Satellite Based Augmentation Systems (SBAS) broadcasts to users Grid Ionospheric Vertical Delays (GIVDs) and associated confidence bounds called Grid Ionospheric Vertical Errors (GIVEs) at Ionospheric Grid Points (IGPs). The correlation structure of ionosphere and undersampled threat should be taken into account in the generation of GIVE to meet the integrity requirement. In this contribution, the correlation structure of ionosphere over China was investigated and a primary undersampled threat model was built for the development of BDSBAS. Planar fit algorithm was used to estimate the GIVDs, the correlation structure was yielded from the differences and residuals of the vertical delays at Ionospheric Pierce Points (IPPs). Then undersampled threat model was created with data deprivation method. Five days’ data collected by 24 stations spread over China was selected to perform the correlation analysis and develop the threat model. Based on the correlation analysis, a maximum fit radius of 1500 km and a decorrelation value of 0.5 m were recommended for China area. Due to the ionospheric irregularity which could affect the region under the latitude of 30°N, a fit radius of 1300 km was suggested to improve the system availability. With the spatial threat model, it was found high latitude region owned a smaller threat value due to the relative quiet ionospheric condition. Relative Centroid Metric (RCM) less than 0.5 was proposed for determining the locations of BDSBAS’ sites. Threat models derived with different cut-off elevation angle were investigated and 5° elevation mask was proposed to generate the threat model.