Reconstruction of the Solar Activity from the Catalogs of the Zurich Observatory

TL;DR

This paper presents a neural network-based method to digitize and reconstruct historical solar activity data from handwritten catalogs of the Zurich Observatory, linking early observations with later datasets and enhancing historical solar activity records.

Contribution

The study introduces a neural network approach for recognizing handwritten solar observation data, creating a comprehensive database that connects historical and modern solar activity records.

Findings

Successfully digitized Zurich Observatory sunspot data

Connected early and later solar observations into a unified database

Demonstrated the potential of machine learning for processing historical astronomical data

Abstract

Catalogs of the Zurich Observatory contain positional information on sunspots, prominences and faculae in late 19th and early 20th centuries. This database is given in handwritten tabular form and was not systematically analysed earlier. It is different from the sunspot number time series made in Zurich and was obtained with a larger telescope. We trained a neural-network model for handwritten text recognition and present the database of reconstructed coordinates. The database obtained connects the earlier observations by Sp\"orer with later programs of the 20th century and supplements the sunspot-group catalogs of the Royal Greenwich Observatory. We also expect that the presented machine-learning approach and its deep capabilities will motivate the processing of a wide bulk of astronomical data, which is still given in non-digitized form or as plain scanned images.