Analysis And Optimization Of Spatial Interpolation Method For Rainfall In Qinghai Province Based On Multi-Core Environment

Through the study of the past regional precipitation distribution,we can draw the conclusion that rasterization precipitation in the whole area is a hot topic at home and abroad in disaster prediction,hydrology,geography and other disciplines.Qinghai Province has complex terrain,dry climate and scarce rainfall,and the number of meteorological stations in Qinghai Province are very rare compared to the mainland provinces.Sparse and discrete site data increase the difficulty of research on the distribution of the overall space.Thus,we need to study the site observation data of discrete space,using spatial interpolation method.In the actual research process of spatial interpolation,the interpolation object is usually a large number of site data.Using the serial interpolation algorithm to interpolate the precipitation data will consume a lot of time.Therefore,the parallelization of interpolation algorithm in multi-core environment is particularly important.In this study,4 kinds of common spatial interpolation models were used to interpolate the daily precipitation data of 50 national sites in Qinghai province.Then the accuracy of the interpolation results are evaluated by the actual verification method,and the optimal interpolation algorithm of the regional precipitation climate resources in Qinghai province is determined.After the analysis upon the platform of SJY datacenter in Qinghai University and "Sunway Taihulight" on the computer system using openMP,MPI and Athread parallel programming model and parallel optimization of optimal interpolation algorithm.Finally,the optimization results under each scheme are evaluated through the acceleration ratio and the speed of acceleration.Results shows that:(1)ordinary Kriging interpolation interpolation method is most suitable for study of precipitation in Qinghai province area;(2)parallel method can effectively improve the efficiency of ordinary Kriging interpolation interpolation tasks;(3)for the common Kriging algorithm,the effect of the Kriging parallel acceleration algorithm based on MPI turns to be the best,and then the effect of Athread thread library is the second best.At last,the effect of OpenMP is the worst.