We investigate the influence of using IGS predicted orbits for near real-time zenith tropospheric delay determination from GPS and implement a new processing strategy that allows the use of predicted orbits with minimal degradation of the ZTD estimates. Our strategy is based on the estimation of the three Keplerian parameters that represent the main error sources in predicted orbits (semi-major axis, inclination, and argument of perigee), and their associated variance directly from the real-time GPS data set. A comparison with the use of IGS final orbits shows a negligible bias and an rms less than 6 mm in the ZTD estimates at all the stations. The improvement compared to a strategy based only on the quality index provided with the predicted orbits is 20%. The level of accuracy and turn-around time shown here meet the current requirements for operational meteorology. ¿ 2000 American Geophysical Union