ArchaeoPhysica Ltd on the WWW

Data Presentation (Geophysics)

Image Plots

The examples below represent the shallow component of the total magnetic intensity & have been chosen to illustrate basic concepts pertaining to image plots. Other sorts of plots are considered below & could be applied to the same data. The principles discussed here apply to all these types of plot regardless of the data they image.

Often, the most effective image is formed from a simple monochrome palette, i.e., the greyscale. It is common to plot magnetic data with darkest tones representing the most positive values, however, some surveyors regularly plot the other way round. Whichever is more correct depends upon the dynamics of the data but historically archaeologists have favoured dark high because the sorts of feature producing these values often looks dark under excavation. Electrical resistance data is often plotted white high for the same reason. The first two images below depict the difference due to palette reversal.

Normal, dark tones high field	Normal, light tones high field	Equal contrast, light tones high field

The mechanics of constructing a greyscale image are fairly simple but a few general principles need consideration. It is important that the plot represents the dynamic range of the data so there should be a wide distribution of mid tones as well as clear extremes. If a linear palette is simply stretched between two limits these must not be set too close to the centre or background values. +/- 1 standard deviation (sd) is often too close, +/- 2 sd is usually more appropriate, or at least 1.5 x interquartile range for non-gaussian data. The red & blue plots show increasing distance from the median background as increasing intensity of colour. Intrinsic to the use of greyscale images is the assumption that mid grey represents the background amplitude. This is only the case if the clip is set symetrically about a representative measure of central tendency, e.g., the mean or median of the data set. For some data a particular value will equate to background & should be used to centre the palette. For magnetic gradient data this is often assumed to be zero but it is often slightly positive which affects the imaging of low amplitude data. It is not always easily identifiable for electrical resistance or magnetic susceptibility data either. For most data there will be a band of values within which most of the detail exists with extreme values outside this. These may be numerous but in themselves do not provide any more information than those within the central band. Simply setting the palette (clipping) to lay within the central band of detail often results in an image dominated by high-contrast extremes. To acheive a better image a non-linear palette should be used, e.g., the arctangent compression below left. This allocates most of the grey tones to the central band & the rest smoothly distributed across the extremes. An alternative is histogram equalisation where grey tones are allocated so as to equalise the number of data within each. This has the advantage of decreasing the allocation of grey tones to the extreme values & maximising them where there is more data. This does not work for every set of data as it assumes that the distribution of the data can be described as an approximation to a bell curve & where this is not the case distortion of the image can result. Finally, there is a more subtle effect at work which relates to how the human eye perceives intensity & luminance. A grey scale divided by amplitude into equal-sized divisions will have approximately half tending to white & half tending to black. The eye, however, responds better to changes in the scale tending to black than that tending to white as differences between pale tones are less visible. The result is less visual mapping between data amplitude & grey tone in the lighter parts of the image. To counteract this it is possible to define a palette with divisions at logarithmic intervals to redistribute the darker tones throughout the image & reduce the affect of the lighter ones. The image top right shows the basic effect; it is slightly clearer than that to the left which has a linear palette over the same range of data.
Arctangent compressed, dark tones high field	Simple trace plot	Trace plot with position bars coloured by polarity

Other types of plot
Geophysical data can be plotted in other formats, e.g. relief, as shown below left, wireframes or wiggle traces as shown above & as contours, below right. For relief plots the issues surrounding palettes again apply although it is in some ways less critical. Colour can also be added to image or relief plots but care must be taken to ensure this is meaningful. Mono- or bi-colour plots tend to convey more meaning than multicoloured ones. The eye responds both to brightness & hue & their combination has a significant affect on how the data is perceived. The differences between colours in a palette may not convey the correct distinction between the classes of data they represent. For this reason, the best software allows colour models like YIQ to be used where palettes can be created with equal brightness but different hues, etc. Contour plots are best suited to data that varies smoothly across a survey e.g., topographic, electrical resistance & magnetic susceptibility data. They can be open with just contour lines or filled with different colours or shading between the contours. If using fills the discussion about grey & colour palettes again applies. Filled contour plots can sometimes be used to simplify magnetic data so that broad trends become more apparent as shown below right. Dot density plots were developed in the 1980s as a means of displaying 2D data on dot-matrix printers. Modern computing power means that the better image & relief plots are easily created & dot density has little direct application. To some extent the same applies for wireframe plots as although they can be invaluable for manual analysis of the data they have been replaced by relief plots. These still allow the data to be visualised full range but the 3D character of the data is more apparent through the use of light sources & colour shading. Trace plots were again a feature of early survey & originated as plots formed in real time from the outputs of magnetometers during survey. They retain considerable function & can now be easily produced as vector images. A key advantage is that they allow the full dynamic range to be displayed but as a plan view, permitting the eye to easily detect spatial patterns. A further advantage is that their vector format allows them to be easily imported into CAD & GIS environments.
3D Relief plot		Contour plot