GB2274951A

GB2274951A - A method of noise reduction applicable to continuous-field and other data acquired along lines

Info

Publication number: GB2274951A
Application number: GB9401287A
Authority: GB
Inventors: Richard John Gleave; John Edward Graham Savage
Original assignee: ARK GEOPHYSICS Ltd
Current assignee: ARK GEOPHYSICS Ltd
Priority date: 1993-01-23
Filing date: 1994-01-24
Publication date: 1994-08-10
Also published as: GB9401287D0; GB9301336D0

Abstract

A method of data surveying for use, for example, in marine gravity surveying wherein lines of gravity readings are corrected for line-to-line instability, and thereby their noise content reduced. Initial lines of data are treated in order to produce first and second sets of lines of data having differing frequency characteristics in respect of noise content, and these sets of lines are combined. The gravity data is acquired at high density. The above steps may be followed by median filtering and higher accuracy and resolution is obtained than is usually considered possible for marine gravity surveying. 2 - D grid filtering is used. <IMAGE>

Description

A METHOD OF NOISE REDUCTION APPLICABLE TO CONTINUOUS-FIELD AND OTHER DATA ACQUIRED ALONG LINES The present invention relates to the reduction of undesirable or low frequency noise superimposed on data acquired along lines, which need not be straight, and where the noise is uncorrelated between nearby lines.

The invention is applicable to but not necessarily confined to the processing of geophysical data e.g.

gravity data, magnetic data, bathymetry data and similar forms of continous-field data when acquired by a vessel such as a ship or aircraft, and where neighbouring lines are approximately parallel and closely spaced, and where data samples are acquired frequently along each line.

Usually the data will have been subjected to appropriate reduction and processing such as application of drift corrections, navigation corrections and the like. The data will also usually have been subjected to network adjustment whereby the difference in values obtained on different lines at the common location where they intersect are minimised by the application of suitably calculated corrections, which vary slowly along line, one set of corrections applying to each line.

We have identified a requirement for improvements in relation to data handling and data processing applicable to the handling and processing of geophysical data, wherein account is taken of the quality and/or integrity of the observed data as it is affected by the conditions under which it is acquired and/or environmental conditions, whereby, for example, correction can be made for line-to-line instability and uncorrelated noise reduced.

It is an object of the present invention to provide a method of geophysical data surveying and a method of noise reduction offering improvements in relation to one or more of the matters discussed above, or generally.

According to the present invention, there is provided a method of noise reduction and a method of geophysical data surveying, as defined in the accompanying claims.

In an embodiment there is provided a method of noise reduction in data acquired along lines comprising low-pass filtering of each initial line of data using a selected cut-off characteristic of each line of data, subtracting the results thereof from the respective initial lines of data to yield a first set of lines showing only high frequency noise and detail; converting said initial lines of data into a grid, filtering of the grid using a low-pass filter that matches the low-pass line filter, sampling back from said filtered grid onto the initial lines of data to yield a second set of lines of filtered data from which low frequency line to line noise has been cancelled; and combining said first and second sets of lines of data.

Also in an ebodiment, the method further includes following steps of further processing such as gridding of the obtained data and median and/or spatial frequency filtering before being displayed.

Also in an embodiment, there is provided a method of geophysical data surveying an area comprising taking geophysical data readings at closely spaced intervals along lines covering the area regularly. For example, between 300 and 1,000 readings per square kilometre may be obtained. The method comprises storing the data given by said readings, processing said data conventionally, and then further processing the corrected data to take account of the quality/integrity of the observed data from line-to-line as it is affected by the data acquisition conditions. In this way, for example, correction can be made for line-to-line instability and thereby uncorrelated noise can be reduced.

The step of further processing the data is preferably carried out by the method including the step of combining the first and second sets of lines of data.

An embodiment of the invention will now be more particularly described by way of example and with reference to the accompanying drawings, in which: Figure 1 illustrates the effect of the method of the invention on a portion of gravity data; and Figure 2 shows a schematic of the steps involved in the method.

Over an examplary North Sea hydrocarbon field gravity readings were taken with a LaCoste and Romberg air sea gravity meter along lines (the "in-lines") running approximately east-west and spaced on average about 100 metres apart. Gravity readings were taken every 25 metres along each line.

The method embodying the invention will hereinafter be described as "Grid Levelling".

Before the application of Grid Levelling, normal corrections applicable to marine gravity readings from a LaCoste and Romberg meter were made. These comprised cross-coupling correction, merging in of navigation data, lag correction and conversion from meter units to miliGals, editing out of poor data, drift correction, base tie, latitude correction, along line filtering with lkm cut-off lowpass filter, Eotvos correction, Bouguer correction and network adjustment. The network adjustment was made with respect to lines (the "cross-lines") which ran perpendicularly to the in-lines and which were spaced about 4 kilometres apart.

The data were gridded and displayed for quality control purposes. A portion of the grid is displayed in Figure 1, panel A. Here a monochrome shaded relief display has been used which enhances the appearance of subtle variations in the gravity field. The result shows prominent striations aligned with the acquisition in-lines. This corresponds to residual noise on the data, the noise being uncorrelated from line-to-line. The root-mean-square amplitude of the noise is estimated to be 0.7 miliGals.

