Lab 3: Seismic Noise Attenuation

Objective:

To improve as much as possible the signal-to-noise ratio (SNR) due to corrupted seismic data with noise or unwanted energy arising from different kinds of sources

Theory:

Noises and unwanted energy that corrupt seismic data can be classified into two main categories:

Coherent noises
Random noise

Coherent noise is an undesirable energy that is added to the primary signals and is generated by the source. Such energy shows consistent phase from trace to trace. Removal of coherent noise can improve the processing techniques and simplifies interpretation of geological structures.

Random noise Disturbances in seismic data which lack phase coherency between adjacent traces are considered to be random noise. Random noise is not related to the source that generates the signal. The attenuation of random noise can be done using different methods including frequency filtering, deconvolution and stacking.

Practical

For analysing the seismic data in the spectrum domain different means exist:

The frequency (or wavenumber) content of one-dimensional (1D) time (or space) signals, Two-dimensional (2D) like the frequency-space (f−x) or the frequency-wavenumber (f−k) spectra; these can be used to obtain meaningful interpretation and helps to apply a suitable filtering technique.by using the following fx.m and fk.m can obtain the f−x and f−k magnitude spectra in dB.

Computer Assignment

1.Load the QC’ed real seismic data called SeismicData_gain.mat. Extract and display shot gather 2 using the wiggle plotting with a scale of your choice. Using the MATLAB functions fft and fftshift, plot the magnitude and phase spectra for traces 1, 10, and 15 of this shot gather. Write down your own observations.

2.For the same shot gather, plot the magnitude and phase spectra of its f−x and f−k data representation. Use both linear and dB magnitudes. Comment on your results. In your opinion, what are the main differences in terms of interpretation between both the f−x and the f–k?

3.Use the MATLAB’s signal processing toolbox to design and apply the following filters:

b. Repeat (a) but with an Infinite Impulse Response (IIR) filter with an order of 5. Design the IIR filter using the bilinear transformation method.

a. An FIR Low-pass filter (LPF) using any windowing method with a filter order of 50 and a cut-off of 15 Hz. Select any trace and apply your designed filter to it and display both the original trace and its LPF version. Comment on both traces by considering their time and frequency domain representations.

c. Apply the FIR and IIR LPFs from parts (a) and (b), respectively, on each trace of shot gather 2. Display the original shot gather and the processed shot gathers. What can you notice from your results?

4.Note: here and in the next parts of this assignment, make sure that you use the same wiggle display scale of your filtered shot gathers in order4. to run fair comparisons.

5.Repeat Part 2 but with a high-pass filter (HPF) with a cut-off 60 Hz. In your opinion, which is better the LP filtered or HP filtered data results?

6.Apply band-pass filtering with cut-off values of 15 and 60 Hz using the bpf_fir.m m-file (provided with the manual) to all the shot gathers and save the resultant data set (with its header) as SeismicData_gain_bpf.mat to be used in the coming chapters

clear all; close all;
load SeismicData_gain.mat;
N=100;
cut_off =[15 ,60];
dt = H(1).dt/1000/1000;
[Dbpf ,Dbpf_f ]= bpf_fir(D ,dt ,N,cut_off);
%Then save the Dbpf and H for the new data for lab 05.
save SeismicData_gain_bpf.mat Dbpf H

Shot gathers before noise filtering

Shot gathers after noise filtering

by applying15 Hz to 60 Hz bandpass filter we can observe that The gorund roll noise is attenuated and primaries are not damaged.the bandpass filtering doenst introduce aliasing due to appropriate parameterization.

f-x (left) and f-k (right) ,shot record 11 to 14,before noise filtering

f-x (left) and f-k (right) ,shot record 11 to 14 ,after noise filtering