Spatial organization and environmental niche modeling of Goguryeo archaeological sites in Northeast China: integrating GIS and machine learning approaches

The spatial organization of archaeological sites provides important evidence for understanding human–landscape interaction, territorial control, and regional adaptation in past societies. Goguryeo archaeological remains in Northeast China, including fortresses, tombs, and settlements, are distributed across mountain-valley landscapes where topography, hydrology, visibility, and movement corridors likely shaped site placement. However, the regional spatial logic of these remains has not been sufficiently examined through an integrated quantitative framework. This study analyzes 187 Goguryeo archaeological sites, including 73 fortresses, 89 tombs, and 25 settlements, by combining geographic information systems, spatial statistics, MaxEnt environmental niche modeling, site-type environmental comparison, and landscape connectivity analysis. The results indicate that Goguryeo sites were not evenly distributed across the study region but were concentrated in river-valley and mountain-corridor settings. Kernel density and nearest-neighbor analyses show a clustered distribution, with 47 sites located in the upper quartile of local density and a mean nearest-neighbor distance of 5.46 km. Ripley’s K-function identifies peak clustering at approximately 18.2 km, suggesting a local to sub-regional scale of spatial organization. The MaxEnt model achieved stable predictive performance, with training AUC = 0.89, testing AUC = 0.86, and spatial cross-validation AUC = 0.84. Elevation, river distance, and agricultural potential were the three most important predictors, together accounting for 63.7% of model contribution and 73.3% of permutation importance. Site-type comparison suggests functional differentiation: fortresses were generally located at higher elevations, on steeper slopes, in more positive topographic positions, and with larger viewsheds, whereas settlements were closer to rivers and had higher agricultural potential. Network analysis further indicates that fortress sites had higher mean betweenness centrality than tombs and settlements, suggesting their importance in modeled movement and territorial networks. A case study of Wunu Mountain illustrates how regional model outputs can be interpreted in relation to defensive visibility, accessibility, and surrounding support landscapes. The study provides a reproducible approach for interpreting Goguryeo archaeological landscapes and for identifying zones that may be prioritized in future survey and cultural heritage management, while also emphasizing the limitations caused by sample-size imbalance, discovery bias, and historical environmental change. The spatial organization of archaeological sites provides important evidence for understanding human–landscape interaction, territorial control, and regional adaptation in past societies. Goguryeo archaeological remains in Northeast China, including ... [2389 chars]