Abstract:Nowadays deep learning-based weather radar echo extrapolation methods have competently improved nowcasting quality. Current pure convolutional or convolutional recurrent neural network-based extrapolation pipelines inherently struggle in capturing both global and local spatiotemporal interactions simultaneously, thereby limiting nowcasting performances, e.g., they not only tend to underestimate heavy rainfalls' spatial coverage and intensity but also fail to precisely predict nonlinear motion patterns. Furthermore, the usually adopted pixel-wise objective functions lead to blurry predictions. To this end, we propose a novel motion-guided global-local aggregation Transformer network for effectively combining spatiotemporal cues at different time scales, thereby strengthening global-local spatiotemporal aggregation urgently required by the extrapolation task. First, we divide existing observations into both short- and long-term sequences to represent echo dynamics at different time scales. Then, to introduce reasonable motion guidance to Transformer, we customize an end-to-end module for jointly extracting motion representation of short- and long-term echo sequences (MRS, MRL), while estimating optical flow. Subsequently, based on Transformer architecture, MRS is used as queries to retrospect the most useful information from MRL for an effective aggregation of global long-term and local short-term cues. Finally, the fused feature is employed for future echo prediction. Additionally, for the blurry prediction problem, predictions from our model trained with an adversarial regularization achieve superior performances not only in nowcasting skill scores but also in precipitation details and image clarity over existing methods. Extensive experiments on two challenging radar echo datasets demonstrate the effectiveness of our proposed method.

Application of Multi-channel 3D-Cube Successive Convolution Network for Convective Storm Nowcasting

A Three-dimensional Convolutional-Recurrent Network for Convective Storm Nowcasting

A Multi-Channel 3D Convolutional-Recurrent Neural Network for Convective Storm Nowcasting

Convolutional Neural Network for Convective Storm Nowcasting Using 3D Doppler Weather Radar Data

Convolutional Neural Network for Convective Storm Nowcasting Using 3-D Doppler Weather Radar Data

A spatiotemporal convolutional network approach for convection storm nowcasting

A Multi-task Two-stream Spatiotemporal Convolutional Neural Network for Convective Storm Nowcasting

Pixel-CRN: A New Machine Learning Approach for Convective Storm Nowcasting

Strong Convective Storm Nowcasting Using a Hybrid Approach of Convolutional Neural Network and Hidden Markov Model

A Deep Belief Network Approach Using VDRAS Data for Nowcasting

A Deep Learning‐Based Methodology for Precipitation Nowcasting with Radar

Application of Spatiotemporal Convolution Network for Precipitation Nowcasting

A Spatio-Temporal Graph-Guided Convolutional LSTM for Tropical Cyclones Precipitation Nowcasting.

Toward a Deep-Learning-Network-Based Convective Weather Initiation Algorithm From the Joint Observations of Fengyun-4A Geostationary Satellite and Radar for 0–1h Nowcasting

Strong Spatiotemporal Radar Echo Nowcasting Combining 3DCNN and Bi-Directional Convolutional LSTM

Extreme and severe convective weather disasters: A dual-polarization radar nowcasting method based on physical constraints and a deep neural network model

A Machine Learning Nowcasting Method Based on Real-time Reanalysis Data

3D-UNet-LSTM: A Deep Learning-Based Radar Echo Extrapolation Model for Convective Nowcasting

Motion-Guided Global–Local Aggregation Transformer Network for Precipitation Nowcasting

The Interpretation of Deep Learning for Convective Storm Nowcasting