Part 2: I only made this one longer because I didn't have time to make it shorter - "A Tufte act to follow"

In Part 1 of the series of articles, I argued the case that well-crafted graphics are key to conveying the complex information that relates to Mineral Resource estimation (MRE) work. In this article, I attempt to summarise the key items of good technical graphic design, with an example from an expert in that field of endeavor, then a first example of MRE work relating to the task of explaining the geometry and relationships of a complex geological model.

Tufte’s principles of graphic design

A technical graphic is a visual representation that should support an MRE practitioner’s ideas or arguments presented in the associated text. However, inadequately designed graphics can fail to provide the desired support or even detract from the intended message. To address this issue, Edward R. Tufte, a renowned statistician and Professor Emeritus at Yale University, has published many books since the 1980s that provide professionals with guidance on how to prepare technical graphics that have 'graphical integrity'. This concept means that graphics should accurately represent data and avoid any distortions or manipulations that could distract or mislead readers. Paraphrasing many of Tufte’s rubrics to be considered when preparing a technical graphic, MRE practitioners should:

• ‘Show the data’ that supports a thesis described in the associated text.
• Avoid dishonest graphical truncations or scale transforms, so show all the data not just the data that supports an opinion.
• Remove all distracting non-data ink decorations from the graphic to encourage readers to think about the graphic’s substance rather than its design.
• ?Present as many numbers as possible in a small space, so maximize the 'data ink' and its density.
• Make large datasets coherent and understandable to facilitate comparisons of different pieces of related data.
• Display nested levels of detail ranging from a fine structure to a broad overview.
• Have a clear purpose in terms of data description, exploration, and decoration.
• Integrate statistical or verbal descriptions directly into the graphic rather than relying on extensive legends or footnotes.

Many of Tufte’s guidelines can be achieved through:

• Comparing and contrasting the differences and similarities between dependent variables,
• Demonstrating the causality of independent variables (inputs) that affect dependent variables (outputs),
• Using multivariate combinations and small multiple images where needed to explain complex narratives,
• Integrating multiple modes of presentation in the same graphic, such as diagrams, text, maps, and so on, to reveal and explain the relationships between the data and findings,
• Document key acknowledgments, provide adequate graphic titles, and include table notes where necessary,
• Always include clear scales and measurements where spatial dimensions are present,
• Provide contextual before and after states, and use trend lines to point to future or between data results, and
• Use muted shades and color palettes that can be used to create a minimal perceptible difference between different graphical elements and ensure the rendering emphasis is on the data rather than its decorations.

In his book ‘Visual Explanations ’, Tufte demonstrates many of his principles for creating effective technical graphics using the two images below, which are based on the numerical simulation of a severe storm (have a look at this link to see the original video).

Both images both above are three-dimensional (3D) projections of the same storm image from the simulation, with the upper image being the basis of Tufte’s redesigned lower image. While this example may appear unrelated to MRE work, the presentation of 3D spatial information is common in many technical graphics presented in the field, making this example relevant.

Tufte’s redesign of the original image is depicted as the second image. His changes to the original image include content from other parts of the full animation, and the redesigned image demonstrates many of his graphical principles including:

• The use of muted shades for the grid and background brings the cloud into principal focus, with the rendering following the minimal perceptible difference principle, as demonstrated by the boundary between the upper storm edge and the background.
• In the original image, the storm cloud is truncated on the left, presumably being the limit of the simulation, while on redesign this technical anomaly almost coincides with the left border of the image and as such, is no longer a distraction. This cropping process has also increased the relative size of the cloud in the graphic.
• Tufte includes black orthogonal direction arrows next to and behind the storm to provide spatial context, which by gauging the long axis trend of the storm’s shadow appears to be moving towards the northeast.
• The tail of the easting arrow is labeled at 100 km to alert the reader to the longest easterly dimension of the storm. From this measure, the reader can then determine through square counting that the white-lined muted grey grid squares have a side length of 10km.
• The vertical axis is labeled at 16 km, which allows the reader to estimate that the storm’s height is about 14km. From this, the reader can conclude that there is a strong vertical exaggeration in this image, a fact that is not discernible in the original.
• In the lower part of the redrafted image, Tufte includes a series of ‘small multiple’ images from different times in the simulation, which depict the development of the cloud at the minute time stamps clearly labeled to the right of each image. These images reveal how the storm’s thunderhead develops and extends more at higher elevations with time. This strip of small multiples also crops the foreground of the main image, further increasing the focus on the main cloud and removing a large chunk of non-data ink.
• Immediately above the small multiples strip is a red bar line depicting the flow of time, which is labeled at both ends with a time stamp. The left-hand stamp labeled ‘5 minutes’, alerts the reader as to the units of time applied across the graphic. This subtle granular addition to the image reveals that the storm in the main part of the graphic is depicted 105 minutes into the simulation. Through visual comparison with the small multiples below, the reader can also discern that the storm’s final shape and size have largely reached the final geometry at 105 minutes, with only inflation of its volume occurring in the remaining 35 minutes of the total simulation.

Comparing Tufte’s redesigned image to the original and also his list of recommendations detailed above, the reader should find that many of Tufte’s guidelines have been applied including the use of:

• Minimal and effective direct labeling,
• Muted shades and minimal perceptible difference principles,
• Before and after states,
• Scales for 3D space and also time, including a scale that alerts to a strong vertical exaggeration.
• Multiple modes are applied in the same graphic.

