A Multi-task Two-stream Spatiotemporal Convolutional Neural Network for Convective Storm Nowcasting
This work addresses the problem of local, short-term prediction of severe convective storms for meteorologists and emergency responders, representing an incremental improvement in machine learning methods for weather forecasting.
The paper tackles convective storm nowcasting by developing a multi-task two-stream convolutional neural network that uses radar and satellite data to predict severe storms, achieving improved classification accuracy and reduced training time compared to existing recurrent neural networks.
The goal of convective storm nowcasting is local prediction of severe and imminent convective storms. Here, we consider the convective storm nowcasting problem from the perspective of machine learning. First, we use a pixel-wise sampling method to construct spatiotemporal features for nowcasting, and flexibly adjust the proportions of positive and negative samples in the training set to mitigate class-imbalance issues. Second, we employ a concise two-stream convolutional neural network to extract spatial and temporal cues for nowcasting. This simplifies the network structure, reduces the training time requirement, and improves classification accuracy. The two-stream network used both radar and satellite data. In the resulting two-stream, fused convolutional neural network, some of the parameters are entered into a single-stream convolutional neural network, but it can learn the features of many data. Further, considering the relevance of classification and regression tasks, we develop a multi-task learning strategy that predicts the labels used in such tasks. We integrate two-stream multi-task learning into a single convolutional neural network. Given the compact architecture, this network is more efficient and easier to optimize than existing recurrent neural networks.