Abstract: The hailstone occurred in Hubei province from 1982 to 1998 had been dirided analyzed in detail. These cases had been divided into two patterns. The same characters of two patterns were that the potential instability were severe before occurred. There were severe increasing humidity and warm in middle-low level and the surface in the front trough pattern, and it resulted in energy level increasing in middle and low level. However, the dry and cold air had no great change in the middle and high level. The trigger of the cases were mainly due to hori zontal energy front and vertical instability increasing. There were severe dry and cold advection in the middle and high level in behind trough pattern. It resulted in humidity and warm decreasing. The weak humidity and warm increasing existed in the middle and low level. The trigger of the cases were mainly due to horizontal energy front increasing in middle and low level. It prompted vertical instability energy releasing. The sum of horizontal energy front ΔθseH/Δn and vertical instability ΔθseH/Δz predicted hailstone region as a criterion. In an ordinary operation, the accuracy of prediction is increased with other conditions. As a result, the cumulus convection paramet erization of MAPS model is improved, i.e, the instability condition of cumulus convection parameterization is improved. The cases in March to May from 1999 to 2001 had been forecasted using improvement MAPS model. So it shows that the MAPS model is able to forecast hailstone region.