Within the scope of pavement monitoring, the use of Ground Penetrating Radar (GPR) methods, especially in the detection of emerging cracks is of greater importance in order to predict the sustaining life period of a pavement structure. However, suitable processing techniques can be used to improve the operator’s interpretation of the pavement conditions in a more efficient way. This matter has motivated the development of specific signal processing methods to conduct early detection of defects such as thin interlayer debondings. The robustness of such signal processing methods depends on their ability to adapt to variations in the GPR data sizes and configurations (e.g., pavement layer geometry and permittivity). The lack of labeled and controlled field data however, demands the need for numerically modeled GPR data to assess the robustness of these methods. Therefore, the numerical analysis is carried out on the EM wave propagation in the pavement data created using the FDTD-based 2D GprMax software. This data can be used to evaluate the influence of pavement characteristics on SVM’s performance. The debonding detection efficiency is susceptible to the pavement layer characteristics, especially the top layer and the debonding layer. Thus, the objective of this paper is to study the influence of top pavement layer characteristics on the debonding detection rate by implementing on independent 1D GPR data (namely, A-scans) using time domain signal features. The use of local features provides the required refined signal analysis and reduces the computational burden. The detection efficiency over larger data sets indicated SVM’s adaptability. On the other hand, the performance over various top pavement layer configurations demonstrated its robustness.