Automatic Minimum and Maximum Alarm Thresholds for Quality Control

Abstract

Automated quality control is a necessity for meteorological measurement networks because of an ever-increasing number of measurements. One of the basic quality control tests is to check that the observations fall into a range of acceptable values. This paper proposes an algorithm for the automatic determination of maximum and minimum thresholds for such a test. The automated algorithm results in more statistically justified thresholds compared to the operator-fed values, which some quality control systems normally use. The suggested algorithm is a tuning-free method for the determination of daily minimum and maximum limits. The basis of the algorithm is, first, to collect the essential statistics of observations during their annual cycle and, second, to use these statistics to determine the acceptable observation range. This paper introduces a semirecursive version with lower computational requirements and discusses additional algorithmic details. The algorithm is demonstrated through numerical examples, which show that limits converge to satisfactory values in a few years.