RainNet v1.0: a convolutional neural network for radar-based precipitation nowcasting

In this study, we present RainNet, a deep convolutional neural network for radar-based precipitation nowcasting. Its design was inspired by the U-Net and SegNet families of deep learning models, which were originally designed for binary segmentation tasks. RainNet was trained to predict continuous precipitation intensities at a lead time of 5 min, using several years of quality-controlled weather radar composites provided by the German Weather Service (DWD). That data set covers Germany with a spatial domain of 900km× 900 km and has a resolution of 1 km in space and 5 min in time. Independent verification experiments were carried out on 11 summer precipitation events from 2016 to 2017. In order to achieve a lead time of 1 h, a recursive approach was implemented by using RainNet predictions at 5 min lead times as model inputs for longer lead times. In the verification experiments, trivial Eulerian persistence and a conventional model based on optical flow served as benchmarks. The latter is available in the rainymotion library and had previously been shown to outperform DWD’s operational nowcasting model for the same set of verification events. RainNet significantly outperforms the benchmark models at all lead times up to 60 min for the routine verification metrics mean absolute error (MAE) and the critical success index (CSI) at intensity thresholds of 0.125, 1, and 5 mmh−1. However, rainymotion turned out to be superior in predicting the exceedance of higher intensity thresholds (here 10 and 15 mmh−1). The limited ability of RainNet to predict heavy rainfall intensities is an undesirable property which we attribute to a high level of spatial smoothing introduced by the model. At a lead time of 5 min, an analysis of power spectral density confirmed a significant loss of spectral power at length scales of 16 km and below. Obviously, RainNet had learned an optimal level of smoothing to produce a nowcast at 5 min lead time. In that sense, the loss of spectral power at small scales is informative, too, as it reflects the limits of predictability as a function of spatial scale. Beyond the lead time of 5 min, however, the increasing level of smoothing is a mere artifact – an analogue to numerical diffusion – that is not a property of RainNet itself but of its recursive application. In the context of early warning, the smoothing is particularly unfavorable since pronounced features of intense precipitation tend to get lost over longer lead times. Hence, we propose several options to address this issue in prospective research, including an adjustment of the loss function for model training, model training for longer lead times, and the prediction of threshold exceedance in terms of a binary segmentation task. Furthermore, we suggest additional input data that could help to better identify situations with imminent precipitation dynamics. The model code, pretrained weights, and training data are provided in open repositories as an input for such future studies.

[1]  G. Ayzel.  hydrogo/rainnet: RainNet v1.0-gmdd , 2020 .

[2]  G. Ayzel.  RYDL: the sample data of the RY product for deep learning applications , 2020 .

[3]  G. Ayzel.  RainNet: pretrained model and weights , 2020 .

[4]  Jason Hickey,et al.  Machine Learning for Precipitation Nowcasting from Radar Images , 2019, ArXiv.

[5]  Joachim Denzler,et al.  Deep learning and process understanding for data-driven Earth system science , 2019, Nature.

[6]  Pengfei Chen,et al.  Log Hyperbolic Cosine Loss Improves Variational Auto-Encoder , 2018 .

[7]  G. Ayzel,et al.  Optical flow models as an open benchmark for radar-based precipitation nowcasting (rainymotion v0.1) , 2018, Geoscientific Model Development.

[8]  Peter Bauer,et al.  Challenges and design choices for global weather and climate models based on machine learning , 2018, Geoscientific Model Development.

[9]  Pierre Gentine,et al.  Could Machine Learning Break the Convection Parameterization Deadlock? , 2018, Geophysical Research Letters.

[10]  Vladlen Koltun,et al.  An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling , 2018, ArXiv.

[11]  Qian Li,et al.  A Method of Weather Radar Echo Extrapolation Based on Convolutional Neural Networks , 2018, MMM.

[12]  Alexey Shvets,et al.  TernausNet: U-Net with VGG11 Encoder Pre-Trained on ImageNet for Image Segmentation , 2018, Computer-Aided Analysis of Gastrointestinal Videos.

[13]  Ting Liu,et al.  Recent advances in convolutional neural networks , 2015, Pattern Recognit..

[14]  Pabitra Mitra,et al.  Leveraging Convolutions in Recurrent Neural Networks for Doppler Weather Radar Echo Prediction , 2017, ISNN.

[15]  Yann Dauphin,et al.  Convolutional Sequence to Sequence Learning , 2017, ICML.

[16]  Roberto Cipolla,et al.  SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[17]  Farhad Nili,et al.  Theoretical Analysis , 2017, Encyclopedia of GIS.

[18]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[19]  Peter Bauer,et al.  The quiet revolution of numerical weather prediction , 2015, Nature.

[20]  Jürgen Schmidhuber,et al.  Training Very Deep Networks , 2015, NIPS.

[21]  Thomas Brox,et al.  U-Net: Convolutional Networks for Biomedical Image Segmentation , 2015, MICCAI.

[22]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[23]  Trevor Darrell,et al.  Fully Convolutional Networks for Semantic Segmentation , 2017, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[24]  Juanzhen Sun,et al.  Use of NWP for Nowcasting Convective Precipitation: Recent Progress and Challenges , 2014 .

[25]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[26]  Tara N. Sainath,et al.  Improving deep neural networks for LVCSR using rectified linear units and dropout , 2013, 2013 IEEE International Conference on Acoustics, Speech and Signal Processing.

[27]  William G. Griswold,et al.  Interpersonal informatics: making social influence visible , 2011, CHI Extended Abstracts.

[28]  Jean Ponce,et al.  A Theoretical Analysis of Feature Pooling in Visual Recognition , 2010, ICML.

[29]  Nigel Roberts,et al.  Intercomparison of Spatial Forecast Verification Methods: Identifying Skillful Spatial Scales Using the Fractions Skill Score , 2010 .

[30]  I. Zawadzki,et al.  Precipitation forecast skill of numerical weather prediction models and radar nowcasts , 2005 .

[31]  I. Zawadzki,et al.  Scale-Dependence of the Predictability of Precipitation from Continental Radar Images. Part I: Description of the Methodology , 2002 .

[32]  Juanzhen Sun,et al.  Nowcasting Thunderstorms: A Status Report , 1998 .

[33]  A. Bellon,et al.  The use of digital weather radar records for short‐term precipitation forecasting , 1974 .

[34]  P. Welch.  The use of fast Fourier transform for the estimation of power spectra: A method based on time averaging over short, modified periodograms , 1967 .