Abstract: Based on observation data from a C-grade network in a city, the impact of four different baseline calculation strategies on the accuracy of geodetic height difference for short baselines is analyzed, which uses different raw observations and ionosphere-free combination. The result shows that it is difficult to evaluate the baseline processing accuracy of different calculation strategies through the analysis of repeated baseline differences. Therefore, external verification of baseline results was carried out through the combination of second-class leveling measurements and a quasi-geoid model. It is found that there is a systematic error of about 2.2 cm in the geodetic height difference in the results of using ionosphere-free combination, while the results of using raw dual-frequency measurements have the highest accuracy which is the best strategy for processing short baselines.