Application of mesoscale meteorological model to environmental assessment

The conventional environmental assessment has been conducted by steady state plume models, considering an occurrence probability of meteorological variables. On the other hand, mesoscale meteorological models have been improved to be able to simulate local scale phenomena down to mesh sizes of 1000 m or less. Therefore, we are now developing a new assessment system using the open source, mesoscale model of RAMS (Regional Atmospheric Modeling System) in Japan. This system can simulate actual meteorological and diffusion phenomena under the unsteady continuous condition for one year, considering the interactive effect of topography and atmospheric stability. This system has an advantage of being to reduce the working time and the cost of meteorological observation for the environmental assessment. We developed two kinds of new techniques to simulate a gas diffusion over complex terrain during a long period by using a PC cluster; one is a new turbulence model called Non-Isotropic Turbulence model in full 3 Dimensions (NIT3D). The other is a new parallel computing scheme called Time Splitting Method (TSM).The NIT3D model can simulate the turbulent intensities caused by topographical effects better than the Mellor-Yamada model because it solves the full 3 dimensional equations for total kinetic energy (k) and turbulent length scale (l). Although the parallel computing scheme using a PC cluster is an efficient tool to reduce computational time, the conventional parallel computing scheme of the domain composition method reaches a maximum limit of high speed computing efficiency when the number of CPU's reaches about 32-48, depending on specific model and hardware configurations. The TSM divides a total simulation period into a number of CPU units, and each CPU solves the divided simulation time independently. For example, if a one year simulation will be conducted with 12 CPU's, each CPU will calculate the meteorology for a one month independently. Therefore, its parallel computing efficiency is about 100%, and computing time is inversely proportional with the number of CPU. This new RAMS code was validated against field experimental data on meteorology and gas concentration observed in Mt. Tsukuba by Japan Atomic Energy Research Institute (JAERI). The simulated results by the new RAMS code showed a good agreement with field data on meteorology and gas concentration, while calculated results by the conventional plume model underestimated ground-level gas concentration in the short distance from a source position and overestimated in the far distance.