Renmin Zhujiang (Jan 2022)
Spatial Dimension Analysis and Judgement of Abnormal Rainfalls
Abstract
The automatic reporting system of water levels and rainfalls is widely used in flood control,hydrology,and meteorology in China.The automatically measured rainfall data is one of the conditions triggering flood control early warning,and its quality and accuracy directly affect the credibility of the warning.In order to avoid false warnings,it is necessary to analyze and filter abnormal rainfalls in real time.This paper explored the correlation of rainfalls in spatial planes and compared the application effects of four statistical methods including the Pauta criterion,Chauvenet criterion,Grubbs test,and Dixon test in the spatial dimension,so as to infer whether the rainfall at a certain point is abnormal.Specifically,the Chauvenet criterion has the optimal comprehensive performance,and its accuracy,precision,recall rate,and F1 score are 0.86,0.78,0.83,and 0.80,respectively.In addition,the Grubbs test and Dixon test obtain similar results,but they both are slightly worse than the Chauvenet criterion since there may be many abnormal rainfalls in regional groups.The Pauta criterion has the worst performance,but its precision is the highest,which is 0.97.In addition,it is effective to optimize rainfalls that have been judged to be abnormal many times by algorithm flows,and the algorithm precision can be significantly improved.It has been proved that it is feasible to judge whether the rainfall at a certain point is abnormal from the spatial plane,which can effectively help water conservancy supervision departments to improve the quality of early warning and reduce labor costs.
