逐时气温质量控制中界限值检查算法的设计

Algorithm Design of Quality Control for Hourly Air Temperature

针对自动气象站数据质量控制中,采用月气候极值对逐时气温资料进行检查所存在的不足,发展了一种基于日极值的小时界限值检查质量控制算法。为统计小时气温界限值,首先采用时间和空间插值原理,基于广义极值分布理论,推求了任意地点的多年日气温极值;然后基于数据插值技术和气温日变化规律,设计了任意地点逐时界限值计算方法。利用中国730个测站逐日最高、最低气温资料,设计了适用于全国范围的小时气温界限值,并采用2010年全国2400个自动气象站的整点、最高、最低气温等逐时资料对该算法进行了试验。结果表明,基于小时界限值质量控制算法与通常的日(月)气候极值的质量检查方法相比,具有漏检率较低,并可在无历史资料台站使用等特点。将该算法的检查结果再进行空间一致性检查,可较大幅度减少误检率。

Although range check based on month climatic extremes is one of the most commonly used data quality control methods for Automatic Weather Stations( AWS) data,if it is applied to quality control of hourly air temperature data,there exist two weaknesses:( 1) The larger threshold range of climatic extremes can make some abnormal data undetected;( 2) Large amount of newly built AWS has no climatic extremes. To improve these weaknesses,a range check method is developed based on hourly threshold,which is computed by day climatic extremes. In order to obtain the lower and upper limits of hourly air temperature threshold,firstly,the principles of time and spatial interpolation are applied to calculate the daily temperature extremes at any site based on the generalized extreme value distribution theory. Then,the computing method of hourly threshold at any site is designed by means of the daily change regularity of temperature and numerical interpolation technology. The daily maximum and minimum air temperature of total 730 AWS in China are applied to design and calculate the lower and upper limits of hourly air temperature threshold which can be used for nationwide. Meanwhile,the punctuality,maximum,minimum air temperature of hourly data produced by 2400 AWS located in the national scale in 2010 are used to demonstrate the effectiveness of our proposed control approach. The results showthat the quality control method based on the range check of hourly threshold has a lower undetected rate compared with the daily or monthly climatic extremes check. Especially,it can be adopted by those stations without historical data. Its check results can be strengthened by combining with spatial consistency check to reduce the false detection rate greatly.