Over lake shores, altimetric waveforms are generally contaminated by lands, rough lake surfaces, and lag effects of the altimeter's automatic gain control. To improve altimeter ranging accuracy and in turn to get better surface height measurement, contaminated waveforms should be retracked against geophysical corrections. In this paper, an improved threshold retracker (ITR) is developed to retrack waveforms over lakes. ITR considers not only the physical characteristics of the reflecting surface, but also the stochastic feature of waveform, and two new retrackers, the N-Beta function model, and the N-5-Beta function model, are also put forward to develop the waveform retracking program of this study. TOPEX/POSEIDON waveforms over Hulun Lake in the North China are retracked to monitor the temporal lake level variations. A comparison with the in situ hydrological data indicates ITR is very efficient to monitor the lake level variations with the retracked altimetric data. The result of our study shows accurate seasonal level variations and the descending trend of Hulun Lake.