| Climate change is a major part of global environment change, and it has beenconcerned by sectors of the community. Climate change has become one of theimportant driving factors, because there is the exchange of energy and matterbetween the climate system and terrestrial system. The temperature is one of themajor climate variables, and it plays a vital role in the process of land use/coverchange. In-depth study and discussion on the relationship between temperaturechange and land use/cover change, is an important direction of the land changeresearch, and has important significance for land use planning and agriculture zoning.However, the study needs to be carried out on the basis of data, and its resultsdepend on the quality of the basic data. Usually, because of the impacts by a varietyof non-natural causes, the quality of meteorological data is distinct, especially for thehistorical data. Using the documented meteorology data in northeast of China, thispaper reconstructs the temperature series for parts of the observed stations. On onehand, the available data is retained, and the unavailable data is removed, whichimproves the reliability and accuracy of the data; on the hand, the fragmentedsequence data that has little value is used as the auxiliary information for thereconstruction, which is useful for improving the utilization of data.This paper made a detailed analysis and study for the entire process of thereconstruction. The source data contains57temperature series, and the time-scale isfrom1909-1950. Firstly, we made a preliminarily control the quality of source data.We followed the idea that gradually control from macro to micro, and used a methodthat combined the qualitative way and quantitative way. In order to achievepreliminary quality controlling, we set a threshold value for the data series or itsstatistics, and then removed the abnormal values which are greater than or less thanthe threshold value. Secondly, we completed the reconstruction of data series. We combined the data from neighboring stations to form a long-term data series, inwhich the missing data was filled by the nearby data values. Specifically, we used aninverse distance weighted method and the expectation maximization method to fillthe missing data values or to extend the data series. In addition, from internalstructure of the series, we adopted a standard normal test method to make ahomogeneity test for the reconstructed series, and revised the test results.The results from the preliminary quality controlling show: the wrong/missingrecord is one of the important factors that affect the quality of the temperature series;the inter-annual variability of temperature is different, in which the variabilitybetween June to September is smaller, and it is larger between October to thefollowing May.The analysis on the correlation between the data series shows: the spatialvariability of the data series has less effect on the correlation between series, whilethe quality status (abnormal and missing values) is its main factor.From the homogeneity test and revision for the temperature series, there shows awell consistency between most of the interpolated series (revision series) and thetrue series. However, in the case of a significance level of95%and specificheterogeneity year, the results from the revision show:12mutation points belongs tonatural variation, while others are caused by the process of revision, and most ofthem are natural structure mutation. The final reconstructed dataset contains46series.The conclusion: through the quality controlling, we can remove the false datavalues and save the true values, and improve the quality of the data series; by usingan average value of a set of reference records, we can exclude the impacts that fromthe possible heterogeneity of a single station records and the spatial-temporalvariability of the temperature; the higher the quality of the original data series, thegreater the value of the reconstructed dataset. |
PDF Full Text Request