RN-Net: A Deep Learning Approach to 0–2 Hour Rainfall Nowcasting Based on Radar and Automatic Weather Station Data
Abstract
:1. Introduction
2. Preliminary
2.1. Data Details
2.2. Problem Definition
3. Method
3.1. Network Structure
3.2. Implementation Details
4. Experiment
4.1. Dataset
4.2. Performance Metric
4.3. Experimental Results and Analysis
4.4. Visualization Results
5. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Knievel, J.C.; Ahijevych, D.A.; Manning, K.W. Using temporal modes of rainfall to evaluate the performance of a numerical weather prediction model. Mon. Weather Rev. 2004, 132, 2995–3009. [Google Scholar] [CrossRef] [Green Version]
- Muhandhis, I.; Susanto, H.; Asfari, U. Determining Salt Production Season Based on Rainfall Forecasting Using Weighted Fuzzy Time Series. J. Appl. Comput. Sci. Math. 2020, 14, 23–27. [Google Scholar] [CrossRef]
- Zhou, J.; Zhang, Y.; Tian, S.; Lai, S. Forecasting Rainfall with Recurrent Neural Network for irrigation equipment. IOP Conf. Ser. Earth Environ. Sci. 2020, 510, 042040. [Google Scholar] [CrossRef]
- Osanai, N.; Shimizu, T.; Kuramoto, K.; Kojima, S.; Noro, T. Japanese early-warning for debris flows and slope failures using rainfall indices with Radial Basis Function Network. Landslides 2010, 7, 325–338. [Google Scholar] [CrossRef]
- Chu, Q.; Xu, Z.; Chen, Y.; Han, D. Evaluation of the ability of the Weather Research and Forecasting model to reproduce a sub-daily extreme rainfall event in Beijing, China using different domain configurations and spin-up times. Hydrol. Earth Syst. Sci. 2018, 22, 3391. [Google Scholar] [CrossRef] [Green Version]
- Hong, W.C. Rainfall forecasting by technological machine learning models. Appl. Math. Comput. 2008, 200, 41–57. [Google Scholar] [CrossRef]
- Sumi, S.M.; Zaman, M.F.; Hirose, H. A rainfall forecasting method using machine learning models and its application to the Fukuoka city case. Int. J. Appl. Math. Comput. Sci. 2012, 22, 841–854. [Google Scholar] [CrossRef] [Green Version]
- Cramer, S.; Kampouridis, M.; Freitas, A.A.; Alexandridis, A.K. An extensive evaluation of seven machine learning methods for rainfall prediction in weather derivatives. Expert Syst. Appl. 2017, 85, 169–181. [Google Scholar] [CrossRef] [Green Version]
- Moon, S.H.; Kim, Y.H.; Lee, Y.H.; Moon, B.R. Application of machine learning to an early warning system for very short-term heavy rainfall. J. Hydrol. 2019, 568, 1042–1054. [Google Scholar] [CrossRef]
- Parmar, A.; Mistree, K.; Sompura, M. Machine Learning Techniques for Rainfall Prediction: A Review. In Proceedings of the 2017 International Conference on Innovations in information Embedded and Communication Systems (ICIIECS), Coimbatore, India, 17–18 March 2017. [Google Scholar]
- Adewoyin, R.; Dueben, P.; Watson, P.; He, Y.; Dutta, R. TRU-NET: A Deep Learning Approach to High Resolution Prediction of Rainfall. arXiv 2020, arXiv:2008.09090. [Google Scholar]
- Ayzel, G.; Heistermann, M.; Sorokin, A.; Nikitin, O.; Lukyanova, O. All convolutional neural networks for radar-based precipitation nowcasting. Procedia Comput. Sci. 2019, 150, 186–192. [Google Scholar] [CrossRef]
- Tian, L.; Li, X.; Ye, Y.; Xie, P.; Li, Y. A Generative Adversarial Gated Recurrent Unit Model for Precipitation Nowcasting. IEEE Geosci. Remote Sens. Lett. 2019, 17, 601–605. [Google Scholar] [CrossRef]
- Shi, X.; Chen, Z.; Wang, H.; Yeung, D.Y.; Wong, W.K.; Woo, W.C. Convolutional LSTM network: A machine learning approach for precipitation nowcasting. Adv. Neural Inf. Process. Syst. 2015, 28, 802–810. [Google Scholar]
- Shi, X.; Gao, Z.; Lausen, L.; Wang, H.; Yeung, D.Y.; Wong, W.k.; Woo, W.C. Deep learning for precipitation nowcasting: A benchmark and a new model. arXiv 2017, arXiv:1706.03458. [Google Scholar]
- Woo, W.C.; Wong, W.K. Operational application of optical flow techniques to radar-based rainfall nowcasting. Atmosphere 2017, 8, 48. [Google Scholar] [CrossRef] [Green Version]
- Wang, Y.; Long, M.; Wang, J.; Gao, Z.; Philip, S.Y. Predrnn: Recurrent Neural Networks for Predictive Learning Using Spatiotemporal Lstms. 2017. Available online: http://ise.thss.tsinghua.edu.cn/~mlong/doc/predrnn-nips17.pdf (accessed on 10 March 2021).
