Abstract: Four typicalmethods to describe the ionospheric biases at the roverpositionwere discussed to develop general formulas of correction generationmodels, both the advantages and disadvantages of each technique were summarized. Test data from both GPSNOS (GPS reference station network of Sichuan) and SCIGN (Southern California integrated GPS network) were used to evaluate the modeling techniques. The experimental results show that the linear interpolationmodel and the linear combinationmodel are more reliable network solution with a correction precision of 2 to 5 cm in reduction of the ionospheric biases for reference station networks with a baseline length of below 200 km.Real-time kinematic differential GPS positioning at cm-level can be considerably improved when the twomodels are used based on the virtual reference station technique to correct the systemic errors, including the ionospheric biases.