Abstract: Fast and correct ambiguity resolution (AR) between long-range base stations is the precondition of realizing high-precision Network Real-time Kinematic (NRTK) positioning. For long baselines in GPS, GLONASS and BDS combined systems, it is always difficult to fix all the ambiguities rapidly and correctly due to the dimensions of ambiguities increased greatly with the influence of measurement noise, residual atmosphere delay and etc., which will be especially difficult for the ambiguities from low-elevation satellites. This paper proposed a fast partial ambiguities resolution (PAR) method with ambiguity subset adaptively selected based on the successively increased elevations for long baselines in GPS, GLONASS and BDS combined NRTK. Satellite cut-off elevation, ambiguity success rate and ratio are treated as major parameters to select the ambiguity subsets with ambiguity subset adaptively selected based on the successively increased elevations, and so that to realize fast and correct ambiguity resolution. The real measured long-baseline data which contain GPS, GLONASS and BDS observations are used in the experiments, and the results show that the PAR strategy can effectively avoid the influence of low-elevation satellites on AR fixing, which increase ambiguity success rate and Ratio significantly and of course shorten the required time of correct AR for long baselines in NRTK.