We investigated a current numerical weather model, known as MAPS (Mesoscale Analysis and Prediction System), to determine if it could precisely define the behavior of GPS signals in the tropospere, ultimately leading to improved GPS-determined ellipsoidal heights. MAPS is the research version of the Rapid Update Cycle (RUC2) generated by NOAA's Forecast System Laboratory. MAPS is generated on an hourly basis and provides coverage in the contiguous United States at a 40-km grid spacing. We processed numerous subsets of GPS data collected over a months-long period on 23 static baselines ranging in length from 62 to 304 km. The GPS data were processed in 1/2-hr, 1-hr, 2-hr, and 4-hr session lengths. The primary effort was to compare the precision of heights obtained using a commonly adopted seasonal weather model with the precision of heights obtained using the MAPS weather model. Our analysis shows that the current version of MAPS can lead to improvement in GPS height precision when session lengths are shorter than two hours. For sessions longer than two hours, comparably precise heights may be obtained using a less accurate seasonal model by introducing appropriate nuisance parameters into the height estimation process. © 2001 John Wiley & Sons, Inc.