- Wang, Y.; Zhang, J.; Zhu, H.; Long, M.; Wang, J.; Yu, P.S. Memory in memory: A predictive neural network for learning higher-order non-stationarity from spatiotemporal dynamics. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Long Beach, CA, USA, 15–20 June 2019; pp. 9154–9162. [Google Scholar]
- Bonnet, S.M.; Evsukoff, A.; Morales Rodriguez, C.A. Precipitation Nowcasting with Weather Radar Images and Deep Learning in São Paulo, Brasil. Atmosphere 2020, 11, 1157. [Google Scholar] [CrossRef]
- Wang, Y.; Gao, Z.; Long, M.; Wang, J.; Philip, S.Y. PredRNN++: Towards A Resolution of the Deep-in-Time Dilemma in Spatiotemporal Predictive Learning. In Proceedings of the International Conference on Machine Learning, Long Beach, CA, USA, 16–20 June 2019; pp. 5123–5132. [Google Scholar]
- Zhang, P.; Jia, Y.; Gao, J.; Song, W.; Leung, H.K. Short-term rainfall forecasting using multi-layer perceptron. IEEE Trans. Big Data 2018, 6, 93–106. [Google Scholar] [CrossRef]
- Oprea, S.; Martinez-Gonzalez, P.; Garcia-Garcia, A.; Castro-Vargas, J.A.; Orts-Escolano, S.; Garcia-Rodriguez, J.; Argyros, A. A Review on Deep Learning Techniques for Video Prediction. arXiv 2020, arXiv:2004.05214. [Google Scholar]
- Skamarock, W.C.; Klemp, J.B.; Dudhia, J.; Gill, D.O.; Barker, D.M.; Duda, M.G.; Huang, X.Y.; Wang, W.; Powers, J.G. A Description of the Advanced Research WRF Version 3. Available online: https://apps.dtic.mil/sti/pdfs/ADA487419.pdf (accessed on 11 March 2021).
- Cornes, R.C.; van der Schrier, G.; van den Besselaar, E.J.; Jones, P.D. An ensemble version of the E-OBS temperature and precipitation data sets. J. Geophys. Res. Atmos. 2018, 123, 9391–9409. [Google Scholar] [CrossRef] [Green Version]
- Chen, F.W.; Liu, C.W. Estimation of the spatial rainfall distribution using inverse distance weighting (IDW) in the middle of Taiwan. Paddy Water Environ. 2012, 10, 209–222. [Google Scholar] [CrossRef]
- Skamarock, W.C.; Klemp, J.B.; Dudhia, J. Prototypes for the WRF (Weather Research and Forecasting) Model. Available online: https://www.researchgate.net/profile/Jimy_Dudhia/publication/242446432_Prototypes_for_the_WRF_Weather_Research_and_Forecasting_model/links/0c9605314d71c38bae000000/Prototypes-for-the-WRF-Weather-Research-and-Forecasting-model.pdf (accessed on 10 March 2021).
