Chiashi Muroi

The Operational JMA Nonhydrostatic Mesoscale Model

Abstract An operational nonhydrostatic mesoscale model has been developed by the Numerical Prediction Division (NPD) of the Japan Meteorological Agency (JMA) in partnership with the Meteorological Research Institute (MRI). The model is based on the MRI/NPD unified nonhydrostatic model (MRI/NPD-NHM), while several modifications have been made for operational numerical weather prediction with a horizontal resolution of 10 km. A fourth-order advection scheme considering staggered grid configuration is implemented. The buoyancy term is directly evaluated from density perturbation. A time-splitting scheme for advection has been developed, where the low-order (second order) part of advection is modified in the latter half of the leapfrog time integration. Physical processes have also been revised, especially in the convective parameterization and PBL schemes. A turbulent kinetic energy (TKE) diagnostic scheme has been developed to overcome problems that arise to predict TKE. The model performance for mesoscale ...

An 80-Fold Speedup, 15.0 TFlops Full GPU Acceleration of Non-Hydrostatic Weather Model ASUCA Production Code

Regional weather forecasting demands fast simulation over fine-grained grids, resulting in extremely memory- bottlenecked computation, a difficult problem on conventional supercomputers. Early work on accelerating mainstream weather code WRF using GPUs with their high memory performance, however, resulted in only minor speedup due to partial GPU porting of the huge code. Our full CUDA porting of the high- resolution weather prediction model ASUCA is the first such one we know to date; ASUCA is a next-generation, production weather code developed by the Japan Meteorological Agency, similar to WRF in the underlying physics (non-hydrostatic model). Benchmark on the 528 (NVIDIA GT200 Tesla) GPU TSUBAME Supercomputer at the Tokyo Institute of Technology demonstrated over 80-fold speedup and good weak scaling achieving 15.0 TFlops in single precision for 6956 x 6052 x 48 mesh. Further benchmarks on TSUBAME 2.0, which will embody over 4000 NVIDIA Fermi GPUs and deployed in October 2010, will be presented.

145 TFlops Performance on 3990 GPUs of TSUBAME 2.0 Supercomputer for an Operational Weather Prediction

Abstract Numerical weather prediction is one of the major applications in high performance computing and demands fast and high-precision simulation over fine-grained grids. While utilizing hundreds of CPUs is certainly the most common way to get high performance for large scale simulations, we have another solution to use GPUs as massively parallel computing platform. In order to drastically shorten the runtime of a weather prediction code, we rewrite its huge entire code for GPU computing from scratch in CUDA. The code ASUCA is a high resolution meso-scale atmosphere model that is being developed by the Japan Meteorological Agency for the purpose of the next-generation weather forecasting service. The TSUBAME 2.0 supercomputer, which is equipped with 4224 NVIDIA Tesla M2050 GPUs, has started operating in November 2010 at the Tokyo Institute of Technology. A benchmark on the 3990 GPUs on TSUBAME 2.0 achieves extremely high performance of 145 TFlops in single precision for 14368×14284×48 mesh. This paper also describes the multi-GPU optimizations introduced into the ASUCA porting on TSUBAME 2.0.

Performance of Long-Term Integrations of the Japan Meteorological Agency Nonhydrostatic Model Using the Spectral Boundary Coupling Method

The spectral boundary coupling (SBC) method, which is an approach used to couple a limited-area model with a large-scale model, was introduced into a nonhydrostatic model. To investigate whether the SBC method works well in a long-term integration of a high-resolution nonhydrostatic model, two numerical experiments were conducted with a model having a horizontal grid interval of 5 km. In one experiment, the SBC method was employed, while it was not in the other experiment. The time integration in both experiments was over a 40-day period. The nonhydrostatic model was nested into objectively analyzed fields, instead of the forecasts from an extended-area model. Predicted patterns of sea level pressure and precipitation were compared with objective analyses, and data provided by the Global Precipitation Climatology Project (GPCP), respectively. The predicted rainfall amounts and surface temperature over the Japanese islands were statistically evaluated, making use of the analyzed rainfall and surface data observed by the Japan Meteorological Agency (JMA). All results examined in the present study exhibited better performances with use of the SBC method than those without the SBC method. It was found that the SBC method was highly useful in long-term simulations by a high-resolution nonhydrostatic model.

Simulations of Forecast and Climate Modes Using Non-Hydrostatic Regional Models

Two applications with a cloud-resolving model are shown utilizing the Earth Simulator. The first application is a case in the winter cold-air outbreak situation observed over the Sea of Japan as a forecast mode. Detailed structures of the convergence zone (JPCZ) and formation of mechanism of transverse convective clouds (T-modes) are discussed. A wide domain in the horizontal (2000 × 2000) was used with a horizontal resolution of 1 km, and could reproduce detailed structures of the JPCZ as well as the cloud streets in the right positions. It is also found that the cloud streets of T-modes are parallel to the vertical wind shears and, thus, similar to the ordinary formation mechanism as longitudinal convective ones. The second application is changes in the Baiu frontal activity in the future warming climate from the present one as a climate mode. At the future warming climate, the Baiu front is more active over southern Japan, and the precipitation amounts increase there. On the other hand, the frequency of occurrence of heavy rainfall greater than 30 mm h-1 increases over the Japan Islands.