DIAGNOSTIC INVESTIGATION OF SIMULATION BIAS WITH THE GRAPES-MESO MODEL FOR A TORRENTIAL RAIN CASE

Abstract:

In this paper, the numerical simulation bias of the non-hydrostatic version GRAPES-Meso (Mesoscale of the Global and Regional Assimilation and Prediction System) at the resolution of 0.18o for a torrential rain case, which happened in May 31st to June 1st 2005 over Hunan province, are diagnosed and investigated by using the radiosondes, intensive surface observation, and the operational global analysis data, and the sensitivity experimental results as well. It is shown in the result that the GRAPES-Meso could reproduce quite well the main features of large-cale circulation and the distribution of the accumulated 24h precipitation and the key locations of the torrential rainfall are captured reasonably well by the model. However, bias exist in the simulation of the mesoscale features of the torrential rain and details of the relevant systems, for example, the simulated rainfall that is too earlier in model integration and remarkable underprediction of the peak value of rainfall rates over the heaviest rainfall region, the weakness of the upper jet simulation and the overprediction of the south-west wind in the lower troposphere etc. The investigation reveals that the sources of the simulation bias are different. The erroneous model rainfall in the earlier integration stage over the heaviest rainfall region is induced by the model initial condition bias of the wind field at about 925hPa over the torrential rainfall region, where the bias grow rapidly and spread upward to about 600hPa level within the few hours into the integration and result in abnormal convergence of the wind and moisture, and thus the unreal rainfall over that region. The large bias on the simulated rainfall intensity over the heaviest rainfall region might be imputed to the following combined factors of (1) the simulation bias on the strength and detailed structures of the upper-level jet core which bring about significant underpredictions of the dynamic conditions (including upper-level divergence and the upward motion) for heavy rainfall due to unfavorable mesoscale vertical coupling between the strong upper-level divergence and lower-level convergence; and (2) the inefficient coupling of the cumulous parameterization scheme and the explicit moisture in the integration, which causes the failure of the explicit moisture scheme in generating grid-scale rainfall in a certain extent through inadequate convective adjustment and feedback to the grid-scale. In addition, the interaction of the combined two factors could form a negative feedback to the rainfall intensity simulation, and eventually lead to the obvious underprediction of the rainfall rate.