Research On Spatial Quality Control Method For Surface Temperature Observations

Nowadays,numerical weather prediction has become a weather forecasting method that affects social development and daily life,and the assimilation technology of meteorological data helps to improve the accuracy of numerical weather prediction.Due to quality problems,surface meteorological observation with low utilization rate is a major problem to limit the development of meteorological data assimilation technology.Therefore,the research on the quality control of surface meteorological observation is one of the most important tasks for the development of numerical weather forecasting.Based on the spatial correlation of surface temperature observation data and the distribution characteristics of surface stations,this paper establishes a series of surface temperature spatial quality control methods based on model accuracy,model optimization of station-dense regional,and supplement of new stations or stations with long missing values.The main contents are as follows:The Moran 'I and semivariogram are used to analyze the spatial distribution characteristics of surface temperature and provide theoretical support for spatial quality control;the artificial random error method invented by Professor Hubbard is used to simulate the observations,and hypothesis test is used to select the quality control parameters to support the evaluation of the quality control model;aiming at the accuracy of the model,an artificial fish swarm algorithm is used to optimize the parameters of the random forest algorithm,and a spatial quality control method(ARF)based on improved random forest is proposed;in view of the poor timeliness and no obvious improvement in accuracy of optimization models for station-intensive regions,spatial regression test is used to screen neighboring stations with high correlation to form new datasets,in combination with the random forest algorithm to build models,a spatial quality control method based on spatial regression test-random forest is proposed(SRP);for some new stations or stations with long missing values,the principal component analysis algorithm is used to map different meteorological elements into the same coordinate system,and random forest algorithm is used to build the model,a spatial quality control method based on principal component analysis-random forest is proposed(PCA-RF)The results of multi-group comparison experiments show that the spatial quality control model based on the improved or combined random forest algorithm can be effectively applied to regions with different spatial correlation and station density,at the same time,effective quality control can be performed for new stations or stations with long missing values.The ARF method is suitable for most stations with no timeliness requirements,it is more universal in different terrain and station density areas,and has higher model accuracy than SRT and IDW methods,however,the precision of the area with high station density is not obvious and the model operation takes a long time;the SRF method is suitable for areas with high station density,which can effectively reduce the running time of the model without reducing the accuracy of the model,however,optimization effect is not obvious at station sparse regions;the PCA-RF method applies to stations that have long missing values,in the case where station surface temperature data is missing,a new data set is supplemented by other meteorological elements,at the same time,it can also increase the diversity of data and supplement the missing data of stations.The above three kinds of spatial quality control methods basically meet the quality control requirements of most of the surface stations in China,and can be adjusted according to the demand and actual situation of the stations.