- Geng, Y.a.; Li, Q.; Lin, T.; Jiang, L.; Xu, L.; Zheng, D.; Yao, W.; Lyu, W.; Zhang, Y. Lightnet: A dual spatiotemporal encoder network model for lightning prediction. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, Anchorage, AK, USA, 4–8 August 2019; pp. 2439–2447. [Google Scholar]
- Paszke, A.; Gross, S.; Massa, F.; Lerer, A.; Bradbury, J.; Chanan, G.; Killeen, T.; Lin, Z.; Gimelshein, N.; Antiga, L.; et al. Pytorch: An imperative style, high-performance deep learning library. arXiv 2019, arXiv:1912.01703. [Google Scholar]
- Kingma, D.P.; Ba, J. Adam: A method for stochastic optimization. arXiv 2014, arXiv:1412.6980. [Google Scholar]
- Le, P.; Zuidema, W. Quantifying the vanishing gradient and long distance dependency problem in recursive neural networks and recursive LSTMs. arXiv 2016, arXiv:1603.00423. [Google Scholar]
- Tran, Q.K.; Song, S.k. Computer vision in precipitation nowcasting: Applying image quality assessment metrics for training deep neural networks. Atmosphere 2019, 10, 244. [Google Scholar] [CrossRef] [Green Version]
Name | Scheme |
---|---|
Microphysics | Thompson scheme |
Cumulus parameterization | Kain–Fritsch (new Eta) scheme |
Planetary boundary layer | Mellor–Yamada–Janjic TKE scheme |
Surface layer | Revised MM5 Monin–Obukhov scheme |
Longwave radiation | Rapid Radiative Transfer Model for GCMs |
Shortwave radiation | Rapid Radiative Transfer Model for GCMs |
Module | Name | CH I/O | Kernel | Stride |
---|---|---|---|---|
Radar Echo Encode or Rainfall Encode | Econv1 ETrajGRU1 Econv2 ETrajGRU2 Econv3 ETrajGRU3 | 1/8 8/64 64/192 192/192 192/192 192/192 | 5 × 5 3 × 3 4 × 4 3 × 3 3 × 3 3 × 3 | 3 1 2 1 3 1 |
Fusion Module | Fconv1 Fconv2 Fconv3 | 128/64 384/192 384/192 | 3 × 3 3 × 3 3 × 3 | 1 1 1 |
Rainfall Predictor | PTrajGRU3 Pdeconv3 PTrajGRU2 Pdeconv2 PTrajGRU1 Pdeconv1 Oconv1 Oconv2 | 192/192 192/192 192/192 192/64 64/64 64/8 8/8 8/1 | 3 × 3 3 × 3 3 × 3 4 × 4 3 × 3 5 × 5 3 × 3 1 × 1 | 1 3 1 2 1 3 1 1 |
Half-Hour of Rainfall (mm) | Rainfall Level |
---|---|
r < 0.25 | No or hardly noticeable |
0.25 ≤ r < 1 | Light |
1 ≤ r < 2.5 | Light to moderate |
2.5 ≤ r | Moderate or greater |
Method | MSE↓ | MAE↓ | r ≥ 0.25 mm TS↑ POD↑ FAR↓ | r ≥ 1 mm TS↑ POD↑ FAR↓ | r ≥ 2.5 mm TS↑ POD↑ FAR↓ |
---|---|---|---|---|---|
WRF | 1.077 | 26.602 | 0.131 0.359 0.763 | 0.094 0.278 0.830 | 0.066 0.190 0.868 |
RF-PredRNN | 0.965 | 19.793 | 0.404 0.495 0.311 | 0.303 0.348 0.298 | 0.216 0.243 0.336 |
RF-ConvLSTM | 0.890 | 18.889 | 0.425 0.506 0.272 | 0.348 0.409 0.299 | 0.268 0.310 0.337 |
RF-TrajGRU | 0.929 | 19.024 | 0.427 0.516 0.287 | 0.326 0.378 0.293 | 0.235 0.265 0.324 |
RE-PredRNN | 1.106 | 22.750 | 0.380 0.380 0.344 | 0.344 0.361 0.389 | 0.189 0.228 0.475 |
