Abstract:The present ground roll elimination methodshave some negative effects in the Fourier trans-form,and that the threshold on wavelet transformis not precisely determined.So ground roll at hy-perbolic top and nearly areas cannot be eliminated.For this propose,we present in the paper a fittingmethod based on mountain climbing to eliminateground roll in wavelet domain. Using the waveletcoefficient characteristics from the wavelet trans-form,this approach searches similar waveform onadjacent traces and calculates the mean.The meanis subtracted from the original wavelet coefficientsif it meets the three ground roll identification crite-rions.Uround roll is removed by wavelet reconstruc-tion. Model and real data examples show that this ap-proach eliminates ground rolls without the shortcom-ings of the existing ground roll elimination methods.