Propagation tools are the cornerstones of the design and planning of communications networks. The aim for greater accuracy and faster computation are two of the conflicting demands placed on propagation models. Usually models characterised by a high level of accuracy are inherently slow and vice-versa. So very often a compromise between these two parameters has to be made by the network planner. Given such a situation an investigation into the speed-v/s-accuracy performance of models is a worthwhile endeavour. In this thesis, we have identified two reference models, the Narrow Angle Parabolic Equation (NAPE) and the slope Uniform Theory of Diffraction (STD-UTD), from the full-wave and canonical classes of propagation models. A description of the underlying principles of these reference models is made. We have sought to improve the speed-v/s-accuracy performance of the reference models by an appropriate selection of input data. The aim is to improve the speed of the NAPE model and the accuracy of the STD-UTD model by focusing on the obstacles of significance in terms of diffraction. The modified models thus obtained are the NAPE-free space (NAPE-FS) and Selected Edges slope UTD (SE-UTD). The input data selection procedure of the modified models is addressed. New hit rate metrics to assess the accuracy of the models have been developed. The models have been run in a rural (terrain-only) environment and a speed-v/s-accuracy analysis has been carried out. It has been found that the NAPE-FS brings a higher saving in runtime for profiles with significant height variations while maintaining reasonable accuracy. The SE-UTD model has also an improved performance for such profiles. In addition, predictions have been made using UTD-based models in an urban environment. The analysis shows that the SE-UTD is not suited to urban environments with significantly uneven building height distributions.