<p>In this contribution, we use all three components of the gravity vector observations to compute a regional geoid and demonstrate the advantages of using the horizontal components alongside the vertical component. We apply the one-step integration method within the remove-compute-restore framework; where the long-wavelength part of the geoid is recovered from Earth’s gravitational models while the harmonicity of the computational space is ensured by removing the topographic effects. We create a system of linear equations using a discretized form of the one-step integration method and use the Tikhonov technique to deal with the numerical instability due to its implicit downward continuation and to determine the geoid at higher resolution, e.g., 1 ′ × 1′. We propose a novel method to estimate the Tikhonov regularization parameter using the discrepancy principal and a stable solution of the geoid at lower resolution, e.g., 3′ × 3′. The results reported are based on real airborne gravity vector observations collected over Colorado, USA. The scattered observations at flight level are directly inverted to the disturbing potential at grid points on the reference ellipsoid, where geoid heights are then computed using Bruns formula. We evaluate the external accuracy of the geoid by comparing it with GNSS/levelling data and estimate the location-based internal uncertainties (error) of the geoid heights through formal error propagation. As part of this contribution, the airborne gravity vector data used in this study are also available for research purposes upon request to the corresponding author.</p>