Time-lapse (4D) full-waveform inversion (FWI) of seismic data can help estimate the sub-surface changes due to hydrocarbon production and CO2 injection. Previously, we have developed a 4D FWI methodology for VTI (transversely isotropic with a vertical symmetry axis) media. However, the VTI algorithm fails to accurately reconstruct the 4D variations in the presence of dipping anisotropic layers that often cause a tilt of the symmetry axis. Here, we extend time-lapse FWI to 2D TI media with a tilted symmetry axis (TTI). The symmetry axis is assumed to be orthogonal to the reflectors, so its orientation can be es- timated from migrated depth images. The proposed algorithm is tested on the BP TTI model using three different time-lapse strategies. If the 4D data are repeatable, the param- eter changes can be reconstructed with sufficient accuracy even in the presence of moderate noise. We incorporate the “source-independent” FWI technique to mitigate the influence of errors in the estimated source wavelet and address the wavelet nonrepeatability in time- lapse data. In addition, we discuss the influence of several common nonrepeatability issues on the time-lapse inversion results. Testing on the BP model shows that the parallel- difference time-lapse method is more sensitive to nonrepeatability-related problems than the other employed 4D strategies.