Dynamic modification of cut marks by trampling: temporal assessment through the use of mixed-effect regressions and deep learning methods

Abstract

Listen

Translate article

Cut marks exposed to a short period of trampling were easily differentiated from trampling marks. Here, through the combined use of cross-validated logistic regression, mixed-effect regression models and computer vision deep learning methods, we approach the timing and modification rate of the microscopic variables that characterize cut marks. This study emphasizes the need to understand that bone surface modifications are dynamic taphonomic entities subjected to context-dependent morphing processes. Every depiction of a mark is just time-determined. Here, we show through experiments that trampling affected mostly the presence and internal configuration of microstriations, as well as the flaking and the morphology of the shoulder of the mark. The presence of overlapping striae and microabrasion were reported as signatures of trampling which did not affect cut mark morphology. Sandy sediment was found more abrasive than gravel sediment because of its potential to get inside the grooves of marks and modify their walls more intensively. The moderate abrasion process modeled here did not impact cut marks to the point of not being differentiated from other types of marks. A comparison with trampling marks successfully classified both types of modifications.