Short‐Term Precipitation Prediction for Contiguous United States Using Deep Learning
Accurate short‐term weather prediction, essential for many aspects of life, relies mainly on forecasts from numerical weather models. Here, we report results supporting strongly deep learning as a viable, alternative approach. A 3D convolutional neural network, which uses a single frame of meteorolo...
Uložené v:
| Vydané v: | Geophysical research letters Ročník 49; číslo 8 |
|---|---|
| Hlavní autori: | , |
| Médium: | Journal Article |
| Jazyk: | English |
| Vydavateľské údaje: |
Washington
John Wiley & Sons, Inc
28.04.2022
|
| Predmet: | |
| ISSN: | 0094-8276, 1944-8007 |
| On-line prístup: | Získať plný text |
| Tagy: |
Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
|
| Shrnutí: | Accurate short‐term weather prediction, essential for many aspects of life, relies mainly on forecasts from numerical weather models. Here, we report results supporting strongly deep learning as a viable, alternative approach. A 3D convolutional neural network, which uses a single frame of meteorology fields as input to predict the precipitation spatial distribution, is developed based on 39‐years (1980–2018) data of meteorology and daily precipitation over the contiguous United States. Results show that the trained network outperforms the state‐of‐the‐art weather models in predicting daily total precipitation, and the superiority of the network extends to forecast leads up to 5 days. Combining the network predictions with the weather‐model forecasts significantly improves the accuracy of model forecasts, especially for heavy‐precipitation events. Furthermore, the millisecond‐scale inference time of the network facilitates large ensemble predictions for extra accuracy improvement. These results demonstrate the promising prospects of deep learning in short‐term weather predictions.
Plain Language Summary
Accurate weather prediction is essential for many aspects of life, notably the early warning of extreme weather events such as rainstorms. Short‐term predictions of these events rely on forecasts from first‐principle numerical weather models, in which, despite much improvement in the past few decades, outstanding issues remain concerning model uncertainties and fast‐growing demands for computation and storage resources. The recent advance of deep learning offers a viable alternative approach. Here we develop a 3D convolutional neural network that uses a single frame of meteorology to predict daily precipitation spatial distribution and demonstrate its superiorities in three aspects. First, the network better predicts daily total precipitation over the contiguous United States than the state‐of‐the‐art weather models for forecast leads up to 5 days. Second, the weighting average of the network predictions and the weather‐model forecasts is more accurate than the model forecasts alone, suggesting a potential approach for improving model forecasts; third, the network prediction via inference is so fast that large ensemble predictions can be conducted to further improve the prediction accuracy. Thus, it is concluded that more research on deep learning for short‐term weather predictions should be actively pursued.
Key Points
A neural network outperforms numerical models in predicting daily total precipitation for contiguous US with forecast leads up to 5 days
Combining the network prediction with the weather‐model forecast can significantly improve the model forecast accuracy
The millisecond‐scale inference time of the network facilitates large ensemble predictions for extra accuracy improvement |
|---|---|
| Bibliografia: | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14 |
| ISSN: | 0094-8276 1944-8007 |
| DOI: | 10.1029/2022GL097904 |