Based on the daily precipitation data from 13meteorological stations in Hexi Region during 1960-2011

the methods of linear regression

inverse distance weighted

Mann-Kendall analysis and rescaledrange analysis were employed to analyze the spatial and temporal variability of precipitation extremes and predict its extremely precipitation variations in future.Four indices of extreme precipitation were studied.Theresults showed that over the 1960-2009period

the regionally averaged occurrence of annual maximum precipitations within 1dand 2～3dincreased whereas annual maximum precipitation of 4ddecreased.With theexception of annual maximum 4dprecipitation

the predicted trends of the other indices in future were opposite compared to the past.The annual maximum 1and 2dprecipitations correlated with annual total precipitation very well

which implies that annual maximum 1and 2dprecipitations are good indicators for annualtotal precipitation.The four indices differed substantially in spatial distribution with varying past and futuretendencies.In addition

abrupt changes in extreme precipitation indices could be readily identified at the 0.05confidence level

although they were not the same year.