Submerged Landscape Modeling using a Virtual World Simulation

Understanding the character of the sea floor and the drowned paleolandscape it represents is essential for the discovery of ancient submerged sites. Yet predicting the actual location of sites has remained problematic. Most predictive models essentially superimpose extant terrestrial environments and patterns of archaeological site distribution onto a dried out sea floor. In so doing they frequently simplify and misrepresent the ancient landscape and its associated fauna and flora, which often has no modern analog. Such models are also limited by their focus on site distributions, rather than on the past cultural behavior that would produce archaeological traces. To address these limitations, submerged site research in the Great Lakes has developed a Virtual World (VW) model of the ancient landscape. The model has three core components: landscape reconstruction; modeling of caribou seasonal movements; and site location prediction. For landscape reconstruction, bathymetry and drainage maps form the base layer which is iteratively updated as new data on bottom characteristics and the environment are acquired. The second component employs AI methods to model the seasonal movement of caribou, the most significant prey species at the time. The third component generates probability values for the occurrence of differing archaeological site types by applying rules derived from ethnographic sources to the overlay of caribou movement and landscape features. As a dynamic model, the VW incorporates new information as research progresses. It also enables change in key variables, such as sea level, or resources distribution to be modeled. The VW system moves beyond static models to create a constantly-evolving representation of the submerged landscape of Lake Huron and its archaeological sites.