Research Of Quality Control For Surface Temperature Observations Based On Geostatistics

Surface observations assimilation technique is conducive to improving the level of Numerical Weather Prediction(NWP),and the effective Quality Control(QC)of surface observations is the primary task of developing surface observations assimilation technique.In view of this,on the basis of analyzing the autocorrelation,factor correlation and distribution of surface temperature,a series of QC algorithms for surface temperature observations from three aspects:function modeling,terrain complexity and density of stations were proposed in the paper,and these algorithms were applied to different cases.Considering that traditional method can't take the autocorrelation of surface temperature and the space distribution of neighboring stations into consideration,a new method called Improved Ordinary Kriging(IOK)was put forward in this paper and it was applied to QC of surface temperature.In IOK method,Genetic Algorithm was adopted to fit the improved Semivariance function on the basis of constructing new fitness function to imbalance the contribution of each neighboring station and further improved the performance of the method.Aiming at the limitation of the IOK method in the area with high terrain complexity,the paper introduced the Gradient plus Kriging(GK)to QC of the surface temperature;results showed that the GK method outperforms IOK and SRT in the regions with higher terrain complexity.Aiming at the limitation of the IOK method in the regions with low density of stations,the paper introduced the Co-Kriging(CK)to QC of the surface temperature according to relationship between temperature and Relative Humidity;results showed that the CK method can make up for the influence of the density of the stations and it is superior to the IOK and SRT in the regions with lower density of stations.After several contrast tests,the QC algorithms based on Geostatistics can be applied effectively to data interpolation for stations which are lack of observations in the case of different terrain complexity and density of stations,or test doubtful and wrong data exist in the original surface temperature observations.At the end of the paper,we discussed how to select the range of the neighboring stations around the target stations in different regions,and analyzed the applicability of each QC method in the whole country.Generally,GK method is used in the regions with high complexity;the CK method is priority selection in regions with lower density of stations.In the regions with low density of stations,if terrain complexity plays a leading role,GK method will still has a certain advantage.In the regions with high density of stations and low terrain complexity,the disparities between three methods are small,the Spatial Regression Test(SRT)is preferred to save the computational cost.