Examining the Relationship Between Tatara Iron Production and Grassland Distribution in Western Japan: An Open Geospatial Approach to Historical Landscape Analysis

Abstract. In western Japan, traditional tatara iron production has long been associated, both in historical narratives and in early ecological work, with the persistence of grassland landscapes. This view was stated by Itow (1962), but the inferred link has rarely been examined with spatially explicit data. We examine the spatial relationship between tatara iron production sites and grassland distribution in Tottori Prefecture using an integrated open-source geospatial workflow. The two data layers are temporally offset by design: tatara sites spanning from the medieval period to the mid-Meiji era, while the earliest spatially explicit municipality-level grassland record—the 1950 World Agricultural Census—captures the “landscape memory” of those activities roughly half a century after the cessation of large-scale tatara. Census tables were digitised using AI-assisted, sub-page table extraction. Tatara site locations (N = 562) were compiled from the Tottori Prefecture cultural-heritage WebGIS. Geology, basin structure, and DEM-derived terrain were processed with GeoPandas, rasterio, scipy, and rasterstats; OLS regression, model comparison, and DBSCAN clustering were implemented in statsmodels and scikit-learn. Tatara sites form well-defined production districts in the granitic Hino River basin, but grassland is most strongly predicted by elevation, with tatara density adding only modest marginal information. The count × slope interaction is significantly negative, indicating that the apparent tatara–grassland association weakens sharply in steep terrain. We interpret these patterns as evidence that both phenomena occupy the same mountain environmental niche, consistent with a niche-construction reading rather than a direct landscape-transformation interpretation.