Journal of Korea Water Resources Association (한국수자원학회논문집)

Korea Water Resources Association (한국수자원학회)
권호 표지
DOI QR코드

DOI QR Code

Radar rainfall prediction based on deep learning considering temporal consistency

시간 연속성을 고려한 딥러닝 기반 레이더 강우예측

  • Shin, Hongjoon (Korea Hydro&Nuclear Power Co. Ltd) ;
  • Yoon, Seongsim (Korea Institute of Civil Engineering and Building Technology) ;
  • Choi, Jaemin (Department of Building Equipment System & Fire Protection Engineering, Gachon University)
  • 신홍준 (한국수력원자력 수력처) ;
  • 윤성심 (한국건설기술연구원 국토보전연구본부) ;
  • 최재민 (가천대학교 설비소방공학과)
  • Received : 2021.02.17
  • Accepted : 2021.03.23
  • Published : 2021.05.31
Download PDF
⟨ Previous Next ⟩

Abstract

In this study, we tried to improve the performance of the existing U-net-based deep learning rainfall prediction model, which can weaken the meaning of time series order. For this, ConvLSTM2D U-Net structure model considering temporal consistency of data was applied, and we evaluated accuracy of the ConvLSTM2D U-Net model using a RainNet model and an extrapolation-based advection model. In addition, we tried to improve the uncertainty in the model training process by performing learning not only with a single model but also with 10 ensemble models. The trained neural network rainfall prediction model was optimized to generate 10-minute advance prediction data using four consecutive data of the past 30 minutes from the present. The results of deep learning rainfall prediction models are difficult to identify schematically distinct differences, but with ConvLSTM2D U-Net, the magnitude of the prediction error is the smallest and the location of rainfall is relatively accurate. In particular, the ensemble ConvLSTM2D U-Net showed high CSI, low MAE, and a narrow error range, and predicted rainfall more accurately and stable prediction performance than other models. However, the prediction performance for a specific point was very low compared to the prediction performance for the entire area, and the deep learning rainfall prediction model also had limitations. Through this study, it was confirmed that the ConvLSTM2D U-Net neural network structure to account for the change of time could increase the prediction accuracy, but there is still a limitation of the convolution deep neural network model due to spatial smoothing in the strong rainfall region or detailed rainfall prediction.

본 연구에서는 시계열 순서의 의미가 희석될 수 있는 기존의 U-net 기반 딥러닝 강우예측 모델의 성능을 개선하고자 하였다. 이를 위해서 데이터의 연속성을 고려한 ConvLSTM2D U-Net 신경망 구조를 갖는 모델을 적용하고, RainNet 모델 및 외삽 기반의 이류모델을 이용하여 예측정확도 개선 정도를 평가하였다. 또한 신경망 기반 모델 학습과정에서의 불확실성을 개선하기 위해 단일 모델뿐만 아니라 10개의 앙상블 모델로 학습을 수행하였다. 학습된 신경망 강우예측모델은 현재를 기준으로 과거 30분 전까지의 연속된 4개의 자료를 이용하여 10분 선행 예측자료를 생성하는데 최적화되었다. 최적화된 딥러닝 강우예측모델을 이용하여 강우예측을 수행한 결과, ConvLSTM2D U-Net을 사용하였을 때 예측 오차의 크기가 가장 작고, 강우 이동 위치를 상대적으로 정확히 구현하였다. 특히, 앙상블 ConvLSTM2D U-Net이 타 예측모델에 비해 높은 CSI와 낮은 MAE를 보이며, 상대적으로 정확하게 강우를 예측하였으며, 좁은 오차범위로 안정적인 예측성능을 보여주었다. 다만, 특정 지점만을 대상으로 한 예측성능은 전체 강우 영역에 대한 예측성능에 비해 낮게 나타나, 상세한 영역의 강우예측에 대한 딥러닝 강우예측모델의 한계도 확인하였다. 본 연구를 통해 시간의 변화를 고려하기 위한 ConvLSTM2D U-Net 신경망 구조가 예측정확도를 높일 수 있었으나, 여전히 강한 강우영역이나 상세한 강우예측에는 공간 평활로 인한 합성곱 신경망 모델의 한계가 있음을 확인하였다.

