Construct Meteorological Similarity Network And Missing Meteorological Elements Data Interpolation

Meteorological elements observation is the basis of meteorological services work. Ground observations of meteorological elements are an important part of meteorological elements observed, it is a process of change the weather conditions and in certain ground spatial extent systematically and continuously observe and measure, for weather forecasting, climate analysis, scientific and meteorological services provide an important basis. Meteorological elements observation is an important task for each weather station. With the development of meteorological services, the number of surface weather observation stations has reached a certain scale. Each individual weather station is characterized by meteorological elements can be observed within a range where the region, within the region to meet the on-site weather information needs. But our ultimate concern is how these isolated meteorological observation stations are connected in a meteorological observation network, study the statistical characteristics of the entire meteorological observation network, the network using the method as a powerful tool for understanding regional meteorological system. This paper is how to build a meteorological observation network of domestic and foreign research focus of this issue.Due to meteorological equipment failure, for various reasons, such as transmission of computer network anomalies, meteorological elements missing data leak test phenomenon exists in varying degrees. The high-quality, full meteorological element data is the basis for high availability of meteorological analysis. Therefore, the missing data meteorological elements using reasonable and effective manner interpolation, making meteorological feature datasets complete and correct is important.To solve the above problem, this paper describes the main work from the following two aspects:(1) Meteorological similarity network constructionFirst, this paper proposes a method to construct meteorological observation network based on similarity. The first step to build a network of meteorological observations is defining the network nodes and connections. Node of the network is the time sequence of a meteorological observation stations of meteorological factors; network connection is the statistics associated between two-node. As used herein, the similarity between the methods to calculate the meteorological observation stations meteorological elements defined time series connection to the network. Common similarity measure is calculated directly similarity of two data objects directly from the whole. For a time series of meteorological elements, this paper presents a meteorological elements one by one to determine whether the two time series data at the same time and then draw on similar overall similarity between two time series methods, to avoid large differences exist meteorological element data time the impact on the similarity measure amplification problems, but can also handle missing data containing similarity calculations. For a time series of meteorological elements all meteorological observation stations, calculates the similarity between any two, got meteorological observation stations of the meteorological elements similarity adjacency matrix. For the adjacency matrix of values, select a threshold for the definition, the matrix is greater than the threshold value has connection points for the adjacency matrix to obtain similar meteorological elements of the network. Binding site location, you can use a graphical representation of the network in the form of put out.(2) Meteorological element data interpolationIn this paper, to improve the existing interpolation method, using the similarity interpolation, two interpolation method and comprehensive three single imputation method-spatial interpolation of meteorological elements similar time interpolation and linear interpolation of mixed interpolation missing data interpolation. To interpolate the missing data, missing data are estimated by known data. Geography’s first law states:things distributed geospatial locations are linked with each other, but the similarity is greater than the similarity between distant things close distance between things. Geospatial of this similarity is the basic assumption of the existing interpolation research. Existing imputation methods are based on this similarity large space geospatial interpolation. Time Interpolation means that in the time dimension to get an interpolated value meteorological elements meteorological elements by analyzing time series features. Similarity interpolation refers site. meteorological computing similarity between meteorological elements, using the site to be interpolated meteorological elements similar site meteorological data interpolation factor when the missing data interpolation of meteorological elements. Two step interpolations get first interpolated values in a single imputation method, and then use another single imputation method to reduce the error of interpolation values.Using two different scales of data sets:1) Domestic 160 meteorological stations in 64 years of monthly data meteorological elements; 2) In Chengdu 13 national meteorological stations half year hourly meteorological elements meteorological observation data to build the network. Meteorological observation network analysis constructed, all similar sites in the geographic space are relatively close, but not the more similar the closer distance sites, some very close to the network nodes are not connected. It also reflects the natural geographic features of spatial complexity. Use Chengdu 13 meteorological stations use half year hourly meteorological element data as the basis of meteorological elements experimental research data interpolation. Select the part of the missing data marker, and then use different interpolation methods accurate interpolation and analysis of the results verify the feasibility and accuracy of the proposed interpolation method.