| The propagation delay induced by the electrically-neutral atmosphere has been recognized as the most problematic modeling error for radiometric space geodetic techniques. A mismodeling of this propagation delay affects significantly the height component of position and constitutes therefore a matter of concern in space-geodesy applications, such as sea-level monitoring, postglacial rebound measurement, earthquake-hazard mitigation, and tectonic-plate-margin deformation studies.;The neutral-atmosphere propagation delay is commonly considered as composed of two components: a "hydrostatic" component, due essentially to the dry gases of the atmosphere, and a "non-hydrostatic" component, due to water vapor. Each one can be described as the product of the delay at the zenith and a mapping function, which models the elevation angle dependence of the propagation delay.;This dissertation discusses primarily the accuracy of zenith delay prediction models and mapping functions found in the scientific literature. This performance evaluation is based on a comparison against 32,467 benchmark values, obtained by ray tracing one-year's worth of radiosonde profiles from 50 stations distributed worldwide, and comprised different phases: ray-tracing accuracy assessment, model development, and model accuracy assessment.;We have studied the sensitivity of the ray-tracing technique to the choice of physical models, processing strategies, and radiosonde instrumentation accuracy. We have concluded that errors in ray tracing can amount to a few centimetres, under special circumstances, but they largely average out for each station's time series of profiles.;In order to optimize the performance of the models, we have established databases of the temperature-profile parameters using 50 additional sites, for a total of 100 radiosonde stations. Based on these large databases, we have developed models for lapse rate and tropopause height determination, which have improved significantly the performance of models using the information.;From our model assessment we have concluded that the hydrostatic component of the zenith delay can be predicted with sub-millimeter accuracy, using the Saastamoinen model, provided accurate measurements of surface total pressure are available. The zenith non-hydrostatic component is much more difficult to predict from surface meteorological data or site dependent parameters, and the best models show values of root-mean-square (rms) scatter about the mean of a few centimetres in the zenith direction.;Notwithstanding the large number of mapping functions we have analyzed, only a small group meet the high standards of modern space geodetic data analysis: Ifadis, Lanyi, MTT, and NMF. For the total number of radiosonde stations analyzed, none of the mapping functions revealed themselves to be superior for all elevation angles. For elevation angles above 15 degrees, Lanyi, MTT, and NMF yield identical mean biases and the best total error performances. At lower elevation angles, Ifadis and NMF are clearly superior. As regards the rms scatter about the mean, Ifadis performs the best for an elevation angles, followed closely by Lanyi. |
