DISCUSSION

As our results show, the presented method for obtaining 3D models from sets of (ordinary) photographs seems to be well suited for palaeontological work. We note that the models obtained by the method are at least sufficiently accurate for the study of the buccal side facets of lower molars of rhinoceroses, although these facets are only about 1-5 mm wide. Hence it is clear that the method will be suitable for many other palaeontological studies as well. A clear benefit of the presented method is that research that requires 3D models can now be performed by almost anyone with a digital camera and a computer, without the need to access expensive and often difficult-to-use 3D scanners. In particular, we note that the simplicity and speed of photography means that some research, which has traditionally had to be done in situ at museums, often under severe time constraints, can with this technique be deferred to off-site investigations of the generated 3D models. We also note that it is, at least in principle, possible to use a digital video camera to even more conveniently obtain the image sequences by extracting individual frames from the recorded movie.

We found that with the generation of 3D reconstruction software we used, the most important aspect of obtaining good 3D models was to have sharp photographs. This is easy to understand when comparing the level of detail found out-of-focus regions of photographs to the level of details in regions that are in focus, as the latter obviously contains more information. In our case this meant that the out-of-focus parts of the photographs had to be masked out by hand. Undoubtedly, once the art of obtaining 3D models from 2D photographs matures, this step will become unnecessary.

We observe that the measured angle between the buccal side enamel edge and the buccal side surface is much larger in those molars with a facet than in those molars with no facets. Hence large average angles seem to generally suggest attrition, whereas small average angles generally suggest abrasion. When the average angles are compared to the upper molar mesowear of the same species, we observe that for browsing species, for which the cusps are on the average high and sharp, the angles are also on the average larger. On the other hand, small average angles seem to correlate well with the low and blunt cusps of grazers. Furthermore, the results obtained for mixed-feeders from the lower molars match those obtained from the upper molars.

There are, however, drawbacks to using the measured angles as attributes of wear. First of all, there are differences in the morphology of the buccal side surfaces in different species, but in this work the angles were measured with respect to these surfaces, with the assumption that the surfaces have a fixed direction. This, however, represents an oversimplification of reality since the curvature, or flatness, of the buccal side surface depends on the morphology of each species. To some extent it is possible to attribute convex buccal side surfaces to browsing species, whereas grazing species have a more vertical buccal side surface. This morphological difference leads to a slight exaggeration of the difference between the measured angles of browsers and grazers. Indeed, the convex buccal side surface found in browsers tends to lead to a slight increase in the measured angle, whereas the straight buccal side surface of grazers tends to slightly decrease the measured angle. It could be possible to eliminate this problem by using, instead of the buccal side surface, some other surface of a tooth, such as for example the growth-direction of the molar roots, as was done by Butler (1972) in a different context. However, since it is difficult to determine in what directions the molar roots grow when studying teeth still attached to the skull, we decided to exclusively use the buccal side surface as reference.

Although it is typically straightforward to consistently measure the enamel edge angle, in some cases it can happen that the facet or the enamel edge itself is convex. In such cases the measurement can be tried at a slightly different location, which usually solves the problem. Otherwise it may be necessary to apply experience to get a reasonable estimate of the angle. We have encountered such difficult cases mainly with teeth from mixed feeders, and the number of such cases is, in any case, so low that they should only slightly affect the obtained results. When we assessed the repeatability of the method by measuring the angle at the hypoconid tip of a single tooth at 10 separate occasions, we found that the difference between the largest and the smallest result deviated by only 2.8 degrees. Since the difference between the largest and the smallest angles of the measured teeth was over 80 degrees, we are therefore confident that, although the deviation when measuring a particular tooth might be larger than 2.8 degrees, the method yields angle measurements, which are sufficiently accurate for our purposes.

We have previously used lower molar mesowear, based on facet development, to study the palaeodiet in the Pliocene species Ceratotherium efficax (Dietrich, 1942) in the Laetoli region in Tanzania (Hernesniemi et al. 2011). In that study the lower molars were divided into three groups; those with no facets (corresponding to blunt upper molars), those with facets with fuzzy boundaries (corresponding to rounded upper molars), and those with facets with sharp boundaries (corresponding to sharp upper molars). In the present study we noticed that the fuzzy boundary is not a very reliable factor, and accordingly proceeded to simplify the classification into molars with facets and molars without. Such a classification does not allow direct identification with upper molar classification and must therefore be separately analysed. However, as Franz-Odentaal and Kaiser (2003) observed from ruminants, the data obtained from the mesowear analysis of the lower teeth tends to be more rounded than the correspondent data from upper teeth. Kaiser and Fortelius (2003) suggested that due to gravity the matter being chewed is in more contact with lower molars than with upper molars, which could explain the differences observed in the corresponding mesowears. Thus mixing upper and lower molars in the same analysis may be inappropriate in any case.

In this work we treated each fossil species as a single group, as all the studied dental fossils come from the fossil localities of Britain. Thus, the dietary regimes were studied at the species level, without a strict differentiation according to age or locality. However, as observed in the species Stephanorhinus hundsheimensis by Kahlke and Kaiser (2010), some intraspecific differentiation in dietary habits between populations are possible, especially since the Pleistocene palaeoenvironments were very variable.