适用于温度和气压秒级数据的质量检查方法初探

利用自动气象站秒级观测数据, 应用一种基于百分位阈值法的界限值检查算法, 对温度、气压2种秒级数据进行了界限值检查试验。结果表明: 针对试验站温度、气压秒级数据进行质量检查, 采用百分位(p=99.9%和p=0.1%)阈值法的检查效率较高, 且具有误判率较低的特点。采用6种百分位阈值法系数组合方案以及2种标准差法进行检查试验, 其中有3种方案的“标记”率低于给定的统计预期值, 误检率较低, 检查结果较好, 检查出的“标记”数据分别为33个、3个和0个, 标记率分别为0.076%、0.007%和0.000%。综合分析, 以1 min滑动窗口结合30 min时间区间系数组合方案的百分位阈值法检查效率最优, 具有“标记”数据少, 误判率低, 且计算机运算负荷小的特点。对连续30 d的温度和气压秒级数据运用该算法检查, 结果均较好。对人工构造的错误数据, 能准确检出。该质量检查算法可应用到无长年代历史观测资料的气象站、缺少临近站对比的气象站以及新建气象站使用, 对于地处偏远、条件恶劣的气象站也同样适用。

Quality control of seconds temperature and pressure data

Based on the high frequency observation data obtained from the automatic weather station, a threshold check algorithm developed using the percentile threshold method (PTM) was used to conduct a detection test to seconds temperature and pressure data.Specifically, six coefficient combination schemes for PTM and two standard deviation methods were adopted to conduct the detection test.Results show that the developed method has higher efficiency and lower misjudgment rate.In these tests, three schemes indicate better test results with a lower flagging percentage relative to the given statistical expected value.The number of "flagging" data for three schemes are 33, 3, and 0, the flagging percentage are 0.076 %、0.007 %, and 0.000 %, respectively.Among three schemes, the scheme combining 1-minute sliding window with a 30-minute time interval is optimal with the features of the least "flagging" data, the lowest misjudgment rate, and a small computation burden for the computer.The continuous 30-day seconds' temperature and pressure data were tested with this method, and the results are satisfying.Besides, this method is also able to test artificially constructed incorrect data.This algorithm could be applied to check those meteorological data at stations that had a short history, are newly constructed or outlying, and have bad conditions and therefore is helpful to identify the emergent problems in the data collection phase and to improve the automation level of data quality check.