In underwater acoustics, shallow water is a complex and dispersive medium. For low frequencies, the propagation is described by normal mode theory. Modes are nonlinear structures overlapped in time, in frequency, and in time-frequency domains. Advanced signal processing methods are needed to study them. This paper presents two different warping methods allowing modal separation and estimation of parameters of the modes. As these transformations are invertible, they also allow modal filtering. Thus, they are a good preprocessing tool for source localization or tomography. This is shown on simulations and real data. Warping is a transformation of a propagated signal and is based on an environment model. Here the environment is modeled with isovelocity or Pekeris waveguide. However, both methods are quite robust to mismatches with the real environment and can be used with minimal preliminary knowledge of the real environment. The first warping method is applied on the time domain. It transforms each mode into a sinus function of frequency fm (with fm the cutoff frequency of the mode m). The second is applied on the frequency domain. It transforms a chosen mode into a dirac, which is localized on the arrival time of the mode.