Daejeon, July 27–31
Between the 8th and 13th centuries, the Christianisation of East-Central Europe produced a dense network of predominantly rural churches. Beyond their liturgical role, these buildings functioned as instruments of governance and social organisation within largely commoner populations (Vargha/Eichert 2025). Churches were consecrated to patron saints whose relics, when present, were deposited in the altar and expressed through architecture, imagery, and spatial symbolism. Such materialisations of sanctity conveyed authority and meaning to largely illiterate communities and may have influenced architectural design.
A long-standing hypothesis suggests that churches were intentionally oriented toward the sunrise on the feast day of their patron saint or, alternatively, toward equinoctial east (McCluskey 2015). Numerous regional architectural and archaeoastronomical studies have addressed this idea (Eichert 2012; Sassin Allen 2015; Abril/Sánchez Velasco 2023), yet results remain inconclusive. While some churches show plausible correlations between axis orientation and solar events, many do not, raising questions of intentionality, methodological consistency, and regional variability.
Large-scale analyses face persistent challenges. Precise orientation data often require labor-intensive field measurements of azimuths and horizon profiles, while historical interpretation depends on detailed knowledge of building phases, patronage histories, and ecclesiastical contexts. Automated approaches, by contrast, risk producing large but misleading datasets if topography or calendrical change is ignored. This study seeks to reconcile scale and precision through a reproducible, data-driven workflow applied to approximately 1000 well-documented medieval church sites in present-day Austria and the Czech Republic.
The workflow integrates archaeological footprints, digital elevation models (DEMs), historical Julian calendar dates, and documented patronal feast days. Church axes are extracted computationally, sunrise directions are modelled with corrections for local topography and calendrical drift, and effective sunrise dates are compared with multiple feast days per patron saint. The aim is to reassess correlations between church orientation and patronal or equinoctial sunrises and to identify spatial, chronological, and ecclesiastical patterns across the region. A further objective is methodological transparency through the use of open data and open-source tools.
A central issue in assessing astronomical alignments is the cumulative drift of the Julian calendar, whose year length exceeds the solar year by roughly 11 minutes. As a result, sunrise on a nominal feast day in the 10th century occurred on a different astronomical date than today. All calculations are therefore calibrated to historically correct Julian dates corresponding to the estimated construction centuries of the churches.
Local topography constitutes a second major challenge. The visible point of sunrise depends strongly on horizon relief: in valleys or mountainous settings, the sun appears at a different azimuth than on flat terrain even for identically oriented buildings. Horizon profiles are therefore derived from DEMs along each church’s nave–apse axis to calculate effective sunrise azimuths that account for topographical obstruction (Čaval 2009; Hinton 2012).
Patron saints often possess multiple feast days, including principal feasts, translations of relics, and regional commemorations (Grotefend 1892–1899). Owing to the symmetry of the solar path around the solstices, identical sunrise azimuths also occur twice per year. This study therefore considers all documented feast days per patron and evaluates both pre- and post-solstitial calendar candidates. Further uncertainties include undocumented changes in patronage, rebuilding or reorientation of churches, ambiguities regarding the intended moment of sunrise, and possible westward or sunset-oriented alignments.
The dataset comprises polygon geometries of standing structures, excavated foundations, and georeferenced archaeological plans of approximately 1000 medieval churches in Austria and the Czech Republic dated between the 8th and 13th centuries. Only sites with sufficiently constrained chronologies and securely identified patron saints were included, together with their diocesan or political affiliations. The data derive from archaeological, architectural, historical, and heritage documentation and are digitally accessible via the OpenAtlas database of the THANADOS network (Eichert 2021; Richards et al. 2023; Eichert et al. 2024).
Open DEMs, including SRTM and EU-DEM, are employed for horizon modelling, with multiple resolutions tested to balance accuracy and computational efficiency. Feast-day data are primarily drawn from Grotefend (1891–1899) and validated against digital scholarly resources (Manuscripta Mediaevalia).
Principal component analysis (PCA) and minimum bounding rectangles (MBR) have been tested for estimating the dominant longitudinal church axis from nave to apse (Gewers et al. 2022). However, for more complex polygons, a custom method based on calculating the maximum inscribed rectangle (MIR) was developed to reconstruct this axis and yielded more accurate results. Results are compared and manually reviewed to correct outliers, particularly in asymmetrical or partially preserved buildings.
For sunrise modelling, each church axis is extended toward the horizon and a DEM-based elevation profile is calculated. The point at which terrain first intersects the line of sight defines the effective horizon angle. Sunrise azimuths corresponding to this horizon are then computed for historically correct Julian dates using established astronomical algorithms (Reda/Andreas 2008).
Calculated sunrise dates are compared with all documented feast days of the respective patron saint. Potential correspondences are evaluated using two continuous measures: angular deviation between church axis and sunrise azimuth, and temporal distance between sunrise date and feast day. Statistical analysis examines alignments with patronal or equinoctial sunrises, deviations from random distributions, regional differences, and chronological trends. Given the circular nature of the data, Rayleigh, Kuiper, and Watson tests are applied (Fisher 1993), complemented by regression modelling and Monte Carlo simulations. Selected cases are validated through targeted field inspections.
This study presents a reproducible digital workflow that combines DEM-based horizon modelling, historical calendrical correction, and semantically rich archaeological data. It offers a critical reassessment of hypotheses concerning astronomical or liturgical intentions in medieval church construction, highlighting both correlations and their limitations. The analysis reveals spatial and temporal variability that points to the influence of regional ecclesiastical structures, political contexts, and landscape settings rather than a universal rule of orientation.
More broadly, the project demonstrates how comparatively small but well-contextualised datasets can be analysed using digital methods that integrate qualitative historical knowledge with quantitative modelling. It thus contributes to ongoing debates in medieval studies and the Digital Humanities on interpretative modelling, methodological transparency, and the responsible scaling of historical data analysis.