Based on East China regional mesoscale numerical forecast model system (Shanghai Meteorological Bureau WRF ADAS Rapid Refresh, SMB WARR), the time lag ensemble forecasting experiment with 7 ensemble members is conducted and routinely updated every hour with hourly output through 6 h forecast length. Evaluation of the hourly area rain of Shanghai from June 17 to September 30, 2011 shows that ensemble mean has better performance than any other members at light to heavy rain levels but worse performance than some members at rainstorm level. Meanwhile, the skill of the latest forecast is not the best and the light to heavy rain could be well forecasted 6 h in ahead. Furthermore, the probability forecast has an advantage over the ensemble mean, and has good directions for the happening of rain in the very short range, also the large (little) probability is more useful at light to moderate rain levels (heavy rain to rainstorm levels) especially the latest forecast.