The performances of Ground-Based Augmentation System (GBAS) designed for landing phase of aircraft rely on the accurate characterization of error models. Among various error sources, the multipath model, which is typically constructed by combining environmental errors at airports, must be modeled in GBAS. However, in practice, the multipath effects at a particular airport differ from other airports due to distinct construction sites and continually changing environments, resulting in inaccurate error model in GBAS operations. Therefore, in this work, we develop and evaluate a two-dimensional ground facility error model from the Global Navigation Satellite System (GNSS) stations at the Suvarnabhumi International Airport in Bangkok, Thailand. The results indicate that the elevation and azimuth grid points require around 7 days of observation data to create the GBAS ground facility error model for GBAS operation. The number of observations per day at each elevation and azimuth grid point will determine the data requirements for the complete building of the two-dimensional ground error model. When the proposed model is applied to the GBAS simulation, it is found that the proposed two-dimensional ground error model reduces the root-mean-square deviation (RMSD) of positioning errors by around 0.4 percent to 3.5 percent when compared to the onedimensional error model and the category B Ground accuracy designator (GAD-B) model, respectively. The maximum vertical protection level (VPL) reduction of the proposed two-dimensional B-value model in comparison with the reference one-dimensional B-value is 0.24 meters, about a 6 percent reduction. I.