The Radon transform is a mathematical tool that has been widely used in seismic data processing and image analysis. Three types of Radon transforms have been used as noise attenuation techniques in seismic data processing: the slant-stack or τ-p transform; the hyperbolic Radon transform; and the parabolic Radon transform. Ground-roll is a main type of strong noises in environmental geophysics. Attenuation of these types of noises is vital for improving the signal-to-noise ratio of seismic data. In the time–offset (t–x) domain, the ground-roll noise and other noisy waves just like; random noises, direct waves and airwaves overlap each other by the terms of time, which makes it difficult to attenuate coherent and incoherent noises in environmental geophysics. Nevertheless, significant different features shown in the time–slowness (τ–p) domain make it possible to separate coherent and incoherent noise and reflection waves effectively. A new method is proposed to separate these waves, using parabolic Radon transform (PRT). Amplitude and phase information is preserved during the proposed transformation. The reversibility and curvature of PRT provide a foundation for ground-roll noise suppression in the time-slowness domain. Synthetic datasets field and data examples have been implemented to illustrate coherent and incoherent noises attenuation can be achieved with a very little distortion of the effective signals. When compared with the linear radon ground-roll attenuation method, the field example results show the superiority of our method in suppressing the ground-roll noise and preserving the amplitude and phase information of effective waves.