This is in agreement with expection as the noise from the LaCoste and Romberg type of air/sea gravity meter operated under seastate conditions pertaining to this survey is typically 1.0 miliGals.

A conventional method for reducing the residual noise is to filter the grid using a low pass filter. Panel B of Figure 1 shows the effect of a 2 kilometre cutoff low-pass filter applied to the data of Panel A. The result is smoother, but the striations are still present. The residual noise, being white in the cross-line direction, still has a long wavelength component associated with it and any spatial frequency filtering is unable to reduce this component without also reducing the part of the signal which has the same wavelength range.

s To make better use of the random nature from line to line of the noise, two noise reduction processes were applied as described below.

The first of these is the Grid Levelling as defined above.

The scheme for applying this is illustrated as Figure 2.

The right hand part of the diagram, which shows how the data is taken through gridding (using a 75 metre x 75 metre pitch), low-pass filtering and sampling back to the original lines, illustrates the formation of the longer wavelength regional component of the output, designated "L-GREG". The filter chosen had a 10 kilometre cutoff wavelength with a Gaussian characteristic. The regional formed in this way is thus composed from data over an area of several kilometres radius, and this averages out the variations due to line to line noise by combining input from the many lines which fall within the circle of influence of the filter.

The left hand side of the diagram shows how the data is taken through low-pass filtering using an equivalent one dimensional filter to that used to filter the grid. The result is subtracted from the original line oriented data to leave a result which contains short wavelength signal and noise, and which is designated "L-RES".

The short wavelength component "L-RES" is added to the regional component "L-GREG". This is the Grid-Levelled result.

The Grid Levelled result, the regridded form of which is illustrated in Figure 1, panel C, still contains short wavelength noise (shorter than the cutoff wavelength selected for the Grid Levelling). After the Grid Levelling, the noise is estimated to be about 0.3 miliGals root mean square. This is a significant improvement, especially as it is the longer wavelength component of the noise which has been reduced.

After regridding, median filtering was applied. This process selects from the input grid samples falling within a rectangular window. Each side of the rectangle is an odd number of samples long. The window is moved over the input grid, and the output value in each case, which aligns with the centre of the rectangle, is the median value (the central value after sorting) of the selected input samples.

After some experimentation with various filter sizes, a single pass filter, 7 samples (525m) in the in-line direction by 11 samples (825m) in the cross-line direction was used.

The result after the application of both Grid Levelling and median filtering to the data of Figure 1, panel A is the result shown in panel D. Here subtle undulations of the gravity field of amplitudes down to 0.05 miliGals and with widths ranging down to 500 metres can be recognised.

These correspond to faulting at depths generally of about 3km depth and greater and which have been mapped using the three dimensional seismic reflection method.

By acquiring gravity data at high density, and by application of Grid Levelling and median filtering it has been possible to achieve much higher accuracy and resolution than is usually considered possible for marine gravity surveying.

Claims

C L A I M S:

1. A method of noise reduction in data acquired along lines, the method comprising low-pass filtering of each initial line of data using a selected cut-off characteristic of each line of data, subtracting the results thereof from the respective initial lines of data to yield a first set of lines showing only high frequency noise and detail; converting said initial lines of data into a grid, filtering the grid using a low-pass filter that matches the low-pass line filter, sampling back from said filtered grid onto the initial lines of data to yield a second set of lines of filtered data from which low frequency line-to-line noise has been cancelled; and combining said first and second sets of lines of data.

2. A method of noise reduction in data acquired along lines, the method comprising treatment of initial lines of data to produce first and second sets of lines of data having, respectively, differing characteristics in relation to high and low frequency noise content, and the method further comprising combining said first and second sets of lines of data.

3. A method according to claim 1 or claim 2, characterised by the step of further processing said combined first and second sets of lines of data.

4. A method according to claim 3, characterised by said further processing comprising gridding.

5. A method according to any one of claims 1 to 4, characterised by the further step of median filtering and/or spatial frequency filtering said combined first and second sets of lines of data.

6. A method of noise reduction in data acquired along lines, substantially as described herein with reference to the accompanying drawings.

7. A method of geophysical data surveying an area comprising taking geophysical data readings at closely spaced intervals along lines regularly covering an area to be surveyed, storing data produced by said readings, processing said data, and then further processing said processed data to take account of the quality and/or integrity of the observed data as it is affected by the data acquisition conditions.

8. A method according to claim 7, characterised by taking said readings at intervals corresponding to a high density of readings per square kilometre.

9. A method according to claim 8, characterised by said readings being taken at from 300 to 1,000 readings per square kilometre.

10. A method of data surveying an area characterised by taking data readings at spaced intervals and processing the data from said readings to take account of the quality and/or integrity of the observed data as it is affected by the data acquisition conditions.

11. A method according to claim 10, characterised by said processing of said data to correct to take account of the quality and/or integrity of the observed data being carried out by a method according to claim 2.

12. A method of data surveying substantially as described herein with reference to the accompanying drawings.