Research On Quality Evaluation And Correction Method Of Satellite Telemetry Precipitation Data

Precipitation,as an important part of the water cycle,is an important element of the study area or basin hydrological cycle system.Precipitation data with high spatial and temporal resolution is the basis for studying regional hydrological processes and exploring regional hydrological changes.It is of great significance for industrial and agricultural production,water conservancy development,drought and flood monitoring and prevention.How to obtain precipitation data with high spatial and temporal resolution is also an important research topic in the field of hydrometeorology.The traditional method to obtain precipitation data is by deploying a network of rainfall stations with a certain spatial density and directly observing the precipitation at regular intervals.The new method of observation is to estimate precipitation and its spatial distribution by weather radar or meteorological satellites.Although the traditional ground station network observation method can obtain the accurate precipitation at each station,it has obvious limitations due to uneven distribution of ground stations and precipitation data is spatially discontinuous.Although radar and satellite estimates of precipitation information have high temporal and spatial resolution,their monitoring data often has large errors.Therefore,it is of great significance to study the use of ground observation precipitation,radar precipitation,and satellite precipitation data to obtain high-precision,high-temporal resolution precipitation data.Taking the Beijing-Tianjin-Hebei region as an example,this paper first evaluates the accuracy of satellite precipitation data,and then focuses on the fusion method between satellite precipitation data and ground station network observation precipitation data.The accuracy of the combined precipitation data and the feasibility of the fusion method were evaluated.The research goal is to obtain precipitation data with high precision and space-time resolution,and improve the application value of satellite precipitation products.Firstly,the accuracy evaluation system of satellite precipitation products was established,and the accuracy of GSMaP_NRT satellite precipitation products was evaluated on the daily scale and monthly scale.The results show that the satellite precipitation data is not high in both time scales,and there will be a phenomenon of high value overestimation,and the false positive rate is higher at the daily scale.Overall,the accuracy of monthly-scale precipitation data is higher than the daily scale,and theaccuracy of daily-scale precipitation data is very low.In order to improve the accuracy of satellite precipitation data,two machine learning methods(random forest and support vector regression)were used to construct a fusion model of satellite precipitation data and ground station observation precipitation data.Apply this fusion model to correct the GSMaP_NRT satellite precipitation data and evaluate the corrected precipitation data.Analyze the effect and feasibility of the fusion model by comparing the accuracy evaluation indicators before and after correction.The results show that the quality of satellite precipitation data after fusion correction is greatly improved,R~2 is greatly improved,RMSE and Bias are significantly decreased;the satellite precipitation data after fusion correction has improved the success rate of daily precipitation events and the error rate is reduced;However,the model-corrected precipitation data sometimes underestimates the precipitation of extreme precipitation events.This paper provides a multi-source precipitation information fusion correction framework that can consider more precipitation factors and can integrate more sources of precipitation data at the same time,and the fusion effect of the method does not depend entirely on the accuracy of the original satellite precipitation data.