MPS-based Multi-source Precipitation Fusion Method In Qaidam Basin

Precipitation is the most important meteorological process affecting the regional water cycle.Improving the accuracy of precipitation data is of great significance to the early warning of flood and drought and related geological disasters.The traditional ground rainfall station observation and ground-based radar observation are limited by the establishment of the base station,and the spatial distribution of measured precipitation data is uneven.Although the spatial distribution of satellite remote sensing precipitation is uniform,the accuracy is not high due to the influence of the atmosphere.At present,with the rapid development of satellite detection technology,obtaining high-quality precipitation products through satellite inversion precipitation and precipitation fusion technology has become the mainstream research and development trend in the world.At present,the commonly used fusion methods focus on the spatial correlation and time dependence of multi-source precipitation data,resulting in the low spatial resolution and low regional accuracy of precipitation fusion results.Based on the multi-point geostatistics Filtersim simulation algorithm,this paper carried out the study on the precipitation of TRMM satellite and ground rain measuring station in Qaidam Basin,and added CLDAS soil moisture data as the reference factor.Firstly,the traditional interpolation method is used to directly Kriging the precipitation of TRMM satellite and station to generate precipitation data.Secondly,the precipitation data was decomposed into the local mean which could represent the precipitation trend and the local residual which could represent the difference of precipitation change,and the precipitation at unknown points was estimated by estimating the precipitation residual at unknown points.The TRMM precipitation residual was used as the training image,the precipitation residual of the station was used as the hard-data,and the Cokriging residual interpolation of the two was used as the soft-data for Filter Sim simulation,so as to generate the high-precision precipitation data.Then,the CLDAS soil moisture data,which can affect the spatial distribution of precipitation,was replaced by the Soft-Data of Filtersim simulation method as the auxiliary simulation of trend surface.Finally,the estimated results of different fusion methods were cross-verified and compared to illustrate the spatial correlation of different precipitation fusion results,and precipitation fusion products with high spatial and temporal resolution were generated.The main research contents are as follows:(1)Through the verification of statistics and comparison with the commonly used Kriging method,the FILTERSIM simulation method effectively improves the accuracy of rainfall fusion.The average absolute error of the original TRMM satellite data and the Kriging method are reduced,and the correlation coefficient is improved.The fusion effect is significantly better than that of the Kriging method.(2)The addition of soil moisture data such as FILTERSIM simulation can better capture the heavy rainfall events.The root mean square error is lower than that of Co Kriging method,and the correlation coefficient is improved.The results show that CLDAS data can provide more complete spatial information for precipitation fusion,and further improve the fusion accuracy.(3)The FILTERSIM simulation method can effectively correct the error of the original TRMM satellite precipitation data,capture the spatial precipitation information ignored by mathematical interpolation,and reflect the more real precipitation distribution.At the same time,according to the analysis of the three simulation results,the FILTERSIM simulation method based on multi-point statistics can be applied to the sparse Qaidam Basin,and effectively improve the accuracy of rainfall fusion.