RE-ConvLSTM | 0.932 | 20.550 | 0.428 0.501 0.254 | 0.354 0.412 0.285 | 0.261 0.301 0.334 |
RE-TrajGRU | 0.872 | 20.514 | 0.455 0.546 0.268 | 0.400 0.492 0.318 | 0.315 0.392 0.383 |
RE-RF-PredRNN | 0.841 | 18.517 | 0.474 0.585 0.285 | 0.393 0.478 0.310 | 0.296 0.356 0.360 |
RE-RF-ConvLSTM | 0.674 | 16.531 | 0.507 0.577 0.193 | 0.4520.5190.221 | 0.3230.4460.266 |
RN-Net | 0.698 | 16.484 | 0.5230.6110.214 | 0.4640.5510.252 | 0.3710.4330.278 |
Method | MSE↓ | MAE↓ | r ≥ 0.25 mm TS↑ POD↑ FAR↓ | r ≥ 1 mm TS↑ POD↑ FAR↓ | r ≥ 2.5 mm TS↑ POD↑ FAR↓ |
---|---|---|---|---|---|
WRF | 3.635 | 53.311 | 0.153 0.419 0.736 | 0.129 0.363 0.759 | 0.098 0.294 0.818 |
RF-PredRNN | 3.104 | 40.523 | 0.395 0.454 0.245 | 0.327 0.295 0.280 | 0.249 0.278 0.261 |
RF-ConvLSTM | 2.963 | 39.124 | 0.424 0.491 0.244 | 0.335 0.374 0.237 | 0.274 0.304 0.262 |
RF-TrajGRU | 3.066 | 39.811 | 0.428 0.503 0.255 | 0.331 0.380 0.276 | 0.258 0.291 0.302 |
RE-PredRNN | 3.515 | 46.225 | 0.352 0.411 0.287 | 0.325 0.398 0.358 | 0.247 0.297 0.407 |
RE-ConvLSTM | 3.083 | 42.151 | 0.365 0.395 0.170 | 0.362 0.407 0.235 | 0.235 0.320 0.261 |
RE-TrajGRU | 2.818 | 41.777 | 0.423 0.508 0.283 | 0.411 0.486 0.272 | 0.356 0.430 0.324 |
RE-RF-PredRNN | 2.713 | 38.258 | 0.454 0.525 0.227 | 0.411 0.494 0.290 | 0.340 0.407 0.325 |
RE-RF-ConvLSTM | 2.260 | 34.573 | 0.465 0.508 0.154 | 0.442 0.489 0.177 | 0.3920.4390.214 |
RN-Net | 2.358 | 34.959 | 0.5030.585 0.217 | 0.4610.5330.226 | 0.3990.467 0.266 |
Method | MSE↓ | MAE↓ | r ≥ 0.25 mm TS↑ POD↑ FAR↓ | r ≥ 1 mm TS↑ POD↑ FAR↓ | r ≥ 2.5 mm TS↑ POD↑ FAR↓ |
---|---|---|---|---|---|
WRF | 12.467 | 130.027 | 0.168 0.466 0.742 | 0.160 0.422 0.758 | 0.132 0.384 0.780 |
RF-PredRNN | 10.196 | 87.180 | 0.341 0.380 0.230 | 0.295 0.333 0.278 | 0.235 0.261 0.295 |
RF-ConvLSTM | 10.018 | 84.924 | 0.352 0.391 0.217 | 0.277 0.297 0.194 | 0.227 0.241 0.202 |
RF-TrajGRU | 10.262 | 87.172 | 0.360 0.409 0.247 | 0.285 0.319 0.275 | 0.228 0.254 0.310 |
RE-PredRNN | 11.308 | 97.919 | 0.315 0.368 0.313 | 0.297 0.361 0.373 | 0.244 0.293 0.407 |
RE-ConvLSTM | 10.635 | 91.343 | 0.293 0.309 0.148 | 0.289 0.310 0.193 | 0.239 0.256 0.219 |
RE-TrajGRU | 9.655 | 90.043 | 0.378 0.441 0.274 | 0.359 0.413 0.266 | 0.318 0.372 0.313 |
RE-RF-PredRNN | 9.167 | 83.913 | 0.399 0.461 0.250 | 0.3680.440 0.307 | 0.314 0.373 0.337 |
RE-RF-ConvLSTM | 8.282 | 77.884 | 0.393 0.421 0.144 | 0.368 0.394 0.154 | 0.325 0.351 0.183 |
RN-Net | 8.465 | 79.188 | 0.4350.495 0.218 | 0.3960.453 0.242 | 0.3500.406 0.283 |
Method | MSE↓ | MAE↓ | r ≥ 0.25 mm CSI↑ POD↑ FAR↓ | r ≥ 1 mm CSI↑ POD↑ FAR↓ | r ≥ 2.5 mm CSI↑ POD↑ FAR↓ |
---|---|---|---|---|---|
WRF | 1.223 | 27.484 | 0.129 0.355 0.784 | 0.092 0.285 0.846 | 0.065 0.190 0.883 |
RF-PredRNN | 1.091 | 21.201 | 0.349 0.424 0.339 | 0.244 0.280 0.349 | 0.163 0.183 0.405 |
