Sudden local severe weather is a threat, and we explore what the highest-end supercomputing and sensing technologies can do to address this challenge. Here we show that using the Japanese flagship “K” supercomputer, we can synergistically integrate “big simulations” of 100 parallel simulations of a convective weather system at 100-m grid spacing and “big data” from the next-generation phased array weather radar that produces a high-resolution 3-dimensional rain distribution every 30 s—two orders of magnitude more data than the currently used parabolic-antenna radar. This “big data assimilation” system refreshes 30-min forecasts every 30 s, 120 times more rapidly than the typical hourly updated systems operated at the world’s weather prediction centers. A real high-impact weather case study shows encouraging results of the 30-s-update big data assimilation system.

Effective assimilation of global precipitation: simulation experiments

Japan's new geostationary satellite “Himawari‐8” enables updating precipitation and flood predictions as frequently as every 10 min and to capture as early as possible the flood risk associated with rapidly changing severe weather. This study focuses on the advantage of the frequent update of precipitation and flood predictions by assimilating all‐sky Himawari‐8 infrared radiances in the case of September 2015 Kanto‐Tohoku heavy rainfall, a major flooding event in Japan. The analyzed tropical cyclone representation and moisture transport are improved by the Himawari‐8 data assimilation. Deterministic runs from the analyses with the Himawari‐8 data provide much improved forecasts of a strong precipitation band. Moreover, every 10 min updates of river discharge forecasts driven by these improved precipitation forecasts show a large improvement in forecast skill with longer lead times. The results demonstrate that the frequent refresh aids to capture a signature of flooding at an earlier stage.

Following the invention of the telegraph, electronic computer, and remote sensing, “big data” is bringing another revolution to weather prediction. As sensor and computer technologies advance, orders of magnitude bigger data are produced by new sensors and high-precision computer simulation or “big simulation.” Data assimilation (DA) is a key to numerical weather prediction (NWP) by integrating the real-world sensor data into simulation. However, the current DA and NWP systems are not designed to handle the “big data” from next-generation sensors and big simulation. Therefore, we propose “big data assimilation” (BDA) innovation to fully utilize the big data. Since October 2013, the Japan's BDA project has been exploring revolutionary NWP at 100-m mesh refreshed every 30 s, orders of magnitude finer and faster than the current typical NWP systems, by taking advantage of the fortunate combination of next-generation technologies: the 10-petaflops K computer, phased array weather radar, and geostationary satellite Himawari-8. So far, a BDA prototype system was developed and tested with real-world retrospective local rainstorm cases. This paper summarizes the activities and progress of the BDA project, and concludes with perspectives toward the post-petascale supercomputing era.

Current methods of assimilation of precipitation into numerical weather prediction models are able to make the model precipitation become similar to the observed precipitation during the assimilation, but the model forecasts tend to return to their original solution after a few hours. To facilitate the precipitation assimilation, a logarithm transformation has been used in several past studies. Lien et al. proposed instead to assimilate precipitation using the local ensemble transform Kalman filter (LETKF) with a Gaussian transformation technique and succeeded in improving the model forecasts in perfect-model observing system simulation experiments (OSSEs). In this study, the method of Lien et al. is tested within a more realistic configuration: the TRMM Multisatellite Precipitation Analysis (TMPA) data are assimilated into a low-resolution version of the NCEP Global Forecast System (GFS). With guidance from a statistical study comparing the GFS model background precipitation and the TMPA data, some modifications of the assimilation methods proposed in Lien et al. are made, including 1) applying separate Gaussian transformations to model and to observational precipitation based on their own cumulative distribution functions; 2) adopting a quality control criterion based on the correlation between the long-term model and observed precipitation data at the observation location; and 3) proposing a new method to define the transformation of zero precipitation that takes into account the zero precipitation probability in the background ensemble rather than the climatology. With these modifications, the assimilation of the TMPA precipitation data improves both the analysis and 5-day model forecasts when compared with a control experiment assimilating only rawinsonde data.

Typhoon Sinlaku (2008) is a case in point under The Observing System Research and Predictability Experiment (THORPEX) Pacific Asian Regional Campaign (T-PARC) with the most abundant flight observations taken and with great potential to address major scientific issues in T-PARC such as structure change, targeted observations, and extratropical transition. A new method for vortex initialization based on ensemble Kalman filter (EnKF) data assimilation and the Weather Research and Forecasting (WRF) model is adopted in this study. By continuously assimilating storm positions (with an update cycle every 30 min), the mean surface wind structure, and all available measurement data, this study constructs a unique high-spatial/temporal-resolution and model/observation-consistent dataset for Sinlaku during a 4-day period. Simulations of Sinlaku starting at different initial times are further investigated to assess the impact of the data. It is striking that some of the simulations are able to capture Sinlaku’s secondary eyewall formation, while others starting the simulation earlier with less data assimilated are not. This dataset provides a unique opportunity to study the dynamical processes of concentric eyewall formation in Sinlaku. In Part I of this work, results from the data assimilation and simulations are presented, including concentric eyewall evolution and the precursors to its formation, while detailed dynamical analyses are conducted in follow-up research.