Tropospheric delay is one of the main factors that affect obtaining a higher precision of long-distance RTK. In order to optimize the tropospheric delay modeling method and obtain a higher positioning precision, this paper uses the measured zenith tropospheric delay data of BJFS and URUM station to calculate the tropospheric zenith delay and RMS of four classical tropospheric delay correction models, the precision and applicability of classical model are analyzed, Hopfield model and Saastamoinen model have a higher precision. But for high precision baseline solution, the precision can not meet the requirements. Therefore, the GPS measurement method based on multi-base station GPS network is adopted. In order to analyze the factors that should be considered in the model, the tropospheric delay of several CORS stations calculated by a classical model is compared and analyzed. It is concluded that there is a positive correlation between the tropospheric delay and elevation, which provides evidence for modeling. And the applicability of several regional modeling methods is theoretically analyzed. The best method to choose base stations of multi-base station modeling is analyzed by several experiments and it is concluded that when the base station is laid in the mountainous area with gentle troposphere change, it should take into account the uniformity of the plane position distribution of the station under the condition of giving priority to the regional elevation range. By this way, the model can reflect regional characteristics better and obtain a higher degree of precision. And the precision and applicability of region modeling method of tropospheric delay is studied through three experimental networks. The tropospheric delay modeling methods based on multi-base station GPS observations obtained in this paper will obtain a higher degree of precision and a better applicability with regional range reducing. In different regions, the different method of tropospheric delay modeling should be chosen. In small-sized and medium-sized areas with gentle tropospheric change, the H1QM3 is the most suitable model for tropospheric delay calculation and the precision is better than 1 cm.
