Home – Home – Jornals – Geography and Natural Resources 2026 number 2

The article presents the results of identifying landscape patterns in the distribution of archaeological sites in Southwestern Tuva and constructing a predictive model of the locations of as-yet-unexplored objects based on a neural network. The analysis was carried out using geoinformation systems and a machine learning algorithm. For this purpose, a database of archaeological sites in Southwestern Tuva was compiled based on literary sources and the authors’ field research. Using a digital elevation model and available archaeological data, a geoinformation analysis was carried out and distribution diagrams for archaeological sites were created based on nine landscape features: absolute height, slope inclination, slope aspect, position relative to watercourses, height above the waterline of a nearby watercourse, solar radiation intensity for December and June, visibility of mountain peaks, and distance from mountain peaks. Based on the analysis of variance for known archaeological sites, a diagram of the importance of landscape features was constructed. The distance from the watercourse and the height above the waterline of the nearby watercourse played the greatest role in choosing the locations for religious structures. According to the resulting predictive model, religious sites can be found primarily along the western periphery of the Khemchik depression, along the river valleys flowing toward this depression, and along the river valleys in the southwest of the study area. The accuracy of the predictive model was approximately 85 %. The obtained data make it possible to assess the role of landscape features in the distribution of archaeological sites and provide an opportunity to search for new archaeological sites.