Description and Analysis of Data and Errors in GPS Meteorology

Doctoral thesis, 2004

This report presents the relation between remote sensing techniques used for estimation of atmospheric water vapor content, and numerical weather prediction. The thesis makes an overview of the methods for estimation of the zenith wet/total delay (ZWD/ZTD) for radio waves in the neutral atmosphere, and a statistical description of data in relation to the sampling problem for atmospheric water vapor. The work includes a description of the ZWD/ZTD estimation errors, and the atmospheric contributions to the uncertainties of global models for propagation delay corrections. The models used in numerical weather prediction (NWP) require correct horizontal covariance structure of the GPS estimation errors in order ZTD data to be assimilated into the forecast system, under the assumptions of homogeneity and isotropy of the two-dimensional ZTD error random field. Three approaches to deriving horizontal correlations of these errors are considered. The first one uses an approximation of the true atmospheric parameter correlations. The second one uses simulations of common error sources that affect a network of GPS receivers, namely satellite clock and orbit model errors. The third one uses ZTD innovations (differences between the GPS estimates, and NWP model forecast fields). For the proper estimation of the correlations of a forecast error field, an ensemble assimilation algorithm is presented and analyzed as an alternative to the currently used "NMC" method. Issues regarding the linearity of the NWP model's observation operator are addressed in the light of the Kalman filter theory. The vertical distribution of the power of atmospheric variations is discussed in the context of local homogeneity and isotropy of turbulence. Radiosonde data are used to estimate the vertical profiles of the Kolmogorov power spectrum density function.