Colorado geoid modeling at the US National Geodetic Survey

The geoid computation method currently used at the US National Geodetic Survey is based on the analytical downward continuation solution of the Molodensky geodetic boundary value problem. The quasigeoid is computed first in 1 ' x 1 ' grids for the whole area, and then, the geoid is obtained by adding the geoid-quasigeoid separation term based on the simple Bouguer anomalies. A few variations in data combination and topographic/gravity reductions are tested, and three models are selected for the Colorado experiment. Historical GPS/leveling data are used as a guide in the computations. The standard deviation of geoid differences between the historical GPS/leveling data and geoid models is 5.2, 5.0 and 4.9 cm for the models of the initial run, the first and second iteration, respectively. After the models were submitted to the Colorado experiment, they were validated using the GSVS17 (Geoid Slope Validation Survey 2017) GPS/leveling data. The standard deviation of the geoid differences on 223 GSVS17 marks is 3.2, 2.6 and 2.3 cm for the three runs, respectively. In addition to the (quasi)geoid models, the geopotential numbers at 223 GSVS17 marks are also computed and submitted to the Colorado geoid computation experiment and it is evaluated in Sanchez et al. (2020 this issue).