Reconstructing Nineteenth-Century River Water Levels with Transformer-Based Computer Vision

Malte Rehbein 

Nineteenth-century Bavarian Danube gauge charts (1826–1894) were converted into daily water-level series referenced to gauge zero through a semi-automated workflow combining light document pre-processing, dewarping, transformer-based line extraction, pixel-to-curve calibration, and targeted human checks. A curated ground-truth sample supported benchmarking and uncertainty quantification. On three representative gauges (Neu-Ulm, Vilshofen, Passau), the pipeline achieved high series-level accuracy while reducing manual effort by roughly an order of magnitude relative to full manual digitisation. Outputs include versioned datasets with page-level provenance, uncertainty flags, and code pointers to ensure transparency and reuse. The approach offers a replicable template for rescuing analogue hydrometric records and enabling long-term analyses of extremes, regulation impacts, and ecological context. Data are openly available under CC BY 4.0 (DOI: 10.5281/zenodo.17296751).