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

Abstract

Short or mid-term rainfall forecasting is a major task for several environmental applications, such as agricultural management or monitoring flood risks. Existing data-driven approaches, especially deep learning models, have shown significant skill at this task, using only rain radar images as inputs. In order to determine whether using other meteorological parameters such as wind would improve forecasts, we trained a deep learning model on a fusion of rain radar images and wind velocity produced by a weather forecast model. The network was compared to a similar architecture trained only on rainfall data, to a basic persistence model and to an approach based on optical flow. Our network outperforms the F1-score calculated for the optical flow on moderate and higher rain events for forecasts at a horizon time of 30 minutes by 8%. Furthermore, it outperforms the same architecture trained using only rainfalls by 7%.

Coupling a Neural Network with a Spatial Downscaling Procedure to Improve Probabilistic Nowcast for Urban Rain Radars

“…In [16], the same U-Net architecture is enriched with attention modules obtaining good results and limiting the size of the model needed. In [19], rain radar images and wind forecasts are used in a deep learning model to transform the rainfall intensity regression problem into a classification problem in which the probability of exceeding a certain precipitation threshold is estimated. In [17], a deep architecture with a self attention mechanism is proposed for the realization of a probabilistic forecast, integrating ground-based radar information with satellite images and weather model forecasts obtaining a forecast over the entire territory of the US.…”