Remote Sensing · 2022 · 20 citations · 33 references
Convolutional Neural NetworkRadar Echo ExtrapolationEngineeringEarth ScienceStorm Nowcasting3D Computer VisionNumerical Weather PredictionData ScienceImaging RadarComputational ImagingRadar Signal ProcessingVideo TransformerConvective Storm NowcastingMeteorologySynthetic Aperture RadarRadar ApplicationDeep Learning3D-convlstm Model3D Data Processing3D Object RecognitionRadar3D VisionAerospace Engineering3D-convlstm LayersRadar Image ProcessingHigh-resolution Modeling
Radar echo extrapolation has been extensively studied for precipitation and storm nowcasting, but most work has focused on two‑dimensional images, leaving three‑dimensional radar volumes largely unexplored. This study introduces a 3D‑convolutional long short‑term memory (ConvLSTM) model to perform three‑dimensional gridded radar echo extrapolation for severe storm nowcasting. The model first uses a 3D‑CNN to extract spatial features from each radar volume, then applies 3D‑ConvLSTM layers to capture spatiotemporal dynamics and generate future hidden states, which are up‑sampled by a second 3D‑CNN to produce the final nowcast. Quantitative results show the 3D‑ConvLSTM outperforms 3D optical flow and other deep‑learning baselines for storms with reflectivity above 35 and 45 dBZ, and case studies confirm more realistic storm evolution and earlier severe‑storm warnings.
Radar echo extrapolation has been widely developed in previous studies for precipitation and storm nowcasting. However, most studies have focused on two-dimensional radar images, and extrapolation of multi-altitude radar images, which can provide more informative and visual forecasts about weather systems in realistic space, has been less explored. Thus, this paper proposes a 3D-convolutional long short-term memory (ConvLSTM)-based model to perform three-dimensional gridded radar echo extrapolation for severe storm nowcasting. First, a 3D-convolutional neural network (CNN) is used to extract the 3D spatial features of each input grid radar volume. Then, 3D-ConvLSTM layers are leveraged to model the spatial–temporal relationship between the extracted 3D features and recursively generate the 3D hidden states correlated to the future. Nowcasting results are obtained after applying another 3D-CNN to up-sample the generated 3D hidden states. Comparative experiments were conducted on a public National Center for Atmospheric Research Data Archive dataset with a 3D optical flow method and other deep-learning-based models. Quantitative evaluations demonstrate that the proposed 3D-ConvLSTM-based model achieves better overall and longer-term performance for storms with reflectivity values above 35 and 45 dBZ. In addition, case studies qualitatively demonstrate that the proposed model predicts more realistic storm evolution and can facilitate early warning regarding impending severe storms.
33
Skilful precipitation nowcasting using deep generative models of radar
Suman Ravuri, Karel Lenc, Matthew Willson et al. · Nature · 2021 · 850 citations · Full text