Statistical distances of measurements for quality control

Abstract

In widely distributed measurement systems, the first guess for the station measurement quality control is to compare the measurement to the neighbor stations. The basic degree of neighborness is often determined by the inverse distance of the measurement stations. However, this will not be descriptive in statistical sense if the system behind the observations is spatially complex or the sensors suffer from quality issues. This paper suggests a framework for measuring the statistical distance between stations within the system. The measures are developed with an assumption of large-scale observation systems, and therefore the computational and database access requirements are desired to keep as low as possible. The special emphasis of the numerical examples is on the meteorological measurements.