Abstract: Climate disasters often appear,especially abnormal rainfall.Normal statistical variables usually being mean values are gotten at a fixed time scale.Areas having the same one statistical variable behave very differently in precipitation course.Thus normal statistical variables can nearly reflect strong fluctuations of precipitation variation.In order to analyze unusual precipitation and search for its relationship on different time scale,we have used fifty years' monthly mean precipitation and analyzed the relative precipitation variation on different time scale from one to twenty years.By making use of random increments of climate series and calculating its high-ordered probability density function,we got hierarchy variables standing for precipitation variation on different time scale and different fluctuation.Through dimensional analysis,we concluded a relationship of hierarchy variables and found that its variety index could identify different preciptation pattern,and reflect strong fluctuations of precipitation variation series.We calculated the variety index of 160 areas all over China.Its value varied from minimum 0.38 to maximum 0.82.We also found that the values accord with actual precipitation variation series.This helps us to judge the possibility of drought and waterlog in an area.The relationship of hierarchy variables maybe helps us to develop short-time climate forecast ability.