Fusion of rain radar images and wind forecasts in a deep learning model applied to rain nowcasting

  • 2021-01-12 16:11:11
  • Vincent Bouget, Dominique Béréziat, Julien Brajard, Anastase Charantonis, Arthur Filoche
  • 0

Abstract

Short- or mid-term rainfall forecasting is a major task with severalenvironmental applications such as agricultural management or flood riskmonitoring. Existing data-driven approaches, especially deep learning models,have shown significant skill at this task, using only rainfall radar images asinputs. In order to determine whether using other meteorological parameterssuch as wind would improve forecasts, we trained a deep learning model on afusion of rainfall radar images and wind velocity produced by a weatherforecast model. The network was compared to a similar architecture trained onlyon radar data, to a basic persistence model and to an approach based on opticalflow. Our network outperforms by 8% the F1-score calculated for the opticalflow on moderate and higher rain events for forecasts at a horizon time of 30min. Furthermore, it outperforms by 7% the same architecture trained using onlyrainfall radar images. Merging rain and wind data has also proven to stabilizethe training process and enabled significant improvement especially on thedifficult-to-predict high precipitation rainfalls.

 

Quick Read (beta)

loading the full paper ...