Geophysical data are always affected by different sources of anomaly. These sources are classified in to three groups: first group is relatively deep sources commonly called as regional effects, second group is near surface sources or local effects and third group is high frequency noises. The only common way to separate the anomaly is separation with respect to the signal frequency. There are several classical techniques in the literature such as polynomial fitting, Griffin method, moving average methods and frequency domain filtering methods.
In this paper a method based on discrete wavelet filtering has been applied. Discrete wavelet is calculated using MATLAB tool box based on rebio6.8 wavelet mother kernel. Discrete wavelet transform decomposes the signal into two parts low frequency (with approximation) and high frequency (whit details). The detail part can be also decomposed in to two or more parts based on the building block frequencies of the signal.
Sphere forward modeling is applied to test the algorithm of the separation method. Synthetic data is calculated assuming two spheres buried at two different depths. White noise with frequency equivalent to sampling interval is added to the synthetic data. The mentioned separation method shows appropriate result in comparison with other separating methods. One of the advantages of the method is automatic denoising process that can be applied during the procedure.
The method has also been applied for real dataset in a salt dome structure located at a station about 25 kilometers from the city of Ghom. The dataset is affected by two different geological sources: a deep fault structure represented as low frequency and a salt dome represented as high frequency in Bouguer gravity map of the region. Bouguer anomaly map of the region represents mixed effect of both structures. The separation process has been prepared comparatively successful. It has been compared to other separating methods. The results obtained from this comparison are:
1- The regional effect due to fault structure is clearly represented and can be applied separately for inversion process. Correspondingly the local effect is separated and presented in residual anomaly map and can be used in inversion modeling.
2- The high frequency noise effect is strongly attenuated during the process automatically.