Strong Convective Weather Extrapolation Based on VAN-ConvLSTM

Inhalt:
Strong convective weather forecasting is an extension of traditional weather forecasting. The accuracy of short-term forecasts of strong convective weather becomes particularly important in early warnings. However, existing forecasts of severe convective weather often have difficulty responding accurately and quickly to weather hazards, while predictions are relatively inaccurate. In this paper, we propose a deep learning model. The model mixes multilayer architectures, including a deep visual analog making network (VAN) and a convolutional long and short term memory network (ConvLSTM). We use ConvLSTM to process the spatial features of the temporal images, while adding VAN to perform image analogy on the single frame image passed into the network. With this method, a continuous image of the same spatial output can be obtained based on a time series. The experimental results show that the prediction model combining VAN and ConvLSTM proposed in this paper has better prediction performance than other existing methods.