Since their advent in the early 20th century, lightning detection and location systems (LLSs) have become a substantial component of weather observation and undergone a remarkable evolution both in their capabilities and their performance. Nowadays, ground-based low-frequency lightning location systems are able to locate the strike point of cloud-to-ground lightning discharges with a median location error that is in the order of 100 - 150 m, and estimate the peak currents of lightning return strokes with remarkable statistical accuracy. The underlying assumption forming the basis for the operating principles of LLSs and the inference of lightning parameters is that of a flat-ground scenario. This thesis investigates several aspects of the impact of irregular terrain on the radiation and propagation of lightning electromagnetic (EM) fields. Employing the three-dimensional finite-difference time-domain (FDTD) technique using open-source FDTD solvers, various simulations are performed to examine different wave propagation scenarios, which are as follows: a. the impact of a lightning strike point located on a mountain on the radiated fields, b. the influence of irregular terrain along the propagation paths on field attenuation, and, c. the impact of the local terrain in the immediate vicinity of the LLS sensors on the incident fields. The deviations from an ideal flat-ground scenario are evaluated and the resulting estimation errors, also known as sensor site errors are analyzed. Further, the influence of buried LLS sensor power supply cables, or any kind of buried conductors, on the fields that are measured by magnetic direction finders, and the related site errors, are investigated by employing a field-to-cable coupling model. Whenever experimental data are available, they are compared with the simulation results. The evaluations indicate that the simulation results are in general in good agreement with the real-world measurements. Although only indirectly related to the preceding topics, the final part of the thesis focuses on flashes with long continuing currents (LCC), which often follow a return stroke and can last for several hundred milliseconds. Using remote single-station E-field measurements, the transferred electric charge is estimated for 140 long CC events assuming a simple monopole charge configuration. Statistical models are then utilized to investigate the expected accuracy of the estimated charges. Overall, the results of the thesis demonstrate the effectiveness of the proposed methodologies and advanced computational models, and indicate their potential to contribute to future improvements in lightning location systems.