Abstract: Runway surface roughness significantly influences aircraft vibrations during takeoff and landing, affecting both flight safety and pavement durability. Aircraft operate at high speeds and wide gear spans, making them sensitive to long-wavelength (15–120 m) and lateral irregularities, which are often overlooked in traditional roughness models. This study aims to construct a three-dimensional runway roughness modeling framework integrating "precise detection-spectrum analysis-spatial reconstruction" in response to this issue. Combining the elevation data of 37 runways (5 asphalt runways and 32 cement runways) measured by a vehicle-mounted laser profilometer and the BeiDou positioning system, the power spectrum analysis was carried out by the Burg method and the spectrum models of asphalt and cement runways were fitted respectively. Meanwhile, a new exponential lateral coherence function was proposed. Finally, the three-dimensional spatial model was reconstructed by using the transfer function and genetic algorithm. The results show that the error of the measured elevation data is less than 1 cm. The spectral characteristics of different pavement types are significantly different. Among them, the R² of the asphalt runway fitted with the Sussman model is greater than 0.9. The cement runway needs to be characterized by a piecewise function to represent the spectral mutation. The fitting error of the new index's lateral coherence function has been reduced to 0.012. The reconstructed three-dimensional model is in good agreement with the theoretical value and the error does not exceed 0.18 mm² m/c. Finally, a three-dimensional model of 0–20 m in the lateral direction and 3000 m in the longitudinal direction is generated, providing support for aircraft vibration simulation and pavement maintenance.