RF-ConvLSTM | 0.890 | 20.424 | 0.347 0.404 0.276 | 0.264 0.303 0.311 | 0.189 0.214 0.365 |
RF-TrajGRU | 1.076 | 20.785 | 0.354 0.425 0.322 | 0.257 0.297 0.353 | 0.176 0.198 0.405 |
RE-PredRNN | 1.217 | 24.064 | 0.337 0.428 0.390 | 0.241 0.298 0.447 | 0.144 0.173 0.540 |
RE-ConvLSTM | 1.086 | 21.913 | 0.371 0.431 0.273 | 0.281 0.322 0.309 | 0.189 0.214 0.371 |
RE-TrajGRU | 1.013 | 22.023 | 0.407 0.490 0.295 | 0.342 0.423 0.361 | 0.257 0.318 0.435 |
RE-RF-PredRNN | 0.986 | 20.351 | 0.416 0.522 0.331 | 0.326 0.401 0.372 | 0.232 0.280 0.433 |
RE-RF-ConvLSTM | 0.840 | 18.321 | 0.446 0.508 0.219 | 0.3770.4330.259 | 0.3030.3500.308 |
RN-Net | 0.867 | 18.599 | 0.4560.5390.254 | 0.3850.462 0.312 | 0.2890.3400.356 |
Method | MSE↓ | MAE↓ | r ≥ 0.25 mm CSI↑ POD↑ FAR↓ | r ≥ 1 mm CSI↑ POD↑ FAR↓ | r ≥ 2.5 mm CSI↑ POD↑ FAR↓ |
---|---|---|---|---|---|
WRF | 1.514 | 35.190 | 0.124 0.357 0.809 | 0.084 0.296 0.873 | 0.062 0.190 0.898 |
RF-PredRNN | 1.270 | 23.446 | 0.273 0.331 0.398 | 0.174 0.199 0.441 | 0.107 0.119 0.519 |
RF-ConvLSTM | 1.246 | 22.634 | 0.249 0.282 0.286 | 0.175 0.197 0.343 | 0.113 0.126 0.447 |
RF-TrajGRU | 1.271 | 23.345 | 0.263 0.313 0.367 | 0.180 0.208 0.459 | 0.115 0.129 0.533 |
RE-PredRNN | 1.384 | 26.327 | 0.268 0.347 0.475 | 0.171 0.211 0.549 | 0.092 0.109 0.650 |
RE-ConvLSTM | 1.303 | 24.053 | 0.269 0.307 0.302 | 0.180 0.202 0.348 | 0.109 0.122 0.431 |
RE-TrajGRU | 1.239 | 24.555 | 0.320 0.385 0.359 | 0.250 0.307 0.448 | 0.171 0.210 0.537 |
RE-RF-PredRNN | 1.198 | 23.227 | 0.331 0.423 0.415 | 0.239 0.295 0.472 | 0.152 0.182 0.547 |
RE-RF-ConvLSTM | 1.093 | 21.132 | 0.342 0.387 0.262 | 0.270 0.307 0.323 | 0.2010.2300.391 |
RN-Net | 1.118 | 21.859 | 0.3550.422 0.333 | 0.2770.335 0.418 | 0.1900.223 0.489 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
Share and Cite
Zhang, F.; Wang, X.; Guan, J.; Wu, M.; Guo, L. RN-Net: A Deep Learning Approach to 0–2 Hour Rainfall Nowcasting Based on Radar and Automatic Weather Station Data. Sensors 2021, 21, 1981. https://doi.org/10.3390/s21061981
Zhang F, Wang X, Guan J, Wu M, Guo L. RN-Net: A Deep Learning Approach to 0–2 Hour Rainfall Nowcasting Based on Radar and Automatic Weather Station Data. Sensors. 2021; 21(6):1981. https://doi.org/10.3390/s21061981
Chicago/Turabian StyleZhang, Fuhan, Xiaodong Wang, Jiping Guan, Meihan Wu, and Lina Guo. 2021. "RN-Net: A Deep Learning Approach to 0–2 Hour Rainfall Nowcasting Based on Radar and Automatic Weather Station Data" Sensors 21, no. 6: 1981. https://doi.org/10.3390/s21061981
APA StyleZhang, F., Wang, X., Guan, J., Wu, M., & Guo, L. (2021). RN-Net: A Deep Learning Approach to 0–2 Hour Rainfall Nowcasting Based on Radar and Automatic Weather Station Data. Sensors, 21(6), 1981. https://doi.org/10.3390/s21061981