G11A-0965
Avoiding negative value with three cornered hat method, with application to Polar Motion data
Monday, 14 December 2015
Poster Hall (Moscone South)
Olivier de Viron, Institut de Physique du Globe de Paris, Paris, France, Michel J. Van Camp, Observatoire Royal de Belgique, Brussel, Belgium, Paul Rebischung, IGN Institut National de l'Information Géographique et Forestière, Paris Cedex 13, France and Christian Bizouard, Paris Observatory, Paris, France
Abstract:
To determine the quality of a measurement time series, there are only three methods: (1) we know the time series with a infinite accuracy by some other mean, and we make a comparison, (2) we measure a phenomenon for which we can determine the signature exactly, and (3) we can use a set of 3 or more independent measurements of the same time series. This last method is known as the three-cornered hat. It allows to estimate the variance of the noise of time series. Applied to geodetic time series, mostly if the time series are short and of inhomogeneous quality, the method returns quite often negative values of the variance, which is unphysical. We discuss the existing solution to this issue, and provide an alternative, more robust solution. We apply this solution to polar motion data.
