Deep Learning for Rain and Lightning Nowcasting @NIPS2016

We describe a deep learning framework for precipitation and lightning nowcasting, applied to weather echo radar and lightning data at regional scale in Trentino-Sudtirol, in the Italian Alps. Nowcasting, i.e. forecasts obtained by extrapolation for a period of 0 up to 6 hours ahead, is based on a Convolutional Long-Short Term Memory model (ConvLSTM) (Shi et al., 2015) and it is embedded in an operational context. The model is able to forecast reflectivity up to 75′ ahead at a spatial resolution of 1.65 km based on 5 frames of recent (25′) radar data. Further, the model can manage blocking effects due to orography. The framework has been also applied to 95.7K events collected from a collaborative lightning location network on the same region, with a predictive accuracy of CSI=0.585, comparable with state of the art machine learning based solutions.

Transcript

1. Deep Learning 
 for Rain and Lightning Nowcasting G. Franch, A.I. Nardelli, C. Zarbo, 
 V. Maggio, G. Jurman, C. Furlanello Fondazione Bruno Kessler (FBK)
 Trento, Italy
3. Motivations •Lightning forecasting can help reduce the risks of injuries and damage to electronic apparatus or even people •Precipitation forecasting has applications in agriculture, boating, hiking, outdoor sports, event planning, assistance to aircraft flight •Nowcasting (0-6 hours forecasting) is a key resource for civil protection purposes
4. We describe a deep learning framework for precipitation and lightning nowcasting applied to weather echo radar and lightning data at regional scale in Trentino, Italy Deep Learning 
 for Rain and Lightning Nowcasting
5. We describe a deep learning framework for precipitation and lightning nowcasting applied to weather echo radar and lightning data at regional scale in Trentino, Italy Deep Learning 
 for Rain and Lightning Nowcasting
6. ConvLSTM [1] X. Shi et al. Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting. (NIPS 2015). Deep Learning 
 for Rain and Lightning Nowcasting Short-time weather data has strong spatio-temporal correlation, 
 continuous domain, big dataset: perfect target for deep learning
7. Deep Learning 
 for Rain and Lightning Nowcasting Metric Rain (75') Lightning (25') CSI 0.585 0.351 FAR 0.180 0.271 POD 0.651 0.403 COR 0.910 0.585
8. [6] J.J. Song. Radar Nowcasting of Cloud-to- Ground (CG) Lightning: A Machine Learning Approach. (2016) Metric Rain (75') Lightning (25') CSI 0.585 0.351 FAR 0.180 0.271 POD 0.651 0.403 COR 0.910 0.585
9. Deep Learning for Rain and Lightning Nowcasting G. Franch, A.I. Nardelli, C. Zarbo, 
 V. Maggio, G. Jurman, C. Furlanello 
 Fondazione Bruno Kessler (FBK), Trento Italy Thank you!