10th International Conference on Geomorphology
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Exploring the relations between archaeological/historical sites and topography in Japan and China by machine learning methods

Yuan Wang1 and Takashi Oguchi2,1
Yuan Wang and Takashi Oguchi
  • 1Graduate School of Frontier Sciences, The University of Tokyo, Kashiwa, Japan
  • 2Center for Spatial Information Science, The University of Tokyo, Kashiwa, Japan

The remains of archaeological and historical sites exist all over the world, and their existence provides the basis of research for many disciplines such as Archaeology, History, and Geography. The distribution of such sites is non-random since human behaviors are controlled by environmental conditions. People in the past depended much more on the natural environment than today because of the lack of advanced technology to cope with nature. Therefore, exploring the relationship between the locations of archaeological sites and surrounding environments is useful to understand why and how past people selected the location of sites. This research establishes a predictive model to determine factors affecting the distribution of the archaeological/historical sites in Japan and China and rank them according to the degree of relevance. Special attention is paid to the influence of topographic factors. The DEMs used for this study reflect the human transformation of land due to the construction of the archaeological/historical sites. To consider the geomorphological environment when the site location was selected, we vectorized the outline of each site, and data only outside of the polygon were used for analysis. Then the topographic condition around each archaeological site was evaluated based on buffer-ring analysis and zonal statistics. The results allowed us to construct a predictive model for the spatial distribution of the sites based on topographic conditions. The model utilizes the Attentional Factorization Machines (AFM) for pairwise factors exploration. Among the sites, 75% were used for model training, and 25% were used for performance tests. The resultant multivariate model suggests that topography and related hydrological conditions affected the site distribution, which is helpful to understand the interaction between ancient people and the environment.

How to cite: Wang, Y. and Oguchi, T.: Exploring the relations between archaeological/historical sites and topography in Japan and China by machine learning methods, 10th International Conference on Geomorphology, Coimbra, Portugal, 12–16 Sep 2022, ICG2022-499, https://doi.org/10.5194/icg2022-499, 2022.