Keywords

Acknowledgement

본 논문은 한국수력원자력(주)에서 재원을 부담하여 한국건설기술연구원에서 수행한 연구결과입니다(No. 2018-기술-20).

References

  1. Agrawal, S., Barrington, L., Bromberg, C., Burge, J., Gazen, C., and Hickey, J. (2019). Machine learning for precipitation nowcasting from radar images. accessed 28 January 2020, .
  2. Ayzel, G., Scheffer, T., and Heistermann, M. (2020). "RainNet v1.0: a convolutional neural network for radar-based precipitation nowcasting." Geoscientific Model Development, Vol. 13, pp. 2631-2644. https://doi.org/10.5194/gmd-13-2631-2020
  3. Kingma, D.P., and Ba, J. (2015). A method for stochastic optimization. 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, 79 May 2015, accessed 10 June 2020 .
  4. Nakakita, E., Ikebuchi, S., Nakamura, T., Kanmuri, M., Okuda, M., Yamaji, A., and Takasao T. (1996). "Short-term rainfall prediction method using a volume scanning radar and GPV data from numerical weather prediction." Journal of Geophysical Research, Vol. 101, No. D21, pp. 26181-26197. https://doi.org/10.1029/96JD01615
  5. Reichstein, M., Camps-Valls, G., Stevens, B., Jung, M., Denzler, J., Carvalhais, N., and Prabhat. (2019). "Deep learning and process understanding for data-driven Earth system science." Nature, 566, pp. 195-204. doi: 10.1038/s41586-019-0912-1
  6. Shi, X., Chen, Z., Wang, H., Yeung, D., Wong, W., and Woo, W. (2015). Convolutional LSTM network: A machine learning approach for precipitation nowcasting, accessed 6 May 2021, .
  7. Shi, X., Gao, Z., Lausen, L., Wang, H., Yeung D., Wong, W., and Woo, W. (2017) "Deep learning for precipitation nowcasting: A benchmark and a new model." 31st Conference on Neural Information Processing Systems, NIPS, Long Beach, CA, U.S.
  8. Shiiba, M., Takasao, T., and Nakakita, E. (1984). "Investigation of short-term rainfall prediction method by a translation model." Proceeding Japanese Conference on Hydraulics, JSCE, Vol. 28, pp. 423-428.
  9. Shin, H.J., and Yoon, S.S. (2021). "AI Competition for rain prediction of Hydropower dam using public data." Water for Future, Vol. 54, No. 1, pp. 87-92.
  10. Sugimoto, S., Nakakita E., and Ikebuchi, S. (2001). "A stochastic approach to short-term rainfall prediction using a physically based conceptual rainfall model." Journal of Hydrology, Vol. 242, pp. 137-155. https://doi.org/10.1016/S0022-1694(00)00390-5
  11. Tran, Q.K., and Song, S.K. (2019). "Computer vision in precipitation nowcasting: Applying image quality assessment metrics for training deep neural networks." Atmosphere, Vol. 10, No. 5, doi: 10.3390/atmos10050244
  12. Yoon, S.S., and Bae, D.H. (2010). "The applicability assessment of the short-term rainfall forecasting using translation model." Journal of Korea Water Resources Association, Vol. 43, No. 8, pp. 695-707. https://doi.org/10.3741/JKWRA.2010.43.8.695
  13. Yoon, S.S., Park, H.S., and Shin, H.J. (2020). "Very short-term rainfall prediction based on radar image learning using deep neural network." Journal of Korea Water Resources Association, Vol. 53, No. 12, pp. 1159-1172. https://doi.org/10.3741/JKWRA.2020.53.12.1159