Lognormal distribution of the observed and modelled neutral thermospheric densities |
| |
Authors: | A Bezděk |
| |
Institution: | (1) Astronomical Institute, Acad. Sci. Czech Republic, Fričova 298, 251 65 Ondřejov, Czech Republic |
| |
Abstract: | For analyzing measurements of any kind, it is important to estimate the probability distribution of the measurement errors.
When modelling the observations using least-squares fitting, the distribution of the errors plays a vital role in choosing
the merit function to be minimized, as unnormally distributed errors (e.g. outliers, or displaying asymmetry around the mean)
may substantially skew a least-squares fit of estimated model parameters. Using the CACTUS accelerometer data covering heights
of 230–750 km, we will show that the statistical relationship between the commonly used semi-empirical models of neutral thermospheric
density (MSIS, DTM) and the observed densities is consistent with lognormal distribution, i.e. the logarithm of the ratio
of the measurements to the predictions is approximately normally distributed. This experimental fact may be applied in modelling
the neutral thermospheric density.
bezdek@asu.cas.cz |
| |
Keywords: | space accelerometers atmospheric density thermosphere least squares estimate multiplicative errors |
本文献已被 SpringerLink 等数据库收录! |
|