovative GSCNN-LSTM model has better stability and skillful correlation prediction than the traditional and some advanced spatiotemporal sequence prediction models.The accuracy of the model suggests important implications that DL methods have good applicability in forecasting the nonlinear system and vortex spatial-temporal characteristics variation in the atmosphere.