Rain Nowcasting from Multiscale Radar Images

Zébiri, Aniss; Béréziat, Dominique; Huot, Etienne; Herlin, Isabelle

doi:10.5220/0007566908920900

Cited by 2 publications

(4 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We compared our model to the persistence model which consists in taking the last rain radar image of an input sequence as the prediction (though simplistic, this model is frequently used in rain nowcasting [12,13,17] and can prove difficult to outperform) and to a similar neural network trained using only radar rain images as inputs. We also compare our approach with an operational and optical flow-based rain nowcasting system [22]. Our models present satisfactory forecasting skill at horizon times of thirty minutes and one hour and outperform both comparison models indicating that data fusion has a significant positive impact.…”

Section: Introductionmentioning

confidence: 88%

See 1 more Smart Citation

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

Bouget¹,

Béréziat²,

Brajard³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

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%.

show abstract

Section: Introductionmentioning

confidence: 88%

“…This approach is known to be limited to small displacements. A solution to set up this issue is to use a data assimilation approach as described in [22]. Once the estimation of velocity field W = ( U, V) computed, the last observation I last is transported, Eq.…”

Section: Rain Nowcasting With Optical Flowmentioning

confidence: 99%

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

Bouget¹,

Béréziat²,

Brajard³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…This approach is known to be limited to small displacements. A solution to fix this issue is to use a data assimilation approach as described in [20]. Once the estimation of velocity field W = ( U, V) is computed, the last observation I last is transported, Equation 15, at the wished temporal horizon.…”

Section: Baselinementioning

confidence: 99%

“…We present a method of data over-sampling to address this issue. We compared our model to the persistence model which consists of taking the last rain radar image of an input sequence as the prediction (though simplistic, this model is frequently used in rain nowcasting [11,12,16]) and to an operational and optical flow-based rain nowcasting system [20]. We also compare the neural network merging radar image and wind forecast to a similar neural network trained using only radar rain images as inputs.…”

Section: Introductionmentioning

confidence: 99%

Fusion of Rain Radar Images and Wind Forecasts in a Deep Learning Model Applied to Rain Nowcasting

Bouget

Béréziat

Brajard³

et al. 2021

Remote Sensing

Self Cite

View full text Add to dashboard Cite

Short- or mid-term rainfall forecasting is a major task with several environmental applications such as agricultural management or flood risk monitoring. Existing data-driven approaches, especially deep learning models, have shown significant skill at this task, using only rainfall 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 rainfall radar images and wind velocity produced by a weather forecast model. The network was compared to a similar architecture trained only on radar data, to a basic persistence model and to an approach based on optical flow. Our network outperforms by 8% the F1-score calculated for the optical flow on moderate and higher rain events for forecasts at a horizon time of 30 min. Furthermore, it outperforms by 7% the same architecture trained using only rainfall radar images. Merging rain and wind data has also proven to stabilize the training process and enabled significant improvement especially on the difficult-to-predict high precipitation rainfalls.

show abstract

Rain Nowcasting from Multiscale Radar Images

Cited by 2 publications

References 0 publications

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

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

Fusion of Rain Radar Images and Wind Forecasts in a Deep Learning Model Applied to Rain Nowcasting

Contact Info

Product

Resources

About