Japan Geoscience Union Meeting 2021

Presentation information

[E] Poster

A (Atmospheric and Hydrospheric Sciences ) » A-AS Atmospheric Sciences, Meteorology & Atmospheric Environment

[A-AS04] Machine Learning Techniques in Weather, Climate, Hydrology and Disease Predictions

Fri. Jun 4, 2021 5:15 PM - 6:30 PM Ch.07

convener:Venkata Ratnam Jayanthi(Application Laboratory, JAMSTEC), Rajib Maity(Indian Institute of Technology Kharagpur), Swadhin Behera(Application Laboratory, JAMSTEC, 3173-25 Showa-machi, Yokohama 236-0001), Takeshi Doi(JAMSTEC)

5:15 PM - 6:30 PM

[AAS04-P03] Development of an integrated NWP-DA-AI system for 30-second-update 3D precipitation prediction

*Shigenori Otsuka1, Yasumitsu Maejima1, Pierre Tandeo2, Takemasa Miyoshi1 (1.RIKEN Center for Computational Science, 2.IMT Atlantique)

Keywords:precipitation nowcast, machine learning, data assimilation, phased array weather radar

The Phased-Array Weather Radar (PAWR), developed by the National Institute of Information and Communications Technology, Osaka University, and Toshiba Corporation, has been in operations since 2012 in Japan. The PAWR scans the whole sky in the 60-km range every 30 seconds at 110 elevation angles. Four PAWRs of the same type have been installed in Osaka, Kobe, Okinawa, and Tsukuba, and two similar ones of other types have been installed in Japan. Taking advantage of the PAWRs’ frequent and dense three-dimensional volume scans, we developed two systems: a high-resolution regional numerical weather prediction (NWP) system (SCALE-LETKF, Miyoshi et al., 2016a,b, Lien et al., 2017), and a three-dimensional (3D) precipitation nowcasting system (Otsuka et al., 2016).

Our 30-second-update 3D nowcasting system is running in real time since 2017; the system adopts an optical-flow-based algorithm in the 3D space. Because convective clouds evolve rapidly within a 10-minute forecast, sometimes the assumption of Lagrangian persistence is violated, and the prediction skill of the optical-flow-based system drops quickly with the forecast lead time. The SCALE-LETKF system, on the other hand, provides physically based predictions; therefore, we would expect that the NWP outperforms the nowcast for longer forecasts. Therefore, merging NWP and nowcast will provide better predictions compared to each of them.

We have been developing an algorithm to integrate data-driven approaches and process-driven approaches for precipitation nowcasting. Previously, we reported that a convolutional long short-term memory (ConvLSTM, Shi et al. 2015) extended to the 3D PAWR data (3D ConvLSTM) can successfully predict convective rain events in the 3D space. In addition, this neural network is designed to accept future data from NWP or optical flow-based precipitation nowcasts: an integrated NWP-Data Assimilation (DA)-AI system. We demonstrated that the use of future data from NWP/optical flow improved the prediction accuracy compared with the baseline experiment.

Recently, this integrated data-driven and process-driven precipitation nowcasting system was updated. First, the computational domain was extended in the vertical direction so that we can take advantage of upper air radar echoes. For that, pooling-unpooling is used to reduce the memory consumption and computation time. Second, the convolution operation in ConvLSTM was changed from 2D to 3D so that the new system can fully consider three-dimensional structures of radar echoes. Preliminary experiments with the new 3D ConvLSTM demonstrated that these updates improved the prediction accuracy. We will present the latest results in the meeting.