The Quality Control of Surface Monthly Climate Data in China

It is generally agreed that outliers detection as well as outliers identification is of primary importance to quality control (QC) of observational data. Using some traditional quality techniques such as high-low extreme check, confidence limit control, internal consistency check etc, China historical surface meteorological data has been examined over and over, but there is a wide variety of erroneous values that are not been detected yet. If the continuity and distribution state of a data series and the outliers existence are not been understood well beforehand, some special erroneous values cannot be detected. The surface climate data series become more complex and inhomogeneous as a result of station moves, changes in the environment surrounding a station, and frequent changes in observational criterion in China. Therefore, the distribution of the data series from a great number of stations in China is not a normal distribution. Though discontinuities and inhomogeneities in time series are not of the field of QC, they have an effect upon traditional QC result. On the other hand, there would be problems in the sequential monthly climate data even the data among years as a result of measurement instrument errors, that of instrument calibration, a gradual shift in the physical characteristics of the instrument apparatus, or misoperation by observers etc if the above problems could not be solved in time. Perhaps the kind of data are not of great difference to the normal data, but they have certain impact on climate analysis. After analyzing the inhomogeneities, distribution state of the series and erroneous data in existence in China historical surface climate data, the QC method of surface monthly climate data in China has been developed, which is a breakthrough to traditional QC techniques of monthly climate data. It turns out that the quality control of China surface monthly climate data should include the following three steps: The check of continuous erroneous data after integrating the 12 monthly time series into a new individual series; the temporal check and spatial check of outliers after time series converted from likely inhomogeneous distribution to homogeneous one; manual advanced identification of continuous suspicious data and outliers. With the above QC method, about 250000 surface monthly climate data of base stations in China from 1971 to 2000 is examined. The climate data contains more than 10 monthly variables: Surface air temperature, surface air relative humidity, wind speed, skin surface temperature, eight layers of soil temperature, sunshine duration, pan evaporation, frozen earth depth and snow depth etc. 136 erroneous monthly climate data referring to various variables are detected in total. The causes of erroneous monthly data according with original data such as hourly data is the following: Use other station data or monthly data to substitute true data; miss-recording such as enlarging or reducing 10 times in original data; the original data should not be equal to 0, but the record data is 0; the measurement instrument is mal functioning.