[Side note: If this is the first time you've heard of Tufte's work I would encourage you to buy at least one of his books, which are available in softback form, or alternatively spend some of your employer's training budget on his online course , where you get all the books as part of the course fee. The only caution here is that you are expected to read the books - or at least browse the pictures]

With Tufte’s guidelines summarised and demonstrated by this example, the next article in this series will describe the author's attempts at applying Tufte’s principles to several Mineral Resource estimation (MRE) related graphics for MRE-related tasks.

Geological interpretation graphics

The first stages of MRE work involve collecting the base data and verifying it to build a reliable MRE database. The next essential stage is to prepare a 3D geological model from the verified data that will guide the estimation process. In the Nova-Bollinger Deposit , IGO's geologists have interpreted 22 distinct estimation zones representing six different styles of nickel-copper-cobalt bearing sulfide mineralization, including breccias, breccia-splays, net-texture, gabbros containing disseminated mineralization, footwall stringer zones, and footwall massive lenses. A low-grade "waste halo" zone is also modeled, which encompasses all the other zones and contains marginal disseminated mineralization. The main purpose of the Waste Halo is to facilitate dilution modeling in mine planning ORE studies.

Visualizing these many 3D zones in a single 2D image projected onto an MRE report page is problematic as the different estimation zones abut, intersect, and overlap with each other. As such, any 2D image captured from a 3D display in an MRE modeling software system, like the graphic below, will always mask some of the estimation zones that the practitioner should depict and describe in the MRE report.

This graphic was prepared several years ago and is an early attempt by the author to follow Tufte's principles. Tufte elements included in this image include the provision of context and scale through a muted background 3D grid, bar scale and north arrow, and the direct labeling of the main visible estimation zones. However, there are many non-data ink distractions in the image including:

• The strong emphasis and image dominance by the large and darkly rendered estimation zone labels. ?
• The many grid labels add nothing to the image. The bar scale included on the northern wall of the background grid adequately provides a sense of scale for the 250m-sided background grid squares.
• ?The color palette applied to the estimation zone is a bit too garish.
• There is no context or explanation for the grey mine development lines, which are mostly obscured by the estimation zones.?

To address (most) of these critiques, the figure below is the author’s recent redesign of the information depicted above.

In this redesigned figure, I've attempted to apply Tufte's small multiples strategy is used to reveal the geometries, spatial locations, and relationships of 21 of the 22 Nova-Bollinger estimation zones, with the all-encompassing Waste Halo zone, excluded. Now depending on the size of your computer screen the image may seem to be too small, but one of the advantages of digital graphics is that it is easy to zoom into features so interest, by holding the ctrl-key and rolling the mouse wheel (give it a try!). Otherwise, this image was ideally designed to occupy a full A4 page in the MRE report.

Now, in the image above, a sequence of small panel images, labeled a) through k), depicts all 21 zones relative to each other, along with a background grid and the mine’s heading development center lines as surveyed on 30 December 2022. Starting from the top left-hand panel a), arrowhead symbols indicate the order in which estimation zone locations, shapes, names, and integer codes for each zone are built up in a sequence from the deposit’s footwall to the hanging wall modeling surface represented by the clipped base of the Leucogabbro upper intrusion in panel k).

Note the direct labels are strategically placed in panel a) to identify the Nova and Bollinger areas, the location of the mine's decline, the dimensions of the faint background grid squares, and the direction of the north with respect to the consistent view direction presented in all 10 panels. The date of the mine development center lines is annotated in panel b), and the Diamond Drilling Ring (DDR) drive, which was the platform used to drill and define the Nova-Bollinger MRE during mine ramp-up, is identified in panel i).

The two images in the third row of the full explanatory image above, labeled as panels k) and l), are included to reveal the overall curviplanar structure of what is effectively a sulfide zone MRE model.?Additional direct labeling is included in panel m) to identify the local Upper and Mid zones in the mine's nomenclature of its primary mining areas.

The small multiples approach reveals relationships that were not apparent or were hidden in the earlier static 3D view. These relationships include the overall curviplanar geometry of the sulfide zones, the correspondence of the mineralization styles at Nova and Bollinger, the large area extent of the Lower Breccia compared to the more compact Bollinger Massive, and the fact that the Upper Breccia appears to connect the Nova and Bollinger zones at a higher level in the sulfide stratigraphy. Furthermore, the orientation and position of the Nova Gabbro, which is interpreted to be the source of the sulfide mineralization, confirms that the source intrusion is now on its side given the assumption that the sulfides have gravitationally settled into fracture systems below the source intrusion. Additionally, the geometry and spatial connectivity of the C5 Breccia and the Leucogabbro confirm the interpretation that this phase of mineralization is later than the other zones and is related to an upper-level intrusion. Close inspection and comparison of different zone styles are also instructive, with the single graphic providing abundant opportunities to draw both corresponding and contrasting observations between similar mineralization styles and between different types. These results can then be assessed and summarised in the main body of the MRE report, providing a reviewer with a clear framework of what to expect when physically reviewing the digital geological model.

My next article in this planned series will be about applying Tufte's principles to the preparation of graphics for sample quality control data.