K Computer Runs Largest Ever Ensemble Simulation of Global Weather

RIKEN: News & Media

Researchers at Japan’s RIKEN Advanced Institute for Computational Science have successfully run 10,240 parallel simulations of global weather on the 10-petaflops K computer–the largest number ever executed–using data assimilation to reduce the range of uncertainties.

A three-week computation of data from the ensembles for simulated global weather was made possible by the eight-fold improvement of the Local Ensemble Transform Kalman Filter’s efficiency by the EigenExa high-performance eigenvalue solver software. The researchers learned that remote observation, even further away than 10,000 kilometers, may have an immediate effect on the ultimate state of the estimation via analysis of the 10,240 equally likely estimates of atmospheric states.

The outcome suggests further research is needed on innovative techniques that can better leverage faraway observations, which could potentially lead to an improvement of weather forecasts.

Contributing to this milestone were several projects financed by Japan Science and Technology Agency CREST (Core Research for Evolutionary Science and Technology) programs. The projects included one focused on advancing big data assimilation technology for revolutionizing short-range severe weather prediction; a project to integrate big data and high-performance computing for yottabyte processing, and an effort to develop an Eigen-supercomputing engine using a post-petascale hierarchical model. Report

RIKEN is Japan’s largest comprehensive research institution renowned for high-quality research in a diverse range of scientific disciplines. Founded in 1917 as a private research foundation in Tokyo, RIKEN has grown rapidly in size and scope, today encompassing a network of world-class research centers and institutes across Japan.