Precise ionospheric modeling is important for single-frequency Global Positioning System (GPS) users to achieve optimal positioning accuracy because the ionospheric signal delay is now the largest error source for positioning and navigation with GPS. The ionospheric modeling during ionospheric storms is particularly critical since the signal delay may be higher than normal and may differ significantly from the broadcast ionosphere model (currently employed by single-frequency users). In this study, a tomographic technique is used to model the ionosphere over North America using data collected from a network of dual-frequency GPS receivers. In support of real-time applications of the ionosphere model, short-term (5-min) forecasts of ionospheric total electron content (TEC) are also performed. To validate the accuracy of the forecast ionospheric TEC, a comparison of the TEC predictions with the observed TEC data (which are inferred from dual-frequency GPS observations) is carried out. Analyses are conducted using GPS data recorded during a 2003 geomagnetic storm event (29--31 October). Results indicate that under less disturbed conditions, an average accuracy of 5 ~ 6.5 total electron content units (TECU, 1 TECU = 1016 el m-2) can be obtained for the vertical TEC prediction and that 80% of slant TEC can be recovered by the model predictions. During extreme ionospheric storm periods (Kp = 9), the vertical TEC forecasting accuracy has a degradation of 2 ~ 3 TECU from the 3-day mean value, and the relative error is several percent to 10% larger than the 3-day average level.