Studying Precipitation Nowcasting Based On Random Forest And Tensor Regression

Precipitation is a common weather phenomenon,which has an important impact on social and economic development and people’s production and life.Accurate and effective precipitation forecast is not only beneficial to the planning of agricultural production development,but also to the prevention and avoidance of natural disasters.Because of the complex changes of meteorological conditions,precipitation nowcasting is always a challenging problem in weather forecasting.At the same time,precipitation data shows great imbalance,which also brings difficulties to build model.Based on the preliminary classification results of random forest and the quantitative prediction of tensor regression including intercept term,this thesis proposes a classification and regression model for precipitation nowcasting.In the first step,the model classifies data through random forest to improve the accuracy of the forecast and reduce the error.Secondly,the tensor regression including intercept term is built on the basis of classification results,and the quantitative precipitation estimation is obtained.Compared with the data centralization method in the regression model,this thesis notes that the intercept term plays an important role in improving the accuracy of quantitative precipitation estimation.Therefore,a tensor regression model including intercept term is established and the corresponding separation ordinary least squares method is derived to solve the unknown parameters in the regression model.Finally,based on the proposed model,this thesis predicts the hourly precipitation of Lanzhou,Baiyin,Dingxi,Pingliang,Tianshui and Qingyang in Gansu Province from April to September 2020.According to the meteorological data of different stations,the model shows better prediction effect in the classification prediction and the estimation of precipitation.Compared with EC fine grid data,the model has smaller mean absolute error and root mean square error.