Profound knowledge of the variable Earth rotation is required to derive accurate and reliable statements and parameters from space geodetic techniques. The geophysical know-how, which forms the basis of Earth rotation studies, benefits in turn from precise measurements and time-series of Earth rotation parameters (ERP). They allow science to broaden the expertise in the field of dynamic processes and interactions of the system Earth and thereby to improve global geophysical models. This dissertation deals with Earth rotation variations in the diurnal and sub-diurnal frequency range with the main objective of developing a new high-frequency ERP prediction model. It should, furthermore, answer the question, if an empirical ocean tide model, being the most independent data source, can be used to derive accurate ERP predictions which could potentially replace the current conventional model. The conventional prediction model for short-period ocean tidal effects on Earth rotation recommended by the International Earth Rotation and Reference Systems Service (IERS) has been found to yield imperfect geodetic results and introduces biases in Global Positioning System (GPS) orbits. An updated model for polar motion and Universal Time 1 (UT1) is, therefore, highly desirable by the scientific community. Due to the high precision of measurements, several empirical ERP models have been developed by harmonic analysis of the residuals of space geodetic techniques. However, they lack of independence, and single-technique models differ in inter-technique comparisons. There are, in addition, combination models determined from several techniques, presumed to be the most precise models currently available. On the other hand, ocean tide models have been improving over the last decades and may serve as input to an independent ERP prediction model. The present thesis addresses the pivotal question of whether an empirical ocean tide model may be used for the development of an high-frequency ERP model and seeks to detect challenges and limitations thereof. As empirical ocean tide models do not include any hydrodynamic equations, the determination of consistent oceanic tidal currents is one major task addressed in this dissertation. An algorithm based on simplified momentum equations and continuity constraints is developed to derive barotropic volume transports and, therewith, motion terms of ocean tidal angular momentum (OTAM). The resultant OTAM values indicate the necessity for different weights of the continuity equation for different components. Equatorial components of diurnal tides require a large weight of 10000 compared to weights between 400 and 800 for the polar component. Semi-diurnal tides, on the other hand, show reasonable results for smaller weights between 100 and 200 for the equatorial components and between 400 to 2000 for polar components. If these variable weights are used, OTAM from estimated volume transports and assimilation models differ by 4.7-19.7% (mean over x-, y-, and z-component). A final model, based on the empirical ocean tide model EOT11a, supplemented by ten additional tidal constituents through quadratic admittance interpolation of angular momentum values, is derived and used as a priori model in the analysis of Very Long Baseline Interferometry (VLBI) observations. In a comparison with other types of high-frequency ERP models, (VLBI) data between 2011 and 2013 are analyzed using the Vienna VLBI Software. Post-fit ERP residual spectra and baseline length repeatabilities are calculated in order to validate four models. The results show that the empirical ERP model yields superior results than those based on ocean tide models. It improves the majority of baselines regarding their repeatabilities and gives generally smallest ERP residuals. The model based on empirical ocean tides is found to perform similarly accurate as the (IERS) conventional model with respect to baseline length repeatabilities. When using the empirical ERP model as reference, the newly derived model shows smaller baseline length variations for 40% of all 161 baselines, the conventional model improves 41%. For the comparison of post-fit residual spectra, the EOT11a-based model yields smaller amplitudes for more tidal frequencies than the conventional model and the same number as the empirical model. However, three main tidal constituents show a significant degradation with respect to post-fit residual spectra. Given that a similar behavior is visible for another ERP model based on ocean tides, imperfect hydrodynamic modeling or observational inaccuracies of altimetry are possible causes. These findings emphasize the need for further studies on high-frequency ERP predictions. An independent and consistently derived model, such as the one developed in the present work, may serve as new reference for diurnal and sub-diurnal polar motion and UT1 variations in the analysis of space geodetic techniques.