For the transformation of recorded seismic reflection data into a depth image a seismic velocity model is required. In this thesis, a new tomographic method for the determination of such velocity models is presented which makes use of traveltime information in the form of kinematic wavefield attributes. These attributes are the coefficients of second-order traveltime approximations and can be extracted from the seismic data by means of coherence analyses, e.g., by applying the common-reflection-surface (CRS) stack method. Compared to conventional reflection tomography which requires picking of reflection events in the prestack data, the use of kinematic wavefield attributes leads to considerable practical advantages: the attributes required for the tomographic inversion are taken from the CRS stack results at a number of pick locations in the stacked section. For each considered data point, these attributes can be interpreted in terms of the second-order traveltimes of an emerging wavefront due to a hypothetical point source in the subsurface. During the inversion process, a model is found that minimizes the misfit between these data and the corresponding quantities modeled by dynamic ray tracing. In the thesis, the complete theory of the method, as well as practical applications are presented. Starting with an overview of the required aspects of ray theory and the CRS stack method, the general concept of the new tomographic inversion approach is developed. The method is then discussed in detail for the case of 1D, 2D, and 3D tomographic inversion and the entire process of deriving a velocity model is demonstrated on a synthetic and on a real 2D seismic dataset.