A Case Analysis of Dust Weather and Prediction of PM10 Concentration Based on Machine Learning at the Tibetan Plateau

Dust weather is common and disastrous at the Tibetan Plateau. This study selected a typical case of dust analyzed its main development mechanism in northeast Plateau, then applied six machine learning methods time series regression model to predict PM10 concentration this area. The results showed that: (1) 24-h pressure change was positive when front intruded on surface; convergence vector winds with sudden drop temperature humidity led by trough 700 hPa; “two troughs one ridge” situation appeared 500 hPa while cold advection behind strong cyclone vorticity formed east Inner Mongolia. (2) trajectory air mass from Hexi Corridor path influencing Xining City, case, since significant lag peak City compared Zhangye City. (3) Multiple Linear Regression not only timely effective predicting but had great abilities for anticipating transition period particle appearance date maximum values such weather. (4) MA MP clean were much lower than that period; as an eigenvalue played important role even periods. Different periods, prediction effect Random Forest Optimized Bayesian hyperparameter superior