- Категория: Обзоры и дискуссии
- Опубликовано 12.03.2014 06:57
PHYSICAL CORRELATION? IT MAY BE DANGEROUS!
Mikhail D.Gerasimenko, Nikolay V.Shestakov
Far East Federal University
Institute of Applied Mathematics FEBRAS, Radio st., 7, Vladivostok 690041, RUSSIA
Abstract. Unknown inestimable systematic errors that influence different observations give the reason for the physical correlations and not with algebraic ones. It is very important especially in the case of GPS measurements. Different investigators have obtained correlation coefficients regarding these reasons of up to 0.85 for identical satellite configuration and they are attempting to use these correlations or different correlation functions in the post processing of GPS measurements. As illustrated in our report, it may be very dangerous for practice and could lead to extraordinary wrong results. One of these unexpected effects may be artificial increasing the weight P of the final results. In limit case it may be that P→∞ if the weight even one observation p→0. This is absurd result clearly and it has no common physical sense. That's why the formal estimates of accuracy do not reflect the real situation with physical correlation of observations. It may be artificial improving of accuracy. We have met this phenomenon when we developed the software of optimal GPS geodetic network design for fault- mechanics studies. Considering this undesirable phenomenon it should be noted that we do not exclude that it is needed to re-estimate different methods of improving of GPS results that are involving in post processing measurements different non-diagonal covariance matrix of observations as well.