Service: Grid stitching joins separate datasets without creating artifacts

Grid stitching and merging service post image

OUR technique of rapidly and automatedly stitching and merging large gridded datasets permits the almost seamless joining of grid boundaries, even those that, with other techniques, are proving problematic. A case study shows how.

Having one continuous compilation of geophysical data over an exploration area of interest is much more practical than having that data divided across many separate chunks.

One example of the elegance of a single, stitched mosaic map is Geoscience Australia's magnetic anomaly map of Australia, the latest edition of which was released a few months ago. [1] The map represents the merging of 795 separate grids, according to GA. [2]

The grid-stitching methods currently available commercially typically employ polynomial corrections.

An arguable drawback to using polynomials to join grids is the rebound effect. If one part of a grid is forced downward to match part of another grid, then there'll be a popping up elsewhere on the first grid. The effect is similar to pushing down on one area of a stiff but flexible plastic slab, such as a plastic kitchen cutting-board — there'll be an upward movement propagating along another part of the slab. It means unwanted regional trends are introduced into the merged grid.

Difference grids

Rather than using polynomials, Fathom Geophysics' grid-stitching method uses difference grids.

Our fine-tunable technique allows us to locally reshape the boundary of a lower-resolution (or a less certain) grid to precisely conform to a higher-resolution (or a more certain) grid.

If it's warranted, we can make both grids' boundaries move. The boundaries could share equally in movement required to make the join, or that movement could be shared unequally.

When difference grids are used, the grid boundaries you elect to reshape behave more like malleable clay than a flexible slab unit. It means the boundary's movement is localized, with none of the undesirable rebound effect occurring.

The method leaves high-frequency data largely intact, unlike most alternative methods. It does tend to alter low-frequency (long-wavelength) features, but this is not a significant issue for explorers wanting to examine small-scale features in the data.

When grids don't overlap

Note that if the two initial grids only meet side-by-side instead of overlapping, then our typical procedure is to extend the grids by extrapolating out from the boundaries to be joined, before undertaking a difference grid.

We extrapolate values using an in-house grid-filling algorithm. (See more details about our grid-filling and extension service.)

And if there is a gap between the two initial grids, we use grid-filling to interpolate values between the data. Our rule of thumb is that the gap should be no wider than a few pixels.

Whenever any extrapolation or interpolation is done as part of grid-stitching, we can supply a polygon layer that identifies the grid values generated by grid-filling, as distinct from the original grid data.

Performs well under pressure

We've tended to obtain good results when using our grid-stitching technique in situations where a satisfactory join has been difficult or impossible to achieve with another stitching technique.

Some might argue that if two grids don't merge nicely in a straightfoward way upon the first few attempts, then perhaps they shouldn't be merged. We agree with this conservative rule of thumb. The risk of falling prey to the rubbish-in-rubbish-out phenomenon is ever-present, especially when intensive, high-volume calculations are involved. One needs to remain vigilant.

However, sometimes explorers are forced to take on board all or most of the exploration data they can lay their hands on, including data acquired in surveys quite some time ago. This is typically true in frontier greenfields exploration areas where geological mapping is relatively scarce. This is also true of places where the cost of a repeat survey may be prohibitive — such as in remote areas, or difficult terrains, or harsh climates.

As always, whenever older or more uncertain datasets are retained rather than binned, the information they contain should be taken with a grain of salt. Ideally, metadata should be made available to define which datapoints in the merged mosaic are likely to be less reliable and why.

Case study: West Kenya

We found our grid-stitching method worked well on Aviva Corporation Limited datasets.** The grids contain airborne magnetic data for part of Aviva's West Kenya joint-venture project area. The data values represent total magnetic intensity (TMI).

An overview image of the case study area shows the difference between the collection of joins attained with our method and with an alternative method (see Figure 1, top left and top right).

Note that the images obtained via the alternative method have had a Butterworth smoothing filter applied across them, to subdue the large suture-site discontinuities as much as possible. We did not apply a Butterworth filter to the images obtained via our method, so that readers can identify our subtle suture sites. We have the capability to remove our suturing traces by applying local, oriented Butterworth filters. The ability to do this style of Butterworth filtering — in a direction-specific manner over small areas of an image — may be currently unique to Fathom Geophysics.

One segment of the joined boundaries runs in a roughly north-south direction in the west. Here, we experienced difficulties when trying to obtain a satisfactory join using commercially available software. The result we finally achieved there, despite our best efforts, still included a prominent stepwise fault-like artifact in the magnetic signature at the suture site. No such artificial fault was introduced at all when using our stitching method (see Figure 1, middle left and middle right).

Once grids have been stitched together, usually the next step in geophysical data processing is reduction to the pole (RTP). The procedure recalculates data so that it appears to have been measured at the earth's magnetic pole. (For a more detailed explanation of RTPing, see the previous ezine article we released on our differential RTP capabilities.)

The case study area, being located essentially on the earth's geographic equator, has a low magnetic inclination. When low-inclination data are RTPed, magnetic lows in the raw TMI data are transformed into magnetic highs.

It's after the RTP process has been carried out that a poor grid stitch can have a significant effect on magnetic data interpretation.

For example, in our case study area, the joined boundary segment running roughly east-to-west behaves very differently upon RTPing according to the grid-stitching method used.

When an alternative stitching method is used, the whole east-west suture area becomes a large, long-wavelength high (see Figure 1, bottom right). The (incorrect) implication of this high is that a large magnetic body exists beneath the surface. No such spurious high was introduced at all when using our stitching method (see Figure 1, bottom left).

Minimizing the number and size of artefacts introduced via grid-stitching is the logical goal to strive for and may help in avoiding wild goose chases during the exploration targeting phase.

Note that grid-stitching can be used to join other types of potential field datasets (such as gravity data). The method also applies to remotely sensed datasets (such as radiometric data and ASTER data) and geographic datasets (such as topography and bathymetry data).

** Acknowledgement: Fathom Geophysics thanks Glen Edwards, Aviva Corporation Ltd, for permission to discuss the company's West Kenya project and to use images of the project in this case study.


[1] Geoscience Australia (2010) Magnetic Anomaly Map of Australia, 5th edition, 1:5 million, Government of Australia, (redirects to the Geophysical Archive Data Delivery System webpage).

[2] P. Milligan (September 2010) "New magnetic datasets to identify energy, geothermal and mineral resources", AusGeo News, 99,

Data stitching and merging service figure 1FIGURE 1: TOP LEFT: Results of the Fathom Geophysics grid stitching. TOP RIGHT: Results of an alternative method of grid stitching. MIDDLE LEFT: Detailed window showing results of the Fathom Geophysics grid stitching. MIDDLE RIGHT: Detailed window showing results of the alternative method of grid stitching. BOTTOM LEFT: Results of reduction to the pole using the Fathom Geophysics method of grid stitching. BOTTOM RIGHT: Results of reduction to the pole using the alternative method of grid stitching.