Abstract
Numerical sensitivity experiments are made to investigate better combination of physical schemes in a meteorological numerical model, i.e., The Weather Research and Forecasting (WRF) by comparing numerical results with radar-AMeDAS data. The numerical experiments are carried out in three nested computational domains to simulate the heavy rainfall event in northern Kyushu in July 2012. It is difficult to predict accurately precipitation at a local point even through the latest meteorological model. The present study also examines whether computational precipitation averaged within spatial scale can provide practical precipitation or not. The time lag of hourly precipitation and the relative error of total precipitation simulated by the WRF model vary depending on the scheme combination and the spatial scale for averaging. The WRF model with the better scheme combination reproduces approximately spatial distributions of 12-hours accumulated precipitation from the radar-AMeDAS data. The present results suggest at there is an effective spatial averaging scale for evaluating practical precipitation.
