A raingauge based local classified average bias adjusted algorithm for radar quantitative precipitation estimate is presented. Based on the elaborate distribution of precipitation with a resolution of 0.01°×0.01° from the proposed algorithm, the meteorological conditions are examined for three flood peak events of Shijiao Catchment (707 km2) in Qijiang, Chongqing Municipality occurring on 28 May 2008， 5 August 2009 and 23 June 2010, respectively. A method of extrapolating lossless runoff series (given all precipitation transforming into runoff) is developed by combining the elaborate distribution of precipitation and the distribution of distance from catchment outlet calculated from DEM (digital elevation model). The results show that: the local classified average bias adjusted algorithm is a better way in the intense rainfall estimate than the local average bias adjusted algorithm. The lossless runoff forecast gives support to alert forecasters to the potential for flash flood.