Journal cover Journal topic
Geoscientific Model Development An interactive open-access journal of the European Geosciences Union
Journal topic

Journal metrics

Journal metrics

  • IF value: 5.154 IF 5.154
  • IF 5-year value: 5.697 IF 5-year
    5.697
  • CiteScore value: 5.56 CiteScore
    5.56
  • SNIP value: 1.761 SNIP 1.761
  • IPP value: 5.30 IPP 5.30
  • SJR value: 3.164 SJR 3.164
  • Scimago H <br class='hide-on-tablet hide-on-mobile'>index value: 59 Scimago H
    index 59
  • h5-index value: 49 h5-index 49
Discussion papers
https://doi.org/10.5194/gmd-2020-30
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.
https://doi.org/10.5194/gmd-2020-30
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Submitted as: model description paper 04 Mar 2020

Submitted as: model description paper | 04 Mar 2020

Review status
This preprint is currently under review for the journal GMD.

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

Georgy Ayzel1, Tobias Scheffer2, and Maik Heistermann1 Georgy Ayzel et al.
  • 1Institute for Environmental Sciences and Geography, University of Potsdam, Potsdam, Germany
  • 2Department of Computer Science, University of Potsdam, Potsdam, Germany

Abstract. 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 five minutes, 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 900 by 900 km, and has a resolution of 1 km in space and 5 minutes in time. Independent verification experiments were carried out on eleven summer precipitation events from 2016 to 2017. In order to achieve a lead time of one hour, a recursive approach was implemented by using RainNet predictions at five minutes lead time as model input 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 minutes for the routine verification metrics Mean Absolute Error (MAE) and Critical Success Index (CSI, at intensity thresholds of 0.125, 1, and 5 mm/h). Apart from its superiority in terms of MAE and CSI, an undesirable property of RainNet predictions is, however, the level of spatial smoothing. At a lead time of five minutes, 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 minutes 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 five minutes, 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 unfavourable 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 to such future studies.

Georgy Ayzel et al.

Interactive discussion

Status: open (until 29 Apr 2020)
Status: open (until 29 Apr 2020)
AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment
[Subscribe to comment alert] Printer-friendly Version - Printer-friendly version Supplement - Supplement

Georgy Ayzel et al.

Data sets

RYDL: the sample data of the RY product for deep learning applications G. Ayzel https://doi.org/10.5281/zenodo.3629951

RainNet: pretrained model and weights G. Ayzel https://doi.org/10.5281/zenodo.3630429

Model code and software

hydrogo/rainnet: RainNet v1.0-gmdd G. Ayzel https://doi.org/10.5281/zenodo.3631038

Georgy Ayzel et al.

Viewed

Total article views: 259 (including HTML, PDF, and XML)
HTML PDF XML Total Supplement BibTeX EndNote
188 68 3 259 14 1 2
  • HTML: 188
  • PDF: 68
  • XML: 3
  • Total: 259
  • Supplement: 14
  • BibTeX: 1
  • EndNote: 2
Views and downloads (calculated since 04 Mar 2020)
Cumulative views and downloads (calculated since 04 Mar 2020)

Viewed (geographical distribution)

Total article views: 226 (including HTML, PDF, and XML) Thereof 223 with geography defined and 3 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 

Cited

Saved

No saved metrics found.

Discussed

No discussed metrics found.
Latest update: 28 Mar 2020
Publications Copernicus
Download
Short summary
In this study, we present RainNet, a deep convolutional neural network for radar-based precipitation nowcasting, which was trained to predict continuous precipitation intensities at a lead time of five minutes. RainNet significantly outperformed the benchmark models at all lead times up to 60 minutes. Yet, an undesirable property of RainNet predictions is the level of spatial smoothing. Obviously, RainNet learned an optimal level of smoothing to produce a nowcast at 5 minutes lead time.
In this study, we present RainNet, a deep convolutional neural network for radar-based...
Citation