Accurate localization of unmanned aerial vehicles (UAVs) in GPS-denied environments is crucial for critical applications like search and rescue, surveying, and remote sensing. Traditional visual methods for UAV navigation, although popular, face limitations such as complexity and high data processing demands, particularly in environments where visual cues are obstructed. This paper explores non-visual navigation alternatives by evaluating the performance of terrain-based methods within Digital Elevation Models (DEMs). Specifically, we compare the Basic Terrain Contour Matching (TERCOM) method with advanced Sequential Iterative Terrain Aided Navigation (SITAN) techniques that utilize the Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF), and Recursive Least Squares (RLS). The results indicate that the UKF significantly outperforms both the EKF and RLS. Additionally, we propose a methodology that integrates the UKF with a k-dimensional (KD) Tree for optimized search space reduction, thereby enhancing localization accuracy and computational efficiency. This integrated approach achieves an 84.6% reduction in Root Mean Square Error (RMSE) compared to the traditional TERCOM method. This considerable improvement highlights the potential of these non-visual methods for real-time UAV localization in challenging GPS-denied environments like urban canyons and dense forests.