Research On The Temporal And Spatial Characteristics Of Surface Rainfall Data And Its Quality Control Metho

rainfall is a key hydrological and climatic variable,and changes in rainfall can drive the occurrence of natural disasters and extreme events such as floods and droughts.The accuracy of monitoring and recording rainfall information will directly affect decision-making and planning in areas such as climate change,water cycle,and flood disaster early warning.Surface rainfall observation data mainly come from timed observations by automatic weather stations.There are a large number of random errors,systematic errors,and gross errors in China’s currently retained data.Using these data for meteorological forecasting and related scientific research may affect the final results.In order to solve this problem,it is necessary to carry out research on the quality control of surface observation data,and the method of rainfall quality control has always been a difficult point in the research.Starting from the characteristics of rainfall,this paper studies the quality control methods of rainfall data on a daily scale,mainly including the following contents:(1)The surface rainfall data from 63 meteorological stations in the Yangtze River Delta region of China from 1978 to 2021 were selected to study the temporal and spatial characteristics of rainfall.On a seasonal scale,the innovative trend analysis method was used to explore the overall characteristics and seasonal changes of rainfall with different intensities in the study area,while the dual ratio analysis method was used to explore the impact of human activities on rainfall in the region.(2)Considering the spatial autocorrelation of rainfall and the spatial sparsity of daily rainfall,this paper compares the applicability of three spatial interpolation methods,namely,Kriging method(OK),generalized Kriging method(GK),and water vapor transport inverse distance weighting method(WVT-IDW),in the study area at large spatial scales,local regions,and seasonal scales.The experimental results show that Water Vapor Transport Inverse Distance Weighting(WVT-IDW)is suitable for processing local spatial rainfall data,and is insensitive to the sparsity of spatial rainfall data.In terms of large-scale rainfall data interpolation,Generalized Kriging interpolation(GK)performs better,but its interpolation accuracy will decrease due to the sparsity of spatial rainfall.(3)Due to global warming,human impacts,and other factors,the localization of rainfall has become stronger,and daily rainfall changes sharply.The above multi station quality control methods have been challenged.Therefore,this paper proposes a seasonal short-term automatic encoder(ASSAE)single station quality control algorithm with attention mechanism based on the single station rainfall sequence from the Yangtze River Delta region.The model is composed of short-term mode and seasonal mode,which can capture the seasonal effects hidden in climate element information,thereby responding to sudden changes and extreme value prediction of rainfall.Combined with quantile loss functions,it effectively predicts zero rainfall values.Compared with several commonly used time series prediction models,ASSAE has proved its superiority.Finally,the error detection ability of ASSAE was evaluated,and the results showed that the model had a higher error detection rate for rainfall data with obvious seasonal trends.