Over 100 individual sauropod appendicular elements
from 10 different institutions were digitized using the Immersion Microscribe
three-dimensional point digitizer. (see Appendix
for procedure). The point
digitizer proved to be transportable, cost effective, and reliable. The input
device for the Microscribe used in this study was modified for use with
either a mouse or foot pedal. A Pentium II PC with a 400MHz processor and 128 MB
of RAM was connected to the digitizer via a serial connection cable (Figure
1).
The Rhinoceros modeling program (Version 1.0) was used to obtain digital
data, which was processed with Surfacer (Art Anderson, Virtual Surfaces). Clay was used for stabilizing
specimens and white paper correction fluid for marking registration points. All
animations were made using Discreets' 3D Studio Max.
Specimen selection is critical when digitizing fossil bones. For the purposes of this study, each element had to be complete, relatively uncrushed, identifiable to genus, and free from the effects of ontogeny. The six major limb elements (humerus, radius, ulna, femur, tibia, and fibula) were considered to be complete if five measurements, length (L), greatest proximal breadth (GP), least breadth (LB), greatest distal breadth (GD), and least circumference (LC) could be made on the bone. The girdle bones were considered complete if approximately 90% of their edges were intact. The degree of crushing in a given specimen was much more difficult to estimate because there are very few sauropod appendicular elements that are totally free of distortion. For this study, elements were assumed to be relatively uncrushed if the key features of the bone (i.e., muscle scars, trochanters, fenestrae, etc.) were clearly visibly, and no offset fractures were apparent.
Any study that involves composite skeletons must consider the effects of ontogeny in reconstructions. The effects of ontogeny on sauropod appendicular material have been addressed by Carpenter and MacIntosh (1994) and Wilhite (1999). Wilhite (1999) found that the limbs of Camarasaurus exhibit isometric growth patterns with very little evidence of allometry. Similarly, Carpenter and MacIntosh (1994) noted isometric growth patterns in the limbs of Apatosaurus. The current study involves only diplodocid and camarasaurid sauropods. Therefore, it is assumed that all appendicular elements digitized are free from the effects of ontogeny and can be scaled to match other elements when composite limbs are necessary.
Specimen size is also an important final consideration when digitizing. In this study, the Microscribe digitizer and Rhinoceros data capture program make it possible to scan any known vertebrate element regardless of size. However, the radius of the digitizer arm used for this study is 70 cm, and elements longer than about 2 m had to be digitized in four or more parts, doubling the time needed to digitize a single element. If access time to museum collections is a consideration, specimens that fit into the size range of the digitizer arm are preferable.
While a detailed description of the digitizing
procedure used here is given in Appendix I, an overview of the procedure is
necessary before the benefits and potential problems associated with it can be
addressed. The Rhinoceros program is capable of storing data from the Microscribe
point digitizer as points or as curves. When stored as points, data is in the
form of a point cloud which must be surfaced. Point cloud data would be a good
choice for very small bones since the entire surface can be sampled. However,
this method was impractical for the large (greater than 30 cm) specimens used in
this study because of the time required to sample the entire surface with the
digitizer. Instead the curve method was employed for this project (points were
used for registration of the two digitized halves). Using this method, a series
of curves is defined across the short axis of the bone at varying intervals
along its length (see Appendix). The intervals between curves varied because
areas of a bone with significant morphological information were sampled at
shorter intervals than those areas which lacked morphological features.
After one side of the bone was digitized, a surface was modeled over the curves using a process called lofting. The process of lofting a surface over the curves is explained in detail in Appendix I, but can be visualized as laying a sheet over an object. Even though the shape of the object is clearly visible, the sheet does not exactly conform to every detail of the object. If the sheet is pulled tightly over the object, however, it conforms much more closely to the shape of the object than if it were loosely draped over it. The LOFT function in Rhinoceros works in much the same way by giving the user multiple options which essentially drape the surface more loosely or more tightly over created curves depending on the settings. Figure 2 shows the digital model of AMNH 6114 lofted using several different settings. The different results obtained using the various settings make it clear that no single method will reproduce the actual surface of the original bone in perfect detail (see Figure 3 for photo of actual element).
After a surface has been lofted over the digitized
curves and found to satisfactorily represent the actual specimen, the specimen
should be turned over and the process repeated for the other side. Both halves
can then be combined using a number of programs (in this case, Imageware’s Surfacer).
The resulting three-dimensional solid can be used for morphological analyses.
Digitized data offers several advantages over other methods of obtaining three-dimensional digital data. The components can be assembled or disassembled rapidly (in less than 10 minutes) so that even brief visits to museum collections may afford an opportunity to digitize important specimens. The cost of digitizing versus other techniques, such as CT scanning, is minimal. The Microscribe digitizer is portable and fits into a case no larger than an average suitcase. Elements also can be digitized very rapidly. For example, with familiarity to the equipment and techniques, I was able to scan a 1 m sauropod humerus in about 15 minutes. Digitized models can be compared to the original element in real time (Figure 3). It is imperative to compare each scan with the original element to check for accuracy, and the ability to correct errors in minutes without redoing the complete scan is an advantage when many elements need to be digitized. Finally, digitized data files are much smaller than those produced by other scanning techniques, making the resulting digital files easy to manipulate even on a modest computer system.
Despite the numerous benefits of digitized data,
there are certain potential problems that may limit the technique’s
usefulness. Elements for this study were scanned as two halves and later registered
together. It can be difficult to assemble bones that were originally very
flat (e.g., scapulae) because there is very little distance between the two
surfaces, which causes difficulty in combining halves. A technique for which no
registration of two halves is necessary would be better for these elements (Bonnan,
personal comm., 2002). Another potential problem is that the model surface is
generated by interpolating between the curves generated using the digitizer, and
therefore, the surface cannot be an exact copy of the original. Some
morphological features, such as the rugosities on the ends of sauropod limb
elements, may not appear in the digital model (Figure
3). Fortunately, the large
size of sauropod limb elements helps to diminish the small inconsistencies
between the model and the original element because morphological features are
much larger.
However, digitizing inconsistencies would prove to be much more
critical for small elements (less than 30 cm) because even minor errors in the
digitizing process, such as holding the point of the digitizer arm a few
millimeters above the surface of a specimen, affect the final shape of the
model. These errors are minimized when digitizing large bones because the same
errors represent such a minor portion of the bone itself. If millimeter-scale
accuracy is important, an alternative technique such as laser scanning (Lyons et
al. 2000) or CT scanning (Ketcham and Carlson
2001) would be more appropriate.
Having considered the benefits and limitations of
three-dimensional digitized data, the applicability of the technique to
different types of studies can be considered. Many morphometric studies could be
greatly facilitated using this equipment. Previous studies relied on digitized
points from photographs (Chapman 1990;
Chapman and
Brett-Surman 1990). The same data points could easily be captured in
three-dimensional morphospace using the point digitizer.
Ontogenetic studies
could also be facilitated using three-dimensional digitizing techniques because
many specimens representing different ontogenetic stages could be digitized
quickly, and the three-dimensional solids generated could easily be scaled to
the same size and qualitative ontogentic differences noted (Figure
4).
Three-dimensional digitized elements are especially useful for modeling joint
articulations in functional studies of exceptionally large animals such as
sauropod dinosaurs (Figure 5). The relatively small data files allow for the
assembly of complete skeletons using a standard personal PC (Figure
6). Also,
digital models of large bones can be illustrated without parallax distortion and
in exactly the same orientation–making comparisons of morphological features
more clear than with traditional photography (Figure
7).