Application And Research On Quality Control Method For Surface Temperature Observations

With the increasing number of surface meteorological observation stations, a large number of observations produced. The quality of the meteorological observations as the base of meteorological service and application are highly valued. So, quality control for surface meteorological observations is essential. But most of the existing quality control methods mainly focus on the research of observation itself, ignoring the environment influence on the quality control effect. In view of this, this paper is mainly focused on the quality control method from multi-station network, and analyzes the application situation for surface hourly temperature observations in different environment, the main work are as follows:General quality control methods for meteorological observations are in temporal dimension or spatial dimension, which are very single. Considering this, Auto Regression based on Inverse Distance Weighting, defined as AI method in this paper is proposed. It enables quality control in both temporal and spatial dimension. In temporal dimension, the quality control for target station is by auto regression; in spatial dimension, the quality control for target station is by the inverse distance weight between neighboring stations and target station, combing both of them effectively, quality control for four cities in Jiangsu province with different geographical conditions and station density are realized. It shows that, AI method can mark suspicious data effectively and is equipped with good quality control effect; it is with different performance effect in different areas, and applicable in both plain and hilly regions with high station density, but it exists limitation in regions with complex environment.Considering the white noise in meteorological observations, Empirical Mode Decomposition (EMD) is used to reduce or eliminate the effect from it, the signal is divided into some Intrinsic Mode Functions (IMF) and a residue, then the noise is eliminated by rebuilt the signal through the IMFs and residue, so AI method based on EMD, called EMD-AI for short is proposed. From the experiment results, we can see that the temperature observations denoised by EMD are used to forecast through AI method, its prediction precision and stability are all improved.The AI quality control effect gets improved by EMD, but another problem that AI method exists limitation in complex environment is still not solved. From this point of view, Spatial Panel Data (SPD) in spatial econometrics is introduced into meteorological observation quality control. Considering spatial effect in SPD and correlation between temperature and relative humidity, we put the relative humidity as an explanatory variable, then analyze SPD after arranging the observations as the requirement of SPD, and comparing SPD with the EMD-AI method. It shows that, the quality control effect of SPD method improved a lot in relative to EMD-AI method; and if the terrain is relative simple and the station distribution is intensive, EMD-AI method is the better choice; otherwise, SPD method is the better choice.After a large number of contrast experiments between the proposed methods in this paper and other methods on multi-station network quality control. It shows that the proposed methods are feasible in quality control for surface meteorological observations, and they are better than other methods, besides, different methods are applicable in different regions with different geographical